EV Charging Planning Models:Some

Investigations

Ph.D. Thesis

SHAKTI VASHISTH
ID No. 2018REE9088

DEPARTMENT OF ELECTRICAL ENGINEERING

MALAVIYA NATIONAL INSTITUTE OF TECHNOLOGY JAIPUR

February 2024
EV Charging Planning Models :Some
Invstigations

Submitted in
partial fulfillment of the requirements for the degree of
Doctor of Philosophy

by

Vipin Chandra Pandey

ID: 2018REE9071

Under the Supervision of

Dr. Praveen Kumar Agrawal

DEPARTMENT OF ELECTRICAL ENGINEERING

MALAVIYA NATIONAL INSTITUTE OF TECHNOLOGY JAIPUR

February 2024
 iii

O
c Malaviya National Institute of Technology Jaipur - 2024

All Rights Reserved

iv
 Declaration

I, Shakti Vashisth declare that this thesis titled, “EV charging Planning Models:
Some Investigations” and work presented in it, is my own. I confirm that:

ˆ This work was done wholly or mainly while in candidature for a research degree at
this university.

ˆ Where any part of this thesis has previously been submitted for a degree or any
other qualification at this university or any other institution, this has been clearly
stated.

ˆ Where I have consulted the published work of others, this is always clearly attributed.

ˆ Where I have quoted from the work of others, the source is always given. With the
exception of such quotations, this thesis is entirely my own work.

ˆ I have acknowledged all main sources of help.

ˆ Where the thesis is based on work done by myself, jointly with others, I have made
clear exactly what was done by others and what I have contributed myself.

Date: 4 February 2024

Shakti Vashisth
(2018REE9088)

v
vi
 Certificate

This is to certify that the thesis entitled “EV charging Planning Models: Some In-
vestigations” being submitted by Shakti Vashisth (ID: 2018REE9088) is a bonafide
research work carried out under my supervision and guidance in partial fulfillment of the
requirement for the award of the degree of Doctor of Philosophy in the Department
of Electrical Engineering, Malaviya National Institute of Technology, Jaipur, India. The
matter embodied in this thesis is original and has not been submitted to any other Uni-
versity or Institute for the award of any other degree.

Dr. Praveen Kumar Agrawal

Place: Jaipur
Associate Professor
Date: 4 February 2024 Department of Electrical Engineering
Malaviya National Institute of Technology Jaipur

vii
viii
 The thesis is dedicated to

My Grand parents, parents, and teachers

ix
 Acknowledgements
The research is a virtue of motivation, knowledge, innovations, perseverance, resilience,
encouragement and discourse of individuals. These instilled me to pursue the journey of
my Ph.D. work. I would like to express my sincere acknowledgments and thankfulness to
those who have been part of this incredible journey directly or indirectly.

First and foremost, I take this opportunity to express sincere gratitude to my supervisor
Dr. Praveen Kumar Agrawal for giving me the opportunity to pursue Ph.D. under his
guidance. I am immensely beholden to him for his faith in my competence and enriching
with his profound knowledge, tenacious and amicable support during the progression of
this thesis. His passionate discussion on the research subjects helped me to explore as a
researcher and to shape my Ph.D. research work. His advises on the research as well as
on life, has infused and evolved my conviction for which, I will remain indebted to him
forever.

I would like to thank my DREC members: Prof. K. R. Niazi , Dr. Anil Swarnkar
and Prof. Nikhil Gupta for their encouraging support, guidance, and invaluable inputs
during the course of research work. My special thanks to Prof. K. R. Niazi for his
inspiring vision and idea to improve and enhance my research work.

I owe my sincere gratitude to Prof. Narayana Prasad Padhy , Director, MNIT, for
extending all kind of infrastructural facilities required for pursuing my Ph.D. My warm
thanks to Prof. Harpal Tiwari , Head of the Department and Dr. Satya Narayan
Neeli , Convener, DPGC, Electrical Department, for encouraging me in all possible man-
ners during the course of my Ph.D.

I am also grateful to my seniors, friends and fellows during my time at MNIT Jaipur. I
thank to Dr. N. K. Meena, Dr. R. K. Thokar, Dr. T. Rawat, Dr. P. Singh, Dr. J. Singh,
Dr. S. Singh, Dr. J. P. Keshri, Bhuwan, Shakti, Vikram, Yogesh, Gurpinder, Krishna,
Hitesh, Abhishek, Tanmay, Prahalad, Gaurav, Manisha, Yashvi and other fellows. They
made my research work and life eventful and enjoyable during my time at MNIT. I will
always cherished the moments spent together. Special mention to my friends, Shivang ,
Uday and Ankur for their continuous encouragement, emotional and moral support.

Finally, I convey my regards to my family members for their constant support and encour-
agement. I am indebted to my father, Mr. Puran Chandra Pandey , mother, Mrs.
Kamla Pandey , elder brother, Manoj , sisters, Meera and Neema for their patience,
emotional support, and understanding during the course of my Ph.D. work.
xi
 xii

At last, thanks to all friends and fellows whom I haven’t mentioned here.

(Shakti Vashisth)
 Abstract
The transition of deregulation, competition, technological advancement, and information
communication technology in the electrical power hierarchy has brought several devel-
opments in the contemporary distribution systems (DS). It has made the contemporary
DS overwhelmed with retail competition, distributed generation, energy storage systems
(ESS), renewable generations, technological based smart loads, advanced metering infras-
tructure, bi-directional communication, etc. These innovations have also transformed the
customers’ perspective from passive to active in addition to DS. The expansion of demand
response (DR) is one of eminent proposition out of it into DS. It is result of persistent
issues such as sudden peak price, price volatility, variable generations, etc. DR is accred-
ited to have ample potential theoretically, and well supported by technology. However,
its implementation is less sufficed in DS. This requires a new, improved modeling and
operational strategies in the viewpoint of contemporary DS. Moreover, the customers’
pragmatic behavioral attributes and the characteristics of DS need to be considered under
DR environment.
This thesis presents DR modeling and its operational strategies in the contemporary DS.
It proposes the various DR modeling approaches considering the customers’ behavioral
attributes in addition to DS characteristics. The proposed DR models are adapted from
the consumer theory, microeconomics, and are suitably re-designed for a pragmatic DR
representation. DR models are developed using the price elasticity model (PEM) and
utility functions (UFs). PEM is improved in price-based DR (PBDR) using the concept of
adaptive elasticity to imitate the customers’ behaviour in DR. Its extension in incentive-
based DR (IBDR) model is addressed by proposing an analytical method to extract the
exact effect of incentives on the elasticity variations. The UFs, cardinal UFs (CUFs)
and ordinal UF (OUF) are developed for PBDR models. Both UFs present the compre-
hensive PBDR models considering the customers’ behavioral attributes such willingness,
risk-behaviour, adaptability, and load recovery. The proposed DR modeling approaches
are performed and investigated using a disaggregated load modeling approach. More-
over, DR operational strategies are developed considering DS characteristics such as the
different customer classes, price cross-subsidy, load patterns, and load diversity, etc. It
is proposed for both (PB and IB) frameworks with the different stakeholders under DR
environment. The strategic interactions among the various stakeholders such as load serv-
ing entity (LSE), demand response providers (DRPs) and customers are devised using
the game theory. Further, the operational frameworks are optimized and solved using a
customized and improved optimization approaches. The proposed frameworks and the
 xiv

solution algorithms are applied on the standard and practical distribution systems con-
sidering ac network constraints in addition to pragmatic DR and DS’ attributes. The
application results are thoroughly investigated and presented from the system and the
customers’ perspectives.
Contents

Certificate vii

Acknowledgements xi

Abstract xiii

Contents xv

List of Tables xix

List of Figures xxi

List of Algorithms xxiii

Abbreviations xxiv

Symbols xxix

1 Introduction 1

2 Literature Review 7
2.1 PBDR and IBDR Modeling using PEM . . . . . . . . . . . . . . . . . . . . 8
2.2 PBDR Modeling using Utility Functions . . . . . . . . . . . . . . . . . . . . 9
2.3 Price Based Demand Response . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4 Incentive Based Demand Response . . . . . . . . . . . . . . . . . . . . . . . 13
2.5 Critical Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.6 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3 Price and Incentive Based Demand Response using Price Elasticity Model
21
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Proposed Price Based DR Framework . . . . . . . . . . . . . . . . . . . . . 22

xv
Contents xvi

3.2.1 Proposed Deterministic PEM (DPEM) . . . . . . . . . . . . . . . . 24

3.2.2 Proposed Stochastic PEM (SPEM) . . . . . . . . . . . . . . . . . . 26
3.3 Disaggregated Load Profiling . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3.1 DR Assessment Index . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.4 Simulation Results and Discussions . . . . . . . . . . . . . . . . . . . . . . . 29
3.5 Proposed Incentive-based DR using PEM . . . . . . . . . . . . . . . . . . . 35
3.5.1 Proposed IBDR Model with Dynamic Rate Charging . . . . . . . . . 37
3.5.2 Proposed IBDR Model with Modified Flat Rate Charging . . . . . . 39
3.6 Economic Assessment of Utility and Customers under DR . . . . . . . . . . 40
3.7 Simulation Results and Discussions . . . . . . . . . . . . . . . . . . . . . . . 42
3.7.1 Difference between the Existing and the Proposed IBDR Modeling
Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4 Price Based Demand Response using Cardinal and Ordinal Utility Func-
tions 53
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.2 Proposed Constrained Economic PBDR Model using CUFs . . . . . . . . . 54
4.2.1 Quadratic CUF Modeling . . . . . . . . . . . . . . . . . . . . . . . . 56
4.2.2 Exponential CUF Modeling . . . . . . . . . . . . . . . . . . . . . . 58
4.2.3 Logarithmic CUF Modeling . . . . . . . . . . . . . . . . . . . . . . 60
4.2.4 Power CUF modeling . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2.5 Modeling of Customer’s Risk Behaviour . . . . . . . . . . . . . . . . 62
4.3 Simulation Results and Discussions . . . . . . . . . . . . . . . . . . . . . . . 64
4.4 Proposed Overlapping Generation based PBDR Model using OUF . . . . . 70
4.4.1 Load Recovery in DR using Altruistic Behaviour . . . . . . . . . . . 72
4.4.2 Modeling of Substitution Effect between Products . . . . . . . . . . 75
4.5 Proposed Fuzzy Framework for Customer Participation in DR . . . . . . . 76
4.6 Procedure of PBDR implementation using OLG model . . . . . . . . . . . 77
4.7 Simulation Results and Discussions . . . . . . . . . . . . . . . . . . . . . . . 78
4.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

5 Hierarchical Price Based Demand Response Framework in Distribution

systems 87
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.2 Proposed Hierarchical PBDR Framework . . . . . . . . . . . . . . . . . . . 88
5.2.1 Upper-level Problem: LSE’s Profit . . . . . . . . . . . . . . . . . . . 89
5.2.2 First Lower-Level Problem: (DRP k) Model . . . . . . . . . . . . . . 90
5.2.3 Second Lower-Level problem: (Customer i) Model . . . . . . . . . . 91
5.3 Stackelberg Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.4 Solution Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.4.1 Slave Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.4.1.1 Inner-Master Problem . . . . . . . . . . . . . . . . . . . . . 99
5.4.1.2 Inner-Slave Problem . . . . . . . . . . . . . . . . . . . . . . 99
5.5 Simulation Results and Discussions . . . . . . . . . . . . . . . . . . . . . . 101
Contents xvii

5.5.1 Standard 33-bus test system . . . . . . . . . . . . . . . . . . . . . . . 102

5.5.1.1 Effect of PBDR on Class-wise Load Pattern . . . . . . . . 102
5.5.1.2 Effect of PBDR on Network Operation Conditions . . . . 104
5.5.1.3 Effect of PBDR on the Economic Prospects of the Different
Entities . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.5.1.4 Effect of PBDR on the Customer Bill and LSE Profit under
Different Pricing Strategies . . . . . . . . . . . . . . . . . . 105
5.5.1.5 Effect of PBDR on Economic Performance . . . . . . . . . 106
5.5.2 108-bus Indian Distribution System . . . . . . . . . . . . . . . . . . 107
5.5.2.1 Effect of PBDR on Class-wise Load Pattern . . . . . . . . 107
5.5.2.2 Effect of PBDR on Network Operation Conditions . . . . 107
5.5.2.3 Effect of DR on Economic Performance . . . . . . . . . . . 107
5.5.3 Comparison of the Proposed Algorithm . . . . . . . . . . . . . . . . 109
5.5.4 Computational Issues . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

6 Bi-level Incentive Based Demand Response Framework in Distribution

Systems 113
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.2 Proposed Bi-level IBDR Framework . . . . . . . . . . . . . . . . . . . . . . 114
6.2.1 Upper-level Problem: LSE’s Operating Cost . . . . . . . . . . . . . . 115
6.2.2 Lower-level Problem: (DRP k) Model . . . . . . . . . . . . . . . . . 116
6.2.3 Equilibrium Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6.2.3.1 Generalized Stackelberg Game . . . . . . . . . . . . . . . . 118
6.2.3.2 Existence of GSE . . . . . . . . . . . . . . . . . . . . . . . 119
6.2.3.3 Uniqueness of GSE . . . . . . . . . . . . . . . . . . . . . . 120
6.3 Solution Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
6.4 Simulation Results and Discussions . . . . . . . . . . . . . . . . . . . . . . 125
6.4.1 Standard 33-bus Test System . . . . . . . . . . . . . . . . . . . . . . 126
6.4.1.1 An aggregated load pattern ADR and DR contribution . . 127
6.4.1.2 Effect of IBDR on Network Operation Conditions . . . . . 128
6.4.1.3 Effect of IBDR on Economic performance . . . . . . . . . 128
6.4.2 Indian 108-bus system . . . . . . . . . . . . . . . . . . . . . . . . . . 130
6.4.2.1 An aggregated Load Pattern ADR and DR Contribution . 131
6.4.2.2 Effect of IBDR on Network Operation Conditions . . . . . 132
6.4.2.3 Effect of IBDR on Economic Performance . . . . . . . . . 132
6.4.3 Comparison of the methods . . . . . . . . . . . . . . . . . . . . . . . 134
6.4.4 Computational Issues . . . . . . . . . . . . . . . . . . . . . . . . . . 135
6.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135

7 Conclusions and Future Scope 137

7.1 Salient Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.2 Future Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
Contents xviii

A Test Distribution Systems 145

A.1 Study system 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
A.2 Study System 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

B Customer Classes and Technical Information 151

Bibliography 153

Publications 165

Brief bio-data 167

List of Tables

3.1 Aggregated customer class bills under the considered states using the pro-
posed DPEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.2 Aggregated class-wise customer electricity bills using the proposed DPEM . 34
3.3 Aggregated class-wise customer electricity bills using the proposed SPEM . 34
3.4 Case considered in PBDR and IBDR study . . . . . . . . . . . . . . . . . . 44
3.5 Aggregated customer class bills under the different cases . . . . . . . . . . . 47
3.6 Aggregated class-wise utility’s profit under the different cases . . . . . . . . 47
3.7 Elasticity evaluation using the standard and proposed IBDR model at 20:00
Hour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.8 Aggregated impact on peak demand curtailment using the standard and the
proposed approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.9 An economic comparison of the standard and the proposed augmented DR
during peak periods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4.1 ARA and RRA indices for the different CUFs . . . . . . . . . . . . . . . . . 63

4.2 Aggregated customer class bills under the different CUFs based PBDR models 69
4.3 Utility profit under the different CUFs based PBDR models . . . . . . . . . 69
4.4 FIS Evaluation based on the customer demand and flexible demand . . . . 77
4.5 Aggregated customer class bills under different PBDR Models . . . . . . . . 84
4.6 Utility profit under the different PBDR models . . . . . . . . . . . . . . . . 84

5.1 Economic performance on the various pricing methods . . . . . . . . . . . . 106

5.2 Class-wise bills BDR and ADR using the proposed pricing strategy . . . . . 107
5.3 Economic performance on the various pricing methods . . . . . . . . . . . . 109
5.4 Class-wise bills BDR and ADR using the proposed pricing strategy . . . . . 109
5.5 Comparison of methods on 33 and 108-bus systems . . . . . . . . . . . . . . 109
5.6 Computational comparison of the methods . . . . . . . . . . . . . . . . . . . 110

6.1 Schedule of class-wise incentive rates and induced flexible demand in the
different peak periods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
6.2 Results in the different peak periods BDR and ADR . . . . . . . . . . . . . 127
6.3 Optimization results of the system for a day . . . . . . . . . . . . . . . . . . 129
6.4 Class-wise bills BDR and ADR using the proposed pricing strategy. . . . . 130
6.5 Schedules of class-wise incentive rates during the different peak periods . . 130
6.6 Schedules of induced demand during the different peak periods . . . . . . . 131
6.7 Results during the different peak periods BDR and ADR . . . . . . . . . . 131

xix
List of Tables xx

6.8 Optimization results of the system for a day . . . . . . . . . . . . . . . . . . 133

6.9 Class-wise bills BDR and ADR . . . . . . . . . . . . . . . . . . . . . . . . . 133
6.10 Incentive rates and induced demand at 20th hour on 33-bus test system . . 134
6.11 Economic comparison for a day using both methods . . . . . . . . . . . . . 134
6.12 Computational comparison of the methods . . . . . . . . . . . . . . . . . . . 135

A.1 Line and load data of the standard 33-bus test distribution system . . . . . 146
A.2 Line and load data of Indian 108-bus distribution system . . . . . . . . . . 147

B.1 Customer classes information used in 33-bus distribution system . . . . . . 151

B.2 Customer classes information used in 108-bus distribution system . . . . . . 151
B.3 Nominal load factor for the various customer classes . . . . . . . . . . . . . 152
B.4 Techno-economic parameter used in the study . . . . . . . . . . . . . . . . . 152
List of Figures

3.1 (a) A sample of load profiles of R Class customers; (b) dynamic price profile
under RTP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.2 A aggregated class-wise elasticity variation using the proposed PEM . . . . 31
3.3 Aggregated class-wise load patterns BDR and ADR using PEM, and the
proposed PEM models for: (a) R, (b) LI, (c) MI, (d) C, (e) A . . . . . . . . 32
3.4 Class-wise curtailed and shifted demand under RTP using (a) PEM, (b)
DPEM, and (c) SPEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.5 Effect of DR on MCP and utility price using PEM . . . . . . . . . . . . . . 42
3.6 Market price, TOU and RTP . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.7 (a) An aggregated price and incentive elasticity variations, (b) Class-wise
demand flexibility under the different cases . . . . . . . . . . . . . . . . . . 45
3.8 Aggregated class-wise load patterns before and after DR under the various
cases using DPEM for: (a) R (b) LI, (c) MI, (d) C, (e) A . . . . . . . . . . 46
3.9 Utility Price variation with DR application under the different cases . . . . 48
3.10 (a) Aggregated Contribution of PBDR and IBDR programs under the dif-
ferent cases; (b) Variation in incentive elasticity with variation in relative
ratio of incentive/change in price . . . . . . . . . . . . . . . . . . . . . . . . 48

4.1 Illustration of a quadratic utility function for: (a) customers’ utility values
(b) customers’ marginal utilities . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2 Illustrations of: (a) the various utility functions, (b) marginal utility curves 55
4.3 Customer’s behaviour under risk . . . . . . . . . . . . . . . . . . . . . . . . 62
4.4 Dynamic price profile under RTP . . . . . . . . . . . . . . . . . . . . . . . . 64
4.5 Variation in (a) ARA; (b) RRA with the change in β for the different CUFs 65
4.6 Variation in (a) ARA; (b) RRA with increasing demand for R customer class 66
4.7 R class customer participation in DR under (a) Quadratic, (b) Exponential,
(c) Logarithmic, (d) Power CUF . . . . . . . . . . . . . . . . . . . . . . . . 66
4.8 Aggregated class-wise load patterns BDR and ADR using the various CUFs
for: (a) R, (b) LI, (c) MI, (d) C, (e) A . . . . . . . . . . . . . . . . . . . . . 67
4.9 (a) Demand flexibility under the various customer classes using the different
CUFs; (b) Aggregated load factor . . . . . . . . . . . . . . . . . . . . . . . . 68
4.10 Average demand flexibility with variation in : (a) Satisfaction coefficient;
(b) price . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.11 (a) Three-tier TOU price framework, (b) A sample of willingness or control
parameter variation for demand level 2 kW . . . . . . . . . . . . . . . . . . 79

xxi
List of Figures xxii

4.12 Customer-wise induced demand flexibility with variation in willingness pa-

rameter for (a) R, (b) LI, (c) MI, (d) C, (e) A . . . . . . . . . . . . . . . . . 80
4.13 Aggregated class-wise load patterns BDR and ADR under the different
PBDR models for: (a) R, (b) LI, (c) MI, (d) C, (e) A . . . . . . . . . . . . 81
4.14 Demand flexibility using (a) PEM; (b) CES; (c) OLG; (d) Proposed OLG
model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.15 (a) Aggregated load pattern, (b) load factor, and relative ratios under the
different DR models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.16 Comparison of curtailment/shifting under the various PBDR models using
(a) Real-time pricing (RTP), (b) Time-of-use (TOU) . . . . . . . . . . . . . 85

5.1 Hierarchical architecture of the proposed PBDR framework . . . . . . . . . 92

5.2 Pricing schedule of LSE and DRP for flexible loads . . . . . . . . . . . . . . 102
5.3 Aggregated class-wise load patterns BDR and ADR of (a) R, (b) LI, (c)
MI, (d) C, (e) A, (f) Class wise curtailment and shifting . . . . . . . . . . . 103
5.4 Effect of DR on (a) System’ load factor and (b) Active power loss . . . . . 104
5.5 Effect of DRF variation on (a) LSE profit; (b) DRP profit, and (c) cus-
tomers’ bills . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.6 Effect of DRF on the total billing and LSE profit under RTP and the pro-
posed pricing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
5.7 Aggregated class-wise load patterns BDR and ADR of (a) R, (b) LI, (c)
MI, (d) C, (e) A, (f) Class wise curtailment and shifting . . . . . . . . . . 108
5.8 Effect of DR on (a) System’s load factor and (b) Active power loss . . . . . 109

6.1 Bi-level Structure of the Proposed IBDR Framework . . . . . . . . . . . . . 115

6.2 RTP profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.3 (a) Load pattern before and after DR and (b) Class-wise demand contribution128
6.4 Effect of DR on (a) system’s load factor and (b) Active power loss (c)
Voltage variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
6.5 (a) Load pattern after DR and (b) Class-wise demand contribution . . . . . 132
6.6 Effect of DR on (a) System’s load factor and (b) Active power loss (c)
Voltage variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

A.1 33-bus distribution system . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

A.2 Indian 108-bus distribution system . . . . . . . . . . . . . . . . . . . . . . . 147
List of Algorithms

4.1 Procedure for the implementation of DR using OLG model . . . . . . . . . 77

5.1 Procedure of Nested R&D algorithm for tri-level problem . . . . . . . . . . 101

6.1 Procedure of diagonalization method based on Gauss-Seidel approach . . . 125

xxiii
Abbreviations

DSO Distribution System Operator

RDG Renewable Distributed Generation
EV Electric Vehicle
ESS Energy Storage System
DR Demand Response
ICT Information Communication Technology
DOE Department Of Energy
PB Price-Based
IB Incentive-Based
PBDR Price-Based Demand Response
IBDR Incentive-Based Demand Response
RTP Real Time Price
TOU Time-of-Use
PTR Peak Time Rebate
IBR Inclining Blocking Rate
CPP Critical Peak Pricing
DLC Direct Load Control
IC/S Interruptible/ Curtailable Service
DB Demand Bidding
ASDR Ancillary Based Demand Response
PEM Price Elasticity Matrix/Model
UF Utility Function
CUF Cardinal Utility Function

xxv
Abbreviations xxvi

OUF Ordinal Utility Function

DS Distribution Systems
IEEE Institute of Electrical and Electronics Engineers
DER Distributed Energy Resources
DisCo Distribution Company
BDR Before Demand Response
NDR No Demand Response
ADR After Demand Response
HP Hourly Pricing
PHEV Plug-in Hybrid Electric Vehicle
CRRA Constant Relative Risk Aversion
CES Constant Elasticity of Substitution
OLG Overlapping Generation
AMI Advanced Metering Infrastructure
DRA Demand Response Aggregator
DRP Demand Response Provider
LA Load Aggregator
LSE Load Serving Entity
SPV Solar Photo-Voltaic
WP Wind Power
GT Gas-Turbine
DG Distributed Generation
PS Peak Saver
SG Stackelberg Game
UL Upper-Level
LL Lower-Level
EUC Electricity Utility Company
SE Stackelberg Equilibrium
KKT Karush-Kuhn-Tucker
B&B Branch & Bound
MPEC Mathematical Programming of Equilibrium Constraints
Abbreviations xxvii

EPEC Equilibrium Problems of Equilibrium Constraints

R&D Reformulation & Decomposition
CCG Column Constraint Generations
DN Distribution Network
GBM Geometric Brownian Motion
DF Demand Flexibility
LF Load Factor
COV Coefficient of Variation
PDF Probability Distribution Function
CDF Cumulative Distribution Function
S/CF Subsidy/ Charging Factor
VPP Variable Peak Pricing
MCP Market Clearing Price
FR Flat Rate
LDI Load Deviation Index
PDI Price Deviation Index
ARA Absolute Risk Aversion
RRA Relative Risk Aversion
MRS Marginal Rate of Substitution
NLP Nonlinear Programming
IPOPT Interior Point Optimizer
OPF Optimal Power Flow
GSG Generalized Stackelberg Game
VI Variational Inequalities
Symbols

Indices/sets
t, τ, ΩT Indices and set of time
k, ΩK Index and set of customer classes/DRPs
i, , ΩI Index and set of customers
n, m, ΩN Indices and set of bus
(n, m), ΩL Index and set of distribution lines
Parameters
dki,t,o Demand of ith customer class k at time t before DR (kW)
ρkt,o /ρk,F
t
R
Nominal/flat rate offered to class k at time t before DR (¢/kWh)
εk,P
i,t
E
Price elasticity of customer i, class k at time t
µP E /(σ P E )2 Mean/standard deviation of stochastic cross-price elasticity
µk,LF
t /(σtk,LF )2 Mean/standard deviation of load factor, class k at time t
COVk,LF
t Coefficient of variation of load factor, class k at time t
k
LFi,t Load factor of ith customer, class k at time t (p.u.)
ρk,IN
t
C
Incentive rate to class k at time t (¢/kWh)
ρk,IN
t,V P
C
Incentive rate to class k at time t
of voluntary IBDR (¢/kWh)
ρk,IN
t,M P
C
Incentive rate to class k at time t
of mandatory IBDR (¢/kWh)
ζt /δt Binary variable associated with voluntary and mandatory IBDR
λk,IN C /λk,P EN Incentive/penalty rate controlling parameter of class k
∆εki Change in elasticity of ith customer, class k
εk,IE
i Incentive elasticity of ith customer, class k

xxix
Symbols xxx

′
k
ρi,o Modified flat rate of class k at time t (¢/kWh)
εM /εU /εAGG Market/utility/aggregated elasticity
ρM U
t,o /ρt,o Market price/utility price at time t before DR (¢/kWh)
ω/β Satisfaction/risk-aversion coefficient
k /ω k
wi,t Satisfaction/modified coefficient of ith customer, class k at time t
i,t

dbki,o Average demand of ith customer, class k before DR (kW)

γik /χki LDI/PDI coefficients of customer i, class k
k
βi,t Risk-aversion coefficient of customer i, class k at time t (in CUFs)
k
θi,t Risk-aversion coefficient of customer i, class k at time t (in OUF)
ηi,pv /ηi,vop Proportions of demand consumption factor between different periods
λ̄ki Non-negative coefficient of degree of altruism of customer i, class k
k
β ki /β i Min/max value of risk-aversion coefficient of ith customer, class k
µkβ /σβk Mean/standard deviation of risk-aversion coefficient of class k
r1k /r2k Price ratios of the different periods
Tki Total budget of customer i, class k (¢/day)
dki,p,tp ,o /dki,v,tp ,o /dki,op,top ,o Demand of ith customer, class k at time t
in peak/valley/off-peak period before DR (kW)
k /σ k
σi,pv k
i,pop /σi,vop Elasticity of substitution of ith customer class k between
peak to valley, peak to off-peak, and valley to off-peak
F Dik Flexible demand factor of ith customer, class k
k
F Dki /F Di Min/max of flexible demand factor of ith customer, class k
µkF D /σFk D Mean/standard deviation of flexible demand, class k
ρFt R LSE’s selling price for inflexible demand at time t (¢/kWh)
PtIF Total inflexible demand at the system level at time t (kW)
ρDR
t
/ρDR
t Min/max selling price for flexible demand at time t (¢/kWh)
ρM
t
CP Market clearing price at time t (¢/kWh)
PL,n,t /QL,n,t Active/reactive load demand of bus n at time t (kW/kVAr)
gnm /bnm Conductance/Susceptance line connected to bus n and m (p.u.)
IF /QIF
Pn,t Active/reactive inflexible demand of bus n at time t (kW/kVAr)
n,t

Sbnm,t Mean apparent power flow between line (n, m) at time t (kVA)
P G,n,t /P G,n,t Min/max of active power generation on bus n at time t (kW)
Symbols xxxi

QG,n,t /QG,n,t Min/max of reactive power generation on bus n at time t (kVAr)

′ k,DR′
ρtk,DR /ρt Min/max of selling price for flexible demand of DRP k
at time t (¢/kWh)
k,F
dk,F
i,n,t /di,n,t Min/max of flexible demand of customer i, located at bus n,
of class k at time t (kW)
k,D k,U
Ri,n,t /Ri,n,t Ramp-up/down limit of flexible demand of customer i, located at
bus n, of class k at time t (kW)
k,F
Ei,n,t,o Total energy consumption of customer i, located at bus n,
of class k at time t (kW)
dk,F
i,n,t,o Flexible demand of customer i, located at bus n, of class
k at time t (kW)
ρRT
t
P LSE’s selling real-time price at time t (¢/kWh)
ρk,F
t
R
LSE’s offered nominal rate to class k at time t (¢/kWh)
ϑkn Binary variable associated with DRP k on bus n
κk Price subsidy/charging factor of class k
k,F
Pt,o Flexible active demand of class k at time t before DR (kW)
ℓk /ℏk Coefficient parameters of incentive function
k,F
P k,F
t /P t Min/max of flexible demand of class k at time t (kW)
Pt,max Total Maximum flexible demand at time t (kW)
ρk,IN C /ρk,IN C Min/max of the offered incentive rates
t t

to class/DRP k at time t (¢/kWh)

V n /V n Min/max of voltage magnitude on bus n (p.u.)
Decision Variables
dki,t Demand of ith customer class k at time t (kW)
ρkt Utility’s offered dynamic price to class k at time t (¢/kWh)
∆dki,t Change in Demand of customer i, class k at time t (kW)
λki Lagrange multiplier of customer i, class k
dki,p /dki,v /dki,op Demand consumption of ith customer, class k during
peak, valley, and off-peak period (kWh/period)
dki,p,tp /dki,v,tp /dki,op,top Demand of ith customer, class k at time t during peak, valley
and off-peak period (kW)
Symbols xxxii

ρDR
t LSE’s selling price for flexible demand at time t (¢/kWh)
PG,n,t /QG,n,t Active/reactive power generation of bus n time t (kW)
en,t /fn,t Real/imaginary distribution voltage of bus n at time t (p.u.)
PtF Total flexible demand at the system level at time t (kW)
dk,F
n,t Flexible demand of bus n under class k at time t (kW)
dk,F
i,n,t Flexible demand of customer i on bus n of class k
at time t (kW)
′
ρk,DR
t DRP’s selling price for flexible demand of DRP k
at time t (¢/kWh)
ρk,IN
t
C
LSE’s offered incentives rate to class k at time t (¢/kWh)
PLoss,t Total active network power loss at time t (kW)
k,F
Pn,t /Qk,F
n,t Active/reactive flexible demand
on bus n of class k at time t (kW/kVAr)
Vn,t Voltage magnitude on bus n at time t (p.u.)
k,F
Pn,t /Qk,F
n,t Active/reactive flexible demand on bus n of class k
at time t (kW/kVAr)
x, y, z State/decision variables
Chapter 1

Introduction

The traditional power system is hierarchically divided into generation, transmission, and
distribution sides. It facilitates a systematic unidirectional power delivery from the gen-
eration side to the customer’s end with a demand-supply balance. Conventionally, such
electrical power exercises have been in inception for many decades. However, the excessive
demand rise, population growth, continuous depletion of fossil resources, modernization
in living standards, etc., have adversely increased the need for higher generation capacity.
This has had a re-look into the conventional power operation and reformed it via electricity
restructuring, deregulation, renewable generation, etc [1]. Such paradigms have also been
propagated at the distribution level in recent times. It is emerging in the forms of retail
competition, renewable distributed generations (RDGs), energy storage systems (ESSs),
advanced communication infrastructure, etc., in the contemporary distribution systems
(DS). Besides, the customers’ perspectives are also changing from passive to active (i.e.,
prosumers), similar to the legacy (passive) to an active DS. It is actively increasing the
customers’ involvement via electric vehicles (EVs), local generation, and demand response
(DR). However, these significant steps are also overwhelming the contemporary DS, leading
to a sudden demand rise, intermittent generation, price volatility, etc [2]. These appre-
hensions made the current DS incorporate customers’ involvement actively to enhance the
security and reliability. This led to the advent of DR, which manifests an efficient energy
utilization through dynamic pricing or financial incentives.
DR is considered a promising innovation in the era of information and communication
technology (ICT) [3–6]. It has efficiently bought supply and demand sides together vir-
tually, which are physically miles apart. Moreover, this proliferates the idea of the smart
grid in the electric grid, enabling the intelligent, controlled, communicated transmission

1
Chapter 1. Introduction 2

of the power from the generation to the customers’ end [4, 6]. It is making DR perva-
sive in the current DS. DR has the potential to facilitate techno-economic benefits for all
stakeholders [7] and has an optimistic attitude to combat unprecedented situations. The
DR has been privileged to offer several advantages, such as alleviating peak demand rise,
customer bill reduction, utility’s techno-economic benefits, deferring network expansion,
competitive environment, etc. Though, theoretically, DR possesses to have many potential
benefits. But its practical implementation suffers from the various barriers such as lack
of educational awareness, a true value of DR, redesigning market regulation under the
changing scenarios, technical assistance, etc [8]. Moreover, DR modeling and operation
in the contemporary DS emphasize more feasible aspects and perspective. This manifests
DR into a prime area of research in DS. Hence, this thesis highlights on the DR modeling,
operation strategy, and assessment in the contemporary DS.
According to the U.S. Department of Energy (DOE), “Demand response is a tariff or
program to motivate end-use customers to changes in electric use in response to changes
in the price of electricity over time. Alternatively, to give incentive payments designed
to induce lower electricity use at the times of high market prices or when grid reliability
is jeopardized.” [7]. It is defined as price-based (PB) and incentive-based (IB) programs
based on how the customers participate in DR. The PBDR programs offer time-varying
prices to adjust the load patterns. In contrast, IBDR programs provide financial incentives
for load curtailment during peak periods. The real-time price (RTP), time-of-use (TOU),
inclining blocking rate (IBR), and critical peak pricing (CPP) are featured pricing models
in PBDR [2, 9–11]. These pricing models exhibit varying degrees of price variation over
time. The well-known IBDR programs are direct load control (DLC), interruptible/cur-
tailable services (ICS), peak time rebate (PTR), demand bidding (DB), ancillary service
based DR (ASDR), etc [9, 12, 13]. The IBDR programs are described as voluntary and
mandatory programs with infrequent and lean operation time [9].
The early DR applications in the literature have been extensively utilized in the whole-
sale market. It measures an aggregated DR response [9–11, 13–15]. This aggregated DR
is usually constituted by the specific customer classes having better instrumentation &
control such as industrial and commercial customers [16, 17]. However, the transition of
ICT, urban development, affluent lifestyles, flexible & controllable loads, etc., have also
expanded the DR reach to residential customers [4, 18, 19]. Though, these customers have
low demand levels but they constitute a large proportion of the energy sectors. Hence, a
substantial DR contribution is coming to the fore from all demand levels customers lately.
The DR representation has usually been described through the modeling approaches. It
illustrates load pattern variations with the change in prices or offered incentives. The
Chapter 1. Introduction 3

price elasticity model (PEM) is a prominently employed method for representing the cus-
tomer’s behaviour in DR modeling [9, 13, 14, 19–22]. It is adapted from microeconomics
and measures the relative change in demand with the relative change in the price [23, 24].
It estimates demand variation through aggregated elasticity and has been widely applied
in the wholesale market and at the distribution level. Though, it discards the effect of
DR reflection at the customer level, which is of prime importance at DS owing to the
different customer classes. Moreover, the standard PEM-based PBDR models have been
developed considering moderate customers’ behavioral attributes such as adaptability and
adjustability, load recovery, risk behaviour, etc. In addition, PEM-based PBDR extension
into the IBDR model has been observed to have an analytical drawback in representing
the incentive/penalty effect on elasticity variation [9, 10, 21, 22]. It has been assessed us-
ing price elasticity even in IBDR. This illustrates the shortcoming of the standard PEM
modeling for PB and IBDR programs in DS.
Another method in PBDR modeling is based on the UFs. It measures the customers’ util-
ity for the demand consumption. The UFs applications in DR have been applied to the
system’s demand as well as to the individual customer’s load modeling [25–31]. The cardi-
nal UF (CUFs) and ordinal UF (OUF) are two majorly employed in DR modeling. These
functions can exhibit the customers’ willingness or preference uniquely. Moreover, the
CUF and OUF-based DR models use fewer parameters for DR representation as opposed
to PEM-based DR models. However, in the case of CUFs, their coefficient attributes and
determination have hardly been part of the literature. Moreover, the utility of the various
functional forms of CUFs are moderately assessed to represent the customers’ rationality
in DR [25, 30, 31]. On the other hand, OUFs applications have been promising in DR
modeling. These OUFs have the potential to model individual customer’s load patterns
using the fewest controlled parameters compared to PEM and CUFs based DR modeling.
However, it has been attempted very limited in literature and is explored minimally in
DS. Moreover, DR models should consider class-wise load patterns, load diversity, price
cross-subsidy, demand level, etc., for a comprehensive illustration of realities in DS.
With increasing sizable demand in the participation, DR has also achieved substantial at-
tention in the operational decisions strategies or frameworks. It has re-positioned DR inte-
gration from the passive participant to an active decision-maker under the current environ-
ment in DS. This has led to the development of decision frameworks or strategies under PB
and IBDR [32–39]. Several works on the economic PBDR decision model in DS have been
reported in the literature. However, the earlier research works have been more devoted
to estimating the system’s techno-economic performance for the modified load pattern. It
put less insight on DR integration, behavioral features, strategic interactions, operational
Chapter 1. Introduction 4

structure, valuations, etc., in the system. In addition, the designed PBDR strategies lack
in representing the pragmatic characteristics of DS [35–37, 40]. It has been significantly
transformed DR under the competitive environment, where DR is evolving into a strategic
decision-maker. It has created multiple stakeholders such as distribution system operator
(DSO)/load serving entity (LSE), demand response providers (DRPs)/demand response
aggregators (DRAs), and customers in DR. This makes economic PBDR based decision
problems more challenging to address each stakeholder’s interest appropriately and ratio-
nally. It has been modeled using a multi-level decision framework in DR. Further, their
strategic interactions have been described using a game theory. However, game theory
application in DR has been developed on the small-size problems using more superfi-
cial assumptions [25, 26, 35, 41]. Moreover, the modeling and optimization of multi-level
problems are highly complex, nonlinear, and non-convex. Hence, the standard analytical,
numerical methods are insufficient to solve such problems. Though, heuristic and meta-
heuristic algorithms can solve such problems but suffer from high computation burden and
parameter tuning. Therefore, the suitable customization and modification in the existing
optimization approaches may need to be devised for the multi-level problems.
Similarly, IBDR strategies have also been implemented in the operational decision frame-
work in DS. It offers financial incentives or reduced bills for load shaving during peak
periods. The various IBDR strategies have been proposed to induce DR in the litera-
ture. It has usually been formulated between LSE/DSO and DRPs or DRAs and the
customers [38, 42–45]. These different DR stakeholders interact to determine incentive
rates while optimizing the system’s operational costs. However, most of the former IBDR
frameworks usually ignore the competitive interaction among the DR stakeholders for
incentive valuation. Moreover, the IBDR problems have been solved using the approxima-
tions. Therefore, a comprehensive IBDR decision framework considering the practical DS
characteristics needs to be developed from the contemporary DS viewpoint.
This thesis presents the DR perspective in the modeling and optimal operational decision
strategies in the contemporary DS. It proposes new and improved DR modeling approaches
using PEM and UFs considering the prevalent features of the DS such as the different cus-
tomer classes, distinct load patterns, load diversity, price cross-subsidy, etc. Moreover, the
customer’s behavioral attributes, such as degree of willingness, load recovery, risk-aversion,
etc., are also incorporated in DR modeling. The proposed DR models are motivated by
the consumer theory in microeconomics. These DR models are tested and investigated
on the different customer classes’ load patterns under the various existing and proposed
pricing strategies. Further, DR as a strategic decision player is incorporated in optimizing
the system’s economic decision framework for both (PB and IB) DR programs. It has
Chapter 1. Introduction 5

been planned in the competitive environment under multiple DR shareholders to achieve

the system and its shareholders’ objectives. Besides, improved and sophisticated solution
methodologies have been proposed to solve the DR problems. The proposed economic DR
frameworks and the solution methodologies have been thoroughly synthesized and opti-
mized on the standard 33-bus test and practical Indian-108 distribution systems. Further,
the application results are investigated and analyzed from the perspectives of the system
and DR’s stakeholders.
The following chapter presents a comprehensive literature review on DR modeling and its
operational framework/strategy in the DS. Then, the critical review is drawn to identify
the research gaps and presented with the research objectives. At last, the brief descriptions
of all subsequent chapters are discussed.
Chapter 2

Literature Review

The expansion of deregulated environment, technological development, and advancement

in information infrastructure has significantly transitioned the contemporary distribution
systems (DS). Accordingly, it has instilled the contemporary DS with the various dis-
tributed energy resources (DERs), advanced metering infrastructure, and the retail com-
petition. This has prompted an overhaul in the operation and coordination between the
supply and demand sides. One of the eminent outcomes of these transformations is demand
response (DR). It is the result of persistent concerns such as sudden peak rise, increasing
variability, intermittent generation, price volatility, etc., in DS [1, 7]. DR has been propo-
sitioned to provide demand flexibility, less capital investment, delay network expansion,
better coordination of DERs, etc. The DR applications have been exhaustively applied in
the wholesale electricity market. It provides an aggregated DR response. However, DR
assessment at the distribution level needs conversion in its evaluation approaches due to
emphasis on the load diversity, class-wise load patterns, societal aspects, customers’ be-
havioral attributes, etc. It requires a sophisticated DR approach to realize the customers’
attributes. Moreover, DR decision operational strategies should be formed and upgraded
to the impeding retail competition in DS. This chapter presents the literature review rel-
evant to the DR modeling approaches and DR strategies in the contemporary DS. Some
research gaps have been identified based on the current state-of-art in the contemporary
DS, then the research objectives of the work are proposed. The literature review has been
categorized in the following sub-areas to cover the scope of the work.

ˆ Review on PBDR and IBDR modeling using price elasticity model (PEM).

ˆ Review on PBDR modeling using utility functions (UFs).

7
Chapter 2. Literature Review 8

ˆ Review on PBDR decision strategies

ˆ Review on IBDR decision strategies

2.1 PBDR and IBDR Modeling using PEM

Elasticity, in general terms, is defined as the change in the demand due to the change
in price. It is usually expressed in per unit, and its complete form is called the price
elasticity matrix or model (PEM). PEM is considered the straightforward and appealing
method in DR modeling. The early applications of PEM have been vastly utilized in the
wholesale electricity market [10–12, 14, 15, 20, 22]. It typically measures the customers’
aggregated response at the system level through elasticity, then its consequences on the
system’s techno-economic factors such as demand, market price, profit are evaluated. In
DR analysis, the elasticity values are usually known beforehand based on the empirical or
surveying approach. In Ref. [10], a comprehensive economic PBDR model is investigated
under the various PBDR programs. The analyses reveal that DR induces better peak
reduction and improved load factor in most PBDR programs. The different functional
forms of PEM-based DR models using logarithmic, exponential, and power functions are
developed in [11, 13]. These functional approaches describe the degree of variation in the
demand under the different time-based prices. The results show that the small elasticity
and small price variation evince similar responses in all PBDR models [13]. On the other
hand, each PBDR model gives distinct DR under a nominal or higher variation [11]. It in-
dicates that the response in the DR depends upon the load patterns. An augmented form
of the different linear and nonlinear functions using a composite weighted DR is suggested
in [46]. It combined all functional forms to encapsulate their distinct behaviour.
PEM-based DR models have also been performed in DS [32, 33, 47–50]. A short and
medium-term decision framework under DR is suggested to measure its impact on the dis-
tribution company’s (DisCo) profit [32, 33]. The DR is measured using PEM under RTP
and TOU, respectively. An integrated distribution planning operation in the presence of
distributed renewable resources and DR is suggested to elevate a low-carbon DS [48]. The
impact of DR on distribution network (DN) operation is performed under critical peak
pricing (CPP) and hourly pricing (HP) [47]. However, most PEM-based DR models in
the electricity market [9–11, 14, 20] or DS [32, 33, 47, 48] have usually been performed at
the system or utility level using the aggregated load approach. Moreover, DR valuation’s
elasticity values are considered a static value over the time horizon. Though, it may not
be appropriate due to distinct demand and price variations over multi-states [12, 51].
Chapter 2. Literature Review 9

Though some aforesaid concerns have also been addressed in the literature [22, 52–56].
The elasticity has been envisioned through the deterministic or stochastic approaches.
It has been suggested as the different elasticity matrices based on the customer’s ratio-
nality [51, 55, 57] in the deterministic method. Though, there is no analytical model to
generalize these elasticity variations. In some cases, the elasticity is estimated using curve-
fitting [46, 56, 58], or regression [54] through the available historical data. The stochastic
approaches have utilized a probabilistic approach to generate elasticity in the predefined
range [53]. In Ref. [19], elasticity for the residential customers is estimated by considering
an appliance-wise elasticity obtained through surveying under IBDR.
The use of PEM in the incentive-based (IB) DR model is modeled by augmenting an incen-
tive/penalty term with the PBDR model during peak hours [9, 10, 19, 22]. The inclusion
of incentives alters the change in price terms, which induces further demand reduction
during peak hours [10, 11, 19, 58]. However, the standard IBDR model has been mod-
eled without considering the effect of incentive term in elasticity variation. It has been
observed that elasticity value in IBDR models usually has been taken as price elasticity
in PBDR [9, 10, 13]. Moreover, less attention has been paid to determining the incen-
tive/penalty value and magnitude scale under the different pricing strategies. It is either
assumed as a fixed [9, 10], or varying value [22].

2.2 PBDR Modeling using Utility Functions

The PEM-based DR models offer a straightforward mathematical approach to represent

the customer’s behaviour in DR. However, its elasticity estimation over multi-states makes
PEM a tedious and rigorous approach. Moreover, it lacks in attributing the customers’
willingness or preferences uniquely [27]. Unlike PEM, the utility functions (UFs) are con-
sidered an appropriate DR modeling approach. These UFs exhibit DR behaviour using
fewer parameters in addition to the customer’s preference [27, 29]. In DR, cardinal utility
functions (CUFs) [25, 26] and ordinal utility functions (OUFs) [27, 29] are predominantly
employed functions. Both functions exhibit satisfaction gains differently, wherein CUFs
measure satisfaction based on the total unit consumption, whereas OUFs measure satis-
faction based on the ordering of units.
Several works based on CUFs in DR have been described in the literature [25,26,30,59,60].
It has been applied as a social welfare or customer’s payoff function [25,59], customer mod-
eling [26], electric vehicles (EVs), plug-in hybrid electric vehicles (PHEVs) [60], etc. The
utilization of CUFs varies with the types of loads or customers. In Ref. [26], EV customers
Chapter 2. Literature Review 10

are modeled as a quadratic function, DR customers [26], and PHEV load [60] as a logarith-
mic function. In Ref. [30], a logarithmic function is utilized to model the customer’s utility
for proportional fair DR allocation. The different types of CUFs are employed to model the
utility gain from the different appliances and associate it with appliances’ control variables
such as temperature, energy, power, etc., in DR [59]. An appliance-based concave utility
function is employed in [61] to define the gain from the load consumption. It defines the
utility’s coefficients based on the total energy consumption over the scheduling horizon.
The setting of coefficients is based on analytical analysis of CUF, which estimates a feasi-
ble value or the range. However, in some studies, these coefficients have been considered
as the fixed [49] or random values [25]. A calibration approach is adopted in [62] to model
the customer’s load behavior based on the available utility data information. Similarly, in
Ref. [63], a dynamic benefit function is proposed using a day’s typical demand and price
curve.
The OUFs are extensively utilized in the production theory, economic growth model, fi-
nances, etc., in economics [23, 64, 65]. It has also been effectively applied in DR with
the feature of minimal parameter modeling. Handful studies based on OUF in DR are
constant relative risk aversion (CRRA) [27], constant elasticity of substitution (CES) [29],
and Cobb-Douglas [28]. In Ref. [29], the authors presented a CES based utility function
to assess the customer’s preference in DR under TOU framework. The substitution rate
parameter adjusts the load consumption level in the different periods with price variation.
A similar DR model is formulated using a Cobb-Douglas function for DR modeling in
two-tier TOU framework [28]. However, it shows inadequacy to imitate the intertemporal
constraint of flexibility constraints, which allows the customer to adjust its demand in the
nearest states [66]. An economic DR framework using the overlapping generation (OLG)
model is proposed in [27]. It is developed for a two-state TOU framework to show its
usefulness in PBDR modeling.

2.3 Price Based Demand Response

DR modeling is a primary and effective tool for realizing the customers’ response. It is
usually considered a preliminary assessment approach. However, DR’s perspectives have
been significantly progressed with the emergence of the smart grid [18]. This has led to
active DR participation in the operational decision-making of the system or the utility [1].
The DR inclusion in the decision making has also created multiple stakeholders such as de-
mand response aggregators (DRA)/ demand response providers (DRP), customers besides
distribution system operator (DSO), or load-serving entity (LSE). This has given rise the
Chapter 2. Literature Review 11

multi-levels decisions models to realize each stakeholder’s interest in the DR environment.

A brief literature review on the optimal PBDR decision frameworks and their operational
strategies in DS are described in the following section.
Safdarian et al. proposed short and medium-term DR decision models to optimize the
DisCos’ profit under RTP [32] and TOU [33] pricing. The DR frameworks are formulated
using PEM in the stochastic environment to account for the uncertainty associated with
price volatility and customer loads. Conjeo et al. suggested a real-time DR model for
optimizing the customer’s utility subjected to minimum energy consumption, minimum
load levels, ramping level, etc [67]. In Ref. [48], an integrated planning for a low-carbon
DS is proposed with the aid of distributed generations (DGs) and DR. The DR is modeled
using PEM, and load recovery in DR using a fading effect.
An optimal DR for the unbalance distribution feeder using the smarts controllable loads
such as air conditioner and heating ventilator is presented to increase the peak shaving
during summers. It proposed a peak saver (PS) program for Ontario, Canada’s residen-
tial and commercial sector to induce DR and optimally control loads, tap-changers, and
shunt capacitors [68]. In Ref. [69], a smart distribution system consisting of uncontrol-
lable price-responsive loads and controllable loads is presented to assess DR impact on
the distribution feeder operations. However, the described literature presents the system’s
techno-economic objectives viewpoint with a moderate emphasis on the DR activators
(i.e., customers and DRPs). This necessitates the inclusion of all the contributing partic-
ipants in DR. It is further amplified with the increasingly competitive environment and
information exchange from the wholesale market to the customer end. This has resulted
into the multi-level economic decision models in DR to explore the interaction among the
various participants, which optimize their interests independently.
A bi-level programming or Stackelberg game (SG) is a widely utilized framework in DR.
Both bi-level and SG are equivalent approaches and differ only on the way of representa-
tions. A bi-level is mathematical programming and is considered a dynamic form, whereas
the SG is based on game theory and is a static form of a two-level problem. The SG is a
sequential process to model the strategic interactions, where one is a leader and others are
followers. The bi-level or SG applications in DR have been broadly demonstrated between
load-serving entity (LSE)/distribution system operator (DSO) and DRP [35, 36, 70] or be-
tween DSO/DRP and customers [37, 40, 71]. These interaction mechanisms are employed
to determine the optimal retail price for DRPs or the customers for inducing DR while
optimizing their interests [30, 31, 34, 36, 71]. In Ref. [70], a multiple DRA in coordination
with wind generation is proposed to minimize the electricity market operator cost in a bi-
level problem. A dynamic PBDR framework is presented as a bi-level problem in DN [35]
Chapter 2. Literature Review 12

for the energy scheduling problem, where the LSE operates at the upper level (UL) and
DRAs at the lower level (LL). In addition, a hybrid pricing, fixed price for inflexible loads,
and dynamic price for flexible load are proposed to induce DR. A SG between electricity
utility company (EUC) and DRAs is formulated to optimize DR integration in DN using
a transitive energy signal [36].
In Ref. [37], a game-theoretic framework is suggested between the energy retailers and cus-
tomers for DR. The energy retailers determine the optimal pricing schedule to maximize
their profits, whereas the customers optimize its flexible consumption. A similar idea is
also suggested in [71], where the optimal TOU pricing encourages the customers to induce
DR. A two-stage, two-level energy pricing and energy dispatch problem is formulated un-
der DR and price uncertainty [40]. It modeled the interaction among the energy market,
retail market, and customers through SG and is tested through robust optimization to
assess the worst-case condition. However, SG-based DR problems have been primarily
concerned with the deduction of Stackelberg equilibrium (SE) to illustrate the strategic
behaviour of customers and DRPs or the utility in games [26, 31, 36, 71].
The tri-level problems have also been applied in the DR-based decision framework compris-
ing DSO, DRP, and customers. A tri-level model consisting of the utility, DRP, and cus-
tomers operating at top, intermediate, and bottom level, respectively, is suggested in [72].
The utility optimizes its operation cost by offering rewards to DRPs, which evinces DR
from the customers at the LL in the exchange for the incentives. A similar framework is
also presented in [73], where the energy supplier in the top hierarchy induces DR through
TOU pricing among DRPs, operating at the intermediate level, who subsequently obtain
it through the customers in return for financial incentives. A tri-level DR framework com-
prising three layers viz., wholesale market, DRPs, and the flexible customers is proposed
in [74]. It optimizes DRP bidding in the wholesale and real-time market while procuring
DR through the customers with flexible loads via rewards. In Ref. [75], the impact of de-
mand flexibility on the electricity retailers is investigated in a tri-level framework, where
the electricity retailers, operating at the UL induce DR among the flexible customers which
are operating at the intermediate level. Further, analyze the impact of demand flexibility
on market-clearing in the wholesale market at the LL. Ref. [40] describes a tri-level retail
market for the energy pricing and load dispatch considering price uncertainty. However,
in most of the research works [72,74,75], DR at the customer’s level is usually activated by
incentives as opposite to price signal except in Ref. [40]. It has also been modeled without
the practical technical constraints in a tri-level optimization [40, 72, 73, 75].
The optimization of multi-level problems has been the subject of wider studies in the
literature. It includes the various optimization approaches such as classical, numerical,
Chapter 2. Literature Review 13

heuristics, meta-heuristic, etc. The classical optimizations comprise single level, gra-
dient descent, and penalty methods. These approaches generally utilize Karush-Kuhn-
Tucker (KKT) condition for the optimality conditions. In numerical methods, linear pro-
gramming, mixed-integer programming, quadratic programming, nonlinear programming,
branch & bound (B&B), etc., are employed to solve the multi-level problems. The heuristic
and meta-heuristic methods are genetic algorithm, particle swarm optimization, differential
evolution, evolutionary, etc. Since the optimization/ algorithms, applications vary with
the problem complexity; hence, the diversified algorithms have been applied in bi-level
or tri-level problems in the literature. Mathematical programming has been extensively
applied on bi-level [35, 36, 76] or tri-level problem [40, 73, 75, 77]. The other approaches,
iterative methods, have been exclusively applied to the small-scale problem [30, 31, 71].
Moreover, decomposition-based algorithms have also been incorporated to solve complex
multi-level problems. It decomposes a large-scale problem into master and sub-problems
to lower the complexity of the problem. Lagrange relaxation [78], column-constraint gen-
eration [79, 80], Dantzig-Wolfe Decomposition [81], and Bender decomposition [82] are the
predominantly employed approaches in the power engineering problems.

2.4 Incentive Based Demand Response

Incentive-based (IB) DR is considered a progressive DR program relative to price-based

(PB), where the customers are incentivized in the former and imposed dynamic price in the
latter. It is based on the customer’s behavioral attributes towards reward/incentive and
punishment, where the incentive is seen as reward and dynamic price as a punishment [9].
The IBDR applications are usually employed during peak periods; hence, they have a lean
operation time, lasting for a few hours in a day, month, or year depending upon their
applications [3]. The existing conventional IBDR programs are direct load control (DLC),
interruptible-curtailable services (ICS), emergency demand response (EDR), peak time
rebate (PTR), etc. These conventional approaches are more system operator or utility
controlled and have higher jurisdiction. However, a paradigm shift has been observed with
the evolution of the deregulated environment in the contemporary DS. This is creating
a competitive atmosphere with more involvement of the various DR enablers. DR has
also been evolving as the strategic decision variable in the economic decision frameworks.
Hence, a brief literature survey on the growing economic decision frameworks or strategies
under the ambit of IBDR in DS is presented below.
Chapter 2. Literature Review 14

The strategic interactions have been vital to illustrate the competition among the differ-
ent entities such as generators, suppliers, customers, DRPs, etc., in the electricity mar-
ket [83–85]. In such problems, an individual’s action is also influenced by rival players’
actions [86] and typically competes non-cooperatively. This leads to multiple leaders’
decision-making problems, which is delineated as a multi-leader-multi-follower game in
the game theory [83, 87]. Its equivalent mathematical form is termed as the equilibrium
problems with equilibrium constraints (EPEC), which is an amalgamation of several math-
ematical problems with equilibrium constraints (MPECs) [86]. MPEC is the simplest form
of SG, which is represented through the KKT conditions to obtain the stationary condi-
tions [88]. Similarly, the EPEC model is formed by combining all leaders’ equivalent KKT
conditions [86]. Further, EPEC applications have been widely applied in the strategic
gaming analysis in deregulated electricity markets [83, 86].
It also has been extended in DS in the literature [89–92]. In Ref. [89], the authors sug-
gested a two-stage optimal decision framework to optimize day-ahead and real-time cost
through distributed generations (DGs) and interruptible loads (ILs) in DS. It considers
only spot price as a strategic interaction. It is better demonstrated in [90], where the
strategic interaction is envisaged among the multiple discos via DGs, considering the dis-
tribution and transmission sides. A DG integration mechanism is suggested to include
DGs as the competing power producers in the wholesale market for the competition [91].
In Ref. [93], the optimal contract pricing scheme for dispatch-able DGs is formulated as a
bi-level problem. DGs owners maximize their profits in the UL, and disco minimizes the
operating cost at the LL. Though, these strategic interactions have been limited to the
asset-light suppliers such as DGs and energy storage systems (ESSs) in DS [92]. However,
DR has also been emerging as a viable option with increase in the customer’s participation.
The IBDR-based decision frameworks have also been investigated in DS in the literature.
A decentralized framework consists of a DR aggregator (DRA), and the customers are
envisioned. In, each entity optimizes its objectives via price and incentives-based DR to
mitigate the overloading of DN [38]. In Ref. [94], a DR aggregation mechanism is proposed
for residential customers by offering financial incentives for minimizing the LSE operating
cost. It also accounts for the degree of comfort levels via a scaling approach. An IBDR
based retail competition in the domain of multiple retailers consisting of DGs and ESS is
proposed to optimize its payoff [45]. It lowers the financial losses and defers the future
capacity charge by employing its asset-light resources in DN. A coupon-incentive-based
DR (CIDR) framework is suggested to optimize the operational cost during peak periods
under the ambit of multiple load aggregators (LAs) to reflect the diversity in their opera-
tional costs [43].
Chapter 2. Literature Review 15

Though, multiple DRPs/DRAs/LAs have been widely included to improve the system’s
operational and technical objectives in DS/DN. However, their valuations in a competitive
environment have rarely been part of the described studies. This makes the competitive-
ness overlooked in determining incentive price and its subsequent effect on the system’s
objectives. It is investigated for a single DRP in [38]. In Ref. [42], a strategic interaction
is established between LSE and DRAs, but it is not considered amongst DRAs. Similar
work is also considered in [94], where multiple residential DRAs interact with LSE only.
In Ref. [44], the competition among the multiple retailers is modeled using DGs and ESS,
but not with DR. Similarly, in Ref. [43], DRP’s strategic behaviour is not considered.
However, incentive rates are weighted using a Shapley value to show their influence in
the system. Moreover, IBDR decision frameworks in DS have been developed using the
straightforward assumptions such as without power flow [43, 44, 94] or linear ac power
flow [38], or dc power flow [42].

2.5 Critical Review

DR is an effective approach for inducing load adjustment. It has been professed to have
adequate competence for providing the various techno-economic benefits in the short and
long run. DR is primarily induced by PBDR and IBDR programs. It has been illustrated
through modeling in the earliest developed DR applications. It models the demand varia-
tion subject to the price variation. The consumer theories in microeconomics, a sub-part
of economics, have widely inspired these modeling approaches. Among, PEM and UFs are
considered primary means for DR assessment. The PEM is an extensively utilized, simple,
and appealing model to represent the customer’s response in PBDR and IBDR. Though,
PEM applications in DR partially reflect the customer’s adaptability and adjustability.
Its elasticity is generally assumed as a fixed value over the different segmented periods.
This diminishes the virtue of elasticity, which is required to be dynamic as the demand
and price are time-dependent over multi-states. Therefore, the suitable changes in PEM
need to be addressed for pragmatic representation of the customer’s behaviour in DR.
The PEM-based PBDR has also been extended in IBDR using the augmentation. It in-
troduces an incentive/a penalty term to transform PEM-based PBDR into IBDR. This
results into an augmented DR model combining the features of both DR models. How-
ever, it analytically lacks representing the exact effect of incentive inclusion on elasticity
variation in the standard IBDR model. It keeps the elasticity value the same as price
elasticity in the IBDR model. Since, the incentive term is equivalent to price, therefore
addition of the incentive term will also change the elasticity in overall, relative to elasticity
Chapter 2. Literature Review 16

in PBDR. This indicates inadequacy of the standard IBDR model, which may further put
erroneous notion on the observations drawn from the obtained demand after DR. Hence,
an appropriate exact IBDR models need to be developed.
Other methods of DR modeling are based on UFs. These UFs have an inherent capability
to model the customer’s attributes through its utility parameters in addition to adaptabil-
ity and adjustability in DR. In, CUFs are extensively employed UFs to model customer’s
response in DR. It makes use of relatively less parameters than PEM. Though, the CUFs’
utilities have been limited to the small-scale problem. In addition, the setting of CUF’s
coefficient parameters is hardly based on rational approaches. Moreover, the usages of the
various functional forms of CUFs in DR modeling are modestly assessed. Besides, CUF’s
coefficient parameters significance in relation to the customer’s behavioral attributes such
as willingness, risk, etc., are given less importance in DR. On the other hand, OUFs ap-
plications in PBDR modeling have been emerged as the promising prospects. It exhibits
likewise features to CUFs, but uses very few parameters than CUFs and PEM. Though, its
applications in PBDR modeling are in early stage and have only a few established models.
Besides, PEM and OUFs based DR models have usually been assessed on the system de-
mand using the load aggregation approach. It provides an aggregated DR response, but
obscures the DR reflection on the customer-wise demand. However, with the current state
of communication infrastructure in the contemporary DS, a customer-wise DR assessment
seem to be more realistic. Hence, a disaggregated load modeling approach should be ap-
propriately considered for better DR assessment in DS.
The competitive environment and communication advancement have actively increased DR
participation in DS. This has also evolved DR as the strategic entity in the economic deci-
sion models. In addition, it has constituted the multiple strategic players, which are being
monetized, leading to multi-level PBDR decision problems. The multi-level problems are
mainly comprised of bi-level or SG and tri-level problems. It determines the optimal DR
price for inducing DR from the customers. Though, the degree of dynamic price greatly
affects the customers and LSE benefits. Hence, a suitable pricing approach price securing
both the entities’ interest needs to be devised. Moreover, bi-level models have been inves-
tigated on the small-scale problem, fixed types of customers, approximate power flows, etc.
In addition, a bi-level problem partially encloses all the contributing participants in DR.
It can be better illustrated in a tri-level decision model, which includes all participants.
However, tri-level models have been designed for a simpler problem without considering
the network constraints. Hence, a comprehensive economic PBDR decision model must
be considered to represent the broader perspective of the customers and DRPs. Further,
the tri-level problem optimization involves numerous decision variables in each level. It
Chapter 2. Literature Review 17

becomes quite a challenging task in a large-scale problems. Therefore, a decomposition-

based algorithm can be a viable option for solving such multi-level problems.
Similarly, the economic decision models have also been extended in IBDR programs in
DS. It offers financial incentives to the customers or DRPs to optimize the LSE/DSO’s
techno-economic cost during the peak periods. These problems have been designed in a
bi-level framework with the strategic interaction between LSE and DRPs. However, most
of the existing works have devised the strategic interaction between LSE and DRP. But
does not consider the strategic interaction among the DRPs. This mitigates the influence
of individual DRP on other DRPs, while optimizing its objective. Such situation occurs
under no competition among the DRPs, which may dissuade them from DR participation.
This lowers the importance of DRPs in determining the incentive price. Besides, the ex-
isting IBDR decision models have a shortcoming in representing a comprehensive IBDR
framework. It has been investigated using the approximate or without power flows. This
shows a shortfall of inclusiveness in the existing studies. Hence, the economic IBDR model
should adequately address the DRPs’ strategic behaviour in the evolving competitive en-
vironment in the contemporary DS.

2.6 Research Objectives

From the aforesaid critical reviews, the following research objectives have been framed.

1. To carry out the comprehensive literature survey in the areas of demand response,
DR modeling approaches, economic decision-based DR frameworks, etc., in the evolv-
ing contemporary DS under DR.

2. To develop new/improved PBDR and IBDR models based on PEM, and investigate
their impacts under the different existing and the proposed pricing strategies.

3. To model the cardinal and ordinal utility functions in PBDR and assess their response
in DR from the customers’ behavioral viewpoint.

4. To develop a comprehensive economic price-based DR decision strategy pertaining

to the different DR stakeholders in the contemporary DS; model their strategic in-
teractions and an efficient solution methodology to solve it.

5. To develop an incentive-based DR decision strategy considering the customer’s con-

straints and network technical constraints under the competitive environment in the
contemporary distribution system.
Chapter 2. Literature Review 18

This thesis’ scope and work focuses on the new/ improved DR modeling approaches and
the operational DR-based decisions strategies in the contemporary DS. It puts forward
the various DR modeling approaches using the microeconomics behavioral model, consid-
ering the customers’ behavioral viewpoints and assess their impact on techno-economic
benefits. Moreover, DR is integrated into the system’s decision-making process to assess
its techno-economic performances under the ambit of different DR stakeholders and com-
petitive environments in DS. The remainder of the thesis is outlined as follows:
Chapter 3 presents an adaptive PBDR and new IBDR model using PEM. The PEM-based
PBDR puts forward an adaptive and adjustable model to illustrate the customer’s rational
behaviour in the DR. It envisages the concept of dynamic elasticity over multi-sates to
make DR adaptive. In IBDR, an improved augmented IBDR model is proposed consider-
ing the effect of incentive inclusion in elasticity. It proposes a mathematical procedure to
extract the incentive effect on elasticity with the addition of incentive. The adequacy of
the proposed DR models are investigated on the class-wise load patterns in the standard
33-bus load data. The simulation results are presented and discussed under the various
existing and the proposed pricing.
Chapter 4 deals in PBDR modeling using UFs. Two different UFs, CUFs and OUFs are
employed. The different forms of the CUFs such as quadratic, exponential, logarithmic,
and power functions are considered in DR modeling. Further, each CUF’s response in
DR is assessed using the risk-based indices, different load patterns, and price variations.
Similarly, an OUF-based Overlapping generation is utilized in a three-tier TOU pricing
for PBDR modeling. It adjusts the demand with the relative change in price, and incor-
porates the customer’s willingness via a degree of risk-aversion. The application results of
the proposed PBDR modeling approaches are tested on the standard 33-us test load data
using disaggregated load modeling.
A hierarchical economic price-based DR framework comprising the different DR stock-
holders such as LSE, DRP, and the customers under the ambit of the DS is presented in
chapter 5. The proposed strategy is formulated as a nonlinear tri-level, two-stage prob-
lem, nested with bi-level in each stage considering ac network constraints. It consists of
LSE and DRPs in the first stage and DRPs and customers in the second stage. Further,
the problem is solved through reformulation & decomposition (R&D) based on the col-
umn constraint generation method (CCG). The applicability of the PBDR framework and
algorithm are thoroughly investigated and discussed on the standard 33-bus and Indian
108-bus distribution system.
In chapter 6, an economic incentive-based DR decision framework consisting of multi-
ple DRPs is presented under the competitive environment in DS. The proposed problem
Chapter 2. Literature Review 19

is formulated as a nonlinear, multi-leader-multi-follower problem in a bi-level framework

considering ac network constraints. It optimizes LSE operating cost at the upper level by
offering incentive prices to the DRPs at the lower level. The DRPs’ interaction is demon-
strated using a generalized Stackelberg game. The proposed problem is tested on the
standard 33-bus test and a practical 108-bus Indian distribution system. The simulation
results are presented and discussed in detail.
Finally, in chapter 7, conclusions and the salient contributions of thesis work are summa-
rized, and its plausible future directions within the scope of the work are presented.
Chapter 3

Price and Incentive Based

Demand Response using Price
Elasticity Model

3.1 Introduction

Demand response (DR) is a conducive phenomenon in the power system to alleviate peak
demand rise, price volatility, congestion, monopolies of generators, etc. It enables the
customer interaction, better energy utilization, deferment of network expansion, and peak
shaving. DR is usually rendered through modeling and is considered as one of the pre-
liminary tools to assess its potential. Price elasticity model (PEM) has been widely part
of the DR studies. It models the customer’s demand sensitivity to price variation via
elasticity. Its early applications in DR have been generally performed in the wholesale or
the electricity markets. However, a paradigm shift in DR assessment has also been per-
ceived with the advent of smart grid technologies in the contemporary distribution system
(DS) in recent times. This has encouraged DR participation from the small retail to large
customers. However, few research works have considered a class-wise DR assessment with
specific customer classes such as residential or industrial. Though, it is vital to consider
the different customers’ classes on account of their discrete load patterns. These distinct
load patterns reflect demand activity usage, diversity, societal costs, etc [95].
Most importantly, PEM-based PBDR analysis puts less insight on DR attributes i.e.,
adaptability and adjustability, as discussed in chapter 2. It is illustrated through two
types of elasticities, i.e., self and cross-elasticity. However, these two elasticities have

21
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 22

typically been considered as a fixed value on the different segmented periods [9, 10, 14].
Even, it is ignored for the cross-elasticity [32, 33]. This shows the partial adaptability
and adjustability under the standard PEM-based PBDR model. Further, it diminishes
the feature of elasticity, as elasticity (demand and price) is being time-dependent. The
standard price-based DR (PBDR) model has also been extended in incentive-based DR
(IBDR) model by introducing an incentive/ a penalty term. However, the standard IBDR
model lack in accounting the incentive rate addition on elasticity and is usually assumed
to the same as in PBDR. This indicates the analytical drawback of the existing IBDR
model. Hence, these shortcomings of both DR models are addressed in this chapter.
This chapter proposes an adaptive PBDR and an improved IBDR framework for DR mod-
eling. It presents an adaptive PBDR framework to illustrate the customer’s attributes of
adaptability and adjustability via a dynamic elasticity. It envisions a dynamic or adaptive
elasticity using two approaches, namely, deterministic and stochastic approaches. Both
approaches illustrate the customer’s load adaptability for the shiftable/flexible loads in re-
lation to price variation. In IBDR, a new incentive DR framework is proposed to account
the effect of incentive/penalty addition on elasticity. It presents an analytical method to
explicitly define incentive elasticity equivalent to price elasticity based on their respective
price and incentive rate terms. The usefulness of the proposed DR modeling approaches
is investigated on the 33-bus test distribution system. It is performed on the different cus-
tomers’ classes load patterns using a disaggregated load modeling approach. Further, the
comprehensive techno-economic DR analyses are performed and critically assessed under
the various existing and the proposed pricing strategies.

3.2 Proposed Price Based DR Framework

The proposed DR modeling is formulated using a disaggregated load approach to measure

the customer’s discrete response in DR [1]. It is evaluated using the net benefit function
or a social welfare function and is defined as the difference between the customer’s benefit
and its payment to the utility [1]. The net benefit function for ith customer of class k at
time t is expressed as follows:

S(dki,t ) = B(dki,t ) − dki,t ρkt (3.1)

where, S(.) and B(.) is called the net benefit/social welfare and the benefit function, re-
spectively. dti,k is the demand of ith customer of class k at time t and ρkt is the price offered
to class k at time t, respectively.
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 23

The formulated framework is based on the two assumptions. First, all customers are
assumed to be behave rationally in DR. Second, the overall consumption of energy re-
mains same before DR (BDR) and after DR (ADR). It is formulated by transforming
the customer benefit function into a constraint optimization problem. Such problems can
be effectively optimized using a Lagrange function [96]. The Lagrange function of the
proposed formulation is expressed as follows:
 
 X X 
L(dki,t ) = B(dki,t ) − dki,t ρkt + λki (dki,t,o − dki,t )− (dki,t − dki,t,o ) (3.2)
 
t∈ΩTp t∈{ΩTv ∪ΩTop }

where, the second term in (3.2) represents the difference of demand exchanged between
the peak ΩTp , valley ΩTv , and off-peak hours ΩTop . It is to be noted that the peak, valley
and off-peak hours are pre-determined and non-overlapping for the sake of simplicity.
The optimal condition of the Lagrange function is obtained by taking the first derivative of
(3.2) under the different periods. This gives a marginal rate of benefit function as follows:

∂B(dki,t )
= ρkt + λki ∀t ∈ ΩT (3.3)
∂dki,t

Further, the second derivative of (3.2) is defined as follows:

∂B 2 (dki,t ) ∂ρkt 1 ρko,t

2 = = (3.4)
∂dki,t ∂dki,t εki (t) dki,t,o

Since εki (t) is resulted due to the price variation. Hence, it is explicitly defined as price
elasticity in this chapter. Extending the customer benefit function in a quadratic form
using the Taylor series expansion, gives the following expression as follows [1].

2
ρkt,o (dki,t − dki,t,o )
B(dki,t ) = B(dki,t,o ) + ρkt,o (dki,t − dki,t,o ) + k,P E (3.5)
εi (t)dki,t,o 2

Differentiating (3.5) and equating it with (3.3), gives the customer’s demand at any time
t ADR as follows:
( )
(ρkt − ρkt,o + λki )
dki,t = dki,t,o 1+ εk,P
i
E
(t) ; ∀t ∈ ΩT (3.6)
ρkt,o
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 24

Equation (3.6) represents the demand ADR using self-elasticity. Similarly, PEM can be
extended into multi-periods by incorporating the effect of cross-elasticity as follows [10]:
 
 X k,P E k k k
(ρτ − ρτ,o + λi ) 
dki,t = dki,t,o 1 + εi (t, τ ) ;∀t ∈ ΩT , ∀τ ∈ ΩT (3.7)
 ρkτ,o 
τ ∈ΩT

Equation (3.7) represents the combined effect of self and cross-elasticity on demand under
DR. Where, self-elasticity manifests increment/decrement relative to change in the price
on the same period, whereas, cross-elasticity describes the transverse effect of the price on
the cross-demand. It is mathematically defined as follows:

∆d(t,τ ) ρo,(t,τ )
   
PE
ε (t, τ ) = × (3.8)
∆ρ(t,τ ) do,(t,τ )

(
εP E (t, τ ) ≤ 0 if t = τ
(3.9)
εP E (t, τ ) > 0 if t ̸= τ

From the electricity point of view, the magnitudes of both elasticities vary with the cus-
tomer’s reaction in DR. Hence, assessing each customer is a rigorous approach due to the
complex behaviour. Though, the analysis can be carried out to assess the impact of DR
with some approximate assumptions. Hence, this chapter presents a deterministic and a
stochastic approach to model DR using PEM.

3.2.1 Proposed Deterministic PEM (DPEM)

In DR, the interrelation between self and cross-elasticities has hardly been part of the
examination [9,10,13]. Though, a few Refs. [14,51] have touched upon the issue of optimal
demand redistribution in DR using PEM. These studies emphasized that the optimal load
redistribution in DR is possible when the elasticities are dynamic and adaptive. This
defines the following condition as follows
X
εP E (t, τ ) = 0 ∀t ∈ ΩT (3.10)
τ ∈ΩT

Equation (3.10) describes that the overall change in elasticity (i.e., self and cross-elasticity)
over the time horizon should be zero for the optimal load redistribution. This indicates
that whatever demand is curtailed using self-elasticity should be adjusted in the cross-time
periods through the cross-elasticities. This defines a lossless DR situation (i.e., no change in
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 25

the consumption BDR and ADR). Equation (3.10) advocates that an interrelation should
exist between self and cross-elasticity. To keeping this in view, a dynamic elasticity model
attributing the features of both elasticities is envisioned.
The basis of the proposed dynamic elasticity is to make DR adaptive. Therefore, a dynamic
elasticity is proposed for exhibiting the customer’s flexibility from peak hours to the low-
price hours. It manifests an interrelation between peak hour elasticity to off-peak/valley
hours. As, the shifted demand during off-peak/valley hours is a reflection of the curtailed
peak demand. Similarly, it can be asserted that elasticity during off-peak/valley hours
will be a reflection of peak hours’ elasticity. For this assertion to stand, the product of
elasticity and demand at any hour t is assumed to be equal to the product of elasticity and
demand at the peak hour. This gives elasticity value relative to the peak hour elasticity
defined as follows:

εk,P
i
E
(t)dki,t,o = εk,P
i
E
(tp )dki,o,tp ∀t ∈ ΩT \{tp } (3.11)

Equation (3.11) defines the elasticity relationship relative to peak hour’s elasticity with
the corresponding demand dki,t,o . If demand and elasticity at peak hour are known, then
the elasticity of the other time states can be evaluated.
The determination of peak hour is another complex task due to the different load patterns
of customer classes. It is presumed that peak price and peak demand typically coincide
with a few exceptions. It is due to the likelihood of a strong correlation between price
and load [97]. Under such case, if peak hour elasticity value is known, then other hours’
elasticities can be calculated relatively to peak hour as follows:
" #
max(dki,t,o )
εk,P
i
E
(t) = εk,P
i
E
(tp ) ; ∀t ∈ ΩT (3.12)
dki,t,o

Then, ADR demand of ith customer of class k at time t is given by the following expression.
" # !
max(dki,t,o ) ρkt − ρkt,o + λcl
i
dki,t = dki,t,o + εk,P
i
E
(tp ) dki,t,o ;∀t ∈ ΩT , ∀i ∈ ΩI (3.13)
dki,t,o ρkt,o

One of the main features of the proposed dynamic elasticity model is that it is a time-
variant model. In addition, it encapsulates the attributes of self and cross-elasticity in
the dynamic elasticity. It can be understood through the distinct trait of elasticity. Since
self-elasticity during off-peak hours will be relatively low due to the small relative change
in the price [55], which makes increase in demand small. Thus, the cross-elasticity should
cause an increase in the demand during the off-peak hours.
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 26

3.2.2 Proposed Stochastic PEM (SPEM)

In this section, elasticity is modeled using the stochastic approach. It is discussed in

the preceding section that self and cross-elasticity should be interrelated to make DR
adaptive. This interrelation can be described through both elasticities using the complete
PEM matrix as expressed in (3.14).
    
∆d1 εP E (1, 1) εP E (1, 2) · · · εP E (1, ΩT ) ∆ρ1
    
 ∆d2   εP E (2, 1) εP E (2, 2) · · · εP E (2, ΩT )  ∆ρ2 
= (3.14)
    
 .. .. .. .. ..  .. 

 .  
  . . . . 
 . 

∆dΩT εP E (t, 1) εP E (t, 2) · · · εP E (t, ΩT ) ∆ρΩT

Equation (3.14) describes the change in demand ∆dt via a corresponding elasticity (ε(t, τ )).
It is resulted due to a change in price ∆ρt at time t. This change in the demand at
time t depends on the change in the price of period t and other cross time-periods τ . If
interrelation between both elasticities holds perfectly, then demand will be lossless in DR.
It means that the overall change in the demand (i.e., ∆dt ) over a time horizon will be zero
[15].
X X X
∆dt = εP E (t, τ )∆ρt dt,o = 0 ∀t ∈ ΩT , ∀τ ∈ ΩT (3.15)
t∈ΩT t∈ΩT τ ∈ΩT

Equation (3.15) is the complete form of (3.10) suggested by [98]. It derives the optimal load
distribution in terms of the change in demand as opposite to (3.10), where only elasticity
is used to obtain the optimized load. It clearly states that the summation of the change
in demand over a time span should be zero. If it does not satisfy that condition, then the
following relation holds:
X
∆dt < 0; ∀t (3.16)
t∈ΩT

Equation (3.16) shows the overall reduction in the demand ADR, indicating the partial load
recovery of the curtailed demand. From the DR perspective, cross-elasticity is responsible
for imitating the curtailed/shifted demand in the cross-periods as a load recovery. However,
load recovery in the cross-periods is also bounded by the flexibility of intertemporal time
constraint. This constraint stipulates that the customer can recover its curtailed demand
to an extent in the cross-periods from the curtailed demand hour [66, 99]. Moreover, it
may increase or decrease in the subsequent periods based on the customer’s reaction to
price. A similar analogy can also be emulated for the cross-elasticity value. Since the
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 27

quantification of cross-elasticity over time is cumbersome due to uncertainty associated

with the customers’ activity usage pattern and price. Hence, cross-elasticity is simulated
using the stochastic process. It is modeled using a geometric Brownian motion (GBM)
to estimate cross-elasticity values over time in PEM [100, 101]. It is an extension of
Brownian motion and is exclusively utilized to predict the stock price as the price being
positive [101]. Likewise, the cross-elasticity is always positive from DR point of view.
A stochastic process {ε(t)}t≥0 is said to be GBM, if it satisfies the following stochastic
differential equation (SDE) expressed as follows:

dεP E (t) = µP E εP E (t)dt + σ P E εP E (t)dW (t) (3.17)

where, εP E is defined as the stochastic variable of cross-elasticity with drift parameter µ

as a propagation of the stochastic variable and volatility parameter σ as a variation in
the propagation. W (t) is called the Weiner/Brownian variable. It satisfies the following
conditions: 1) W (t0 ) = 0 and 2) W (tk ) = W (tk−1 ) + N (0, tk − tk−1 ); where, N being
a normal distribution with zero-mean and variance equal to (tk − tk−1 ). If t0 < t1 <
.... < tk−1 < tk < ΩT are the successive interval in the time horizon, then increments
W (tk )−W (tk−1 ) ∼ N (0, tk −tk−1 ) are the stationary and independent normally distributed
random variables. On solving (3.17), the following expression is obtained.

σ2
  
PE
ε (t) = εP0 E PE
exp µ − PE
t + σ W (t) (3.18)
2

where, εPo E is the initial value at t = 0. The generalized GBM expression is given by
(3.19).

2
" ! #
(σ P E ) p
εP E (tk+1 ) = εP E (tk ) exp µP E − ∆tk + σ P E ∆tk N (0, 1) (3.19)
2

The solution in (3.19) is obtained using Ito’s formula [101,102]. It represents a time- based
cross-elasticity in the exponential form. The successive ratios of cross-elasticity in the time
interval are independent random variables.
The solution of GBM characterizes the stochastic process using the drift µP E and volatility
parameters (σ P E )2 . These two parameters govern the movement of the stochastic process.
 2

If µP E ≤ (σP E ) /2 , then stochastic process decay exponentially in the successive time
intervals, which is equivalent to a decreasing cross-elasticity. It gives a partial load recovery.
 2

Conversely, for µP E > (σP E ) /2 , the stochastic process grows exponentially, which offers
an increasing cross-elasticity in the cross-periods. This in turn give the higher load recovery
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 28

than the curtailed demand. It may result in a rebound effect in DR (i.e. when a new peak
demand higher than the base peak emerges in the post-DR period) [103].

3.3 Disaggregated Load Profiling

In distribution systems (DS), load patterns of the different customer classes typically
demonstrate their activity usage, societal aspects, etc. It also varies within the customer
class for each individual owing to their social, economic, financial, other exogenous factors,
etc [104]. This manifests the diverse variations in load patterns within the same class,
which will also exhibit the discrete variability in DR [95]. Therefore, diversified load profiles
are synthesized to simulate each customer’s discrete response in DR. It is modeled using a
disaggregated load modeling approach. First, a mean load factor of the customer classes is
considered. Then, the expected variation in load factor at each hour in a considered class
is modeled using a coefficient of variation (COV). Based on COV, the standard deviation
at each hour is calculated as

σtk,LF = COVk,LF
t × µk,LF
t (3.20)

where, µk,LF
t , σtk,LF , and COVk,LF
t denote mean load factor, standard deviation, and
COV of the customer class k at time t, respectively. Using mean and standard deviation,
the permissible interval of load factor variation within each customer class is obtained as
follows:

k ∈ [ak , bk ] ∈ [µk,LF − 3σ k,LF , µk + 3σ k,LF ]

LFi,t t t t t t t
(3.21)
∈ [µk,LF
t − 3COVk,LF
t × µk,LF
t , µk,LF
t + 3COVk,LF
t × µk,LF
t ]

k is denoted as load factor of ith customer of class k at time t. In order to

where, LFi,t
k , a truncated normal distribution is utilized.
restrict the value of randomly generated LFi,t
The probability distribution function (PDF) of the truncated normal distribution for each
class of customers is expressed as follows [105]:

 0 k < ak
ifLFi,t t


ϕ(µk,LF ,σtk,LF ;LFi,t
k )

Ψ(µk,LF
t , σtk,LF , akt , bkt ; LFi,t
k
)= t k ≤ LF k ≤ bk
ifLFi,t t t
 Φ(µk,LF
t ,σtk,LF ;bkt )−Φ(µkt ,σtk ;akt )
k > bk


 0 ifLFi,t t

(3.22)
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 29

To generate a variable from a truncated normal distribution, the following approach is

employed

repeat :
x = rand()
(3.23)
k
LFi,t = Φ−1 (µk,LF
t , (σtk,LF )2 , x)
until(akt ≤ LFi,t
k ≤ bt )
t

To sample the load factor, a random number x is first generated and is evaluated using an
inverse cumulative distribution function (cdf) (Φ−1 (.)). Then, a sampled load factor value
is checked against the condition. If it violates the bounds limits, repeat the process. Once
load factor is calculated, then the load profile of the customers at each hour is synthesized
using the following expression.

dki,t,o = LFi,t
k
× dki,o (3.24)

3.3.1 DR Assessment Index

The impact of DR on the class-wise customers is assessed by observing the proportional

change in demand in the considered time-states BDR and ADR. It is denoted by an index
called demand flexibility (DF). It is expressed in terms of percentage as follows:
 
,
X   X
DF =  dkt,o − dkt dkt,o  × 100% (3.25)
 
t∈ΩTp/v/op t∈ΩTp

where, dkt,o and dkt denote the class-wise k demand BDR and ADR. The class-wise demand
k is obtained by aggregating all customers’ demands.

3.4 Simulation Results and Discussions

The effectiveness of the proposed PBDR modeling approaches is investigated on a 33-bus

test distribution load data. The proposed models are developed in MATLAB ® platform.
The load data information is taken from Ref. [106] and is given in Table A.1 of Appendix
A. For class-wise DR assessment, the load buses are segmented into five different customer
classes, i.e., residential: (R), commercial: (C), agriculture: (A), large industrial: (LI),
and medium industrial (MI) customers. Each load bus is assigned discretely to individual
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 30

customer class, and the available load bus is stipulated as their demand level for the ease
of exposition. The demand proportions of each customer class are set according to energy
share in the Indian energy perspective [107]. The mix has a total 235 number of customers.
The detail of assigned nodes, total allocated demand, number of customers, S/CF, and
customer demand range for each class can be referred from Table B.1 in Appendix B.
The class-wise load factor is given in Table B.3 in Appendix B and is referred from
Ref. [108]. The COV of each customer class is assumed to be set normally distributed
in the range of (0-30) %. A sample of diversified load profiles for residential customer class
is shown in Fig. 3.1 (a). The employed dynamic price (RTP) for DR analysis is shown
in Fig. 3.1 (b). It is taken from Ontario Energy Board [109]. The class-wise rates are
obtained using ρkt = ρt × (1 + κk ), where κk is the subsidy/charging factor (S/CF). It
models the distinct pricing rates to the customer class in the practical DS. These discrete
rates imitate the co-existing demand diversity, activity usage, and cross-subsidy [110].
The time span is partitioned into three periods namely: off-peak (0:00-7:00, 23:00-24:00),
valley: (12:00-17:00) and peak: (8:00-11:00, 18:00-22:00) for the sake of simplicity.

15 30
Load Demand (kW)

Price (¢/kWh)

25
10
20
5
15

0 10
1 3 5 7 9 11 13 15 17 19 21 23 1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour) Time (Hour)
(a) (b)

Figure 3.1: (a) A sample of load profiles of R Class customers; (b) dynamic price profile
under RTP

The class-wise elasticity value for R, LI, MI, C and A customers are -0.30, -0.43, -0.54,
-0.30 and -0.23, respectively as reported in Ref. [111]. The standard PEM-based DR ap-
proach is utilized for the comparison analysis. The elasticity values in the case of DPEM
and SPEM are evaluated using the proposed dynamic and stochastic models. Though,
peak hour elasticities for DPEM and SPEM are the same as the standard PEM, while
in the case of SPEM, the initial value of cross elasticity is assumed as the fraction of
self-elasticity. It is based on the notion of the low value of cross-elasticity [9, 10, 14]. The
cross-elasticity initial value is taken as (ε(t, τ ) × 0.15|t=τ ). The GBM parameters, µP E
and σ P E are taken as 0.2 and 1.2, respectively, for the analysis purpose.
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 31

R LI MI C A Avgε
-5 -1.0

Average Elasticity
-4 -0.8

Elasticity
-0.6
-3
-0.4
-2
-0.2
-1 0.0
0
1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour)

Figure 3.2: A aggregated class-wise elasticity variation using the proposed PEM

The aggregated class-wise elasticity profiles using the proposed DPEM are shown in
Fig. 3.2. The figure shows that each customer class exhibits discrete elasticity patterns
owing to the different load patterns. It displays the nominal variations in elasticity for R
and LI classes, while MI, A and C class customers exhibit wide variations. The rationale
behind such a vast elasticity variation can be inferred from the customer aggregated load-
/price pattern for more clarity. It shows that the customer class with a high demand/price
ratio between peak to valley or peak to off-peak set forth wide elasticity variation and vice-
versa. These elasticity variations indicate that load patterns with high relative ratios will
need to exhibit higher load flexibility to achieve true DR. Otherwise, they may face an
overall reduction in their energy consumption. On the contrary, the customers with nomi-
nal proportions among the different periods can easily display the load flexibility over the
time-horizon.

BDR_R PEM_R DPEM_R SPEM_R

1000
Load Demand (kW)

800

600

400

200
1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour)
(a) (b)

Fig. 3.3 (a)-(e) shows the aggregated class-wise customer demand for load BDR and ADR
using the PEM, the proposed DPEM, and SPEM model. It can be observed from Fig. 3.3
(a)-(e) that PEM exhibit DR behaviour persuasively during peak hours. However, its
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 32

(c) (d)

(e)

Figure 3.3: Aggregated class-wise load patterns BDR and ADR using PEM, and the
proposed PEM models for: (a) R, (b) LI, (c) MI, (d) C, (e) A

recovery during off-peak/valley hours is partial. The figures show that R, C, and MI cus-
tomers exhibit a partial increase in demand during off-peak hours. Though, LI and A
customers display much better load adjustment during the off-peak hours. On the con-
trary, the proposed DPEM and SPEM display much better curtailment and shifting for all
the customers’ classes. In case of DPEM, dynamic elasticity makes the curtailed load to
recover in the off-peak and valley periods, indicating complete load recovery. In addition,
the proposed DPEM is showing its feasibility to all classes’ load patterns. Likewise, the
proposed SPEM also exhibits better load recovery of the curtailed load using the stochastic
process accompanied by the inter-temporal constraint.
Further, the class-wise DR participation is better demonstrated via DF, as shown in
Fig. 3.4. Fig. 3.4 (a) depicts that R, C, and MI classes customers partially shift the
curtailed demand, while LI and A classes customers exhibit close to the curtailed load
demand. This partial load adjustment indicates the existence of a high price/demand
ratio between the segmented periods. It necessitates that elasticity cannot be static. It
should rather be adaptive or dynamic, as proposed in the the DPEM and SPEM based
DR model. In DPEM, a dynamic elasticity shows better load adjustment. It exhibits the
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 33

Off-peak Valley Peak Off-peak Valley Peak

20 20

Curtailment/Shifting (%)
Curtailment/Shifting (%)

15 15

10 10

5 5

0 0
R LI MI C A R LI MI C A
Customer Classes Customer Classes
(a) (b)
Off-peak Valley Peak
20
Curtailment/Shifting (%)

15

10

0
R LI MI C A
Customer Classes
(c)

Figure 3.4: Class-wise curtailed and shifted demand under RTP using (a) PEM, (b)
DPEM, and (c) SPEM

nearly equal percentage change in the curtailed and shifted demand for each customers’
class as demonstrated in Fig. 3.4 (b) as opposite to PEM. Similarly, the SPEM gives better
load recovery using the stochastic process, as can be observed from Fig. 3.4 (c). Though,
it does not maintain equal energy consumption BDR and ADR due to stochasticity.
From the economic point of view, an aggregated class-wise customer electricity bills BDR,
NDR and ADR for the considered states using the proposed DPEM framework are sum-
marized in Table 3.1. Here, BDR represents before DR state at the flat rate, NDR is the
state after DR implementation with zero participation and ADR is the state after DR with
participation. The table indicates that peak hours bills are increased as the customers shift
from BDR to NDR state, whereas the electricity bills during the valley and off-peak periods
are decreased. On the other hand, the customers bills are significantly reduces for peak,
valley and off-peak periods under ADR condition. The results suggest that shifting from
FR to RTP with zero DR participation increases the customer bills remarkably. However,
with DR participation, the customers bills are changed notably ADR. It increases during
peak while decreases during valley and off-peak periods relative to BDR. This indicates
that participating in DR will certainly benefits the customers under PBDR application.
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 34

Table 3.1: Aggregated customer class bills under the considered states using the pro-
posed DPEM

Case Class R (¢) LI (¢) MI (¢) C (¢) A (¢) Total (¢)

Peak 115110.26 210234.26 32671.00 148834.80 36303.67 543153.99
BDR Valley 72379.75 147044.53 24710.61 109778.81 7825.74 361739.44
Off-peak 56868.55 195118.28 11800.84 28230.88 43946.21 335964.76
Peak 148895.83 269854.36 41967.17 196242.43 46819.97 703779.75
NDR Valley 65993.88 135789.43 22722.46 100258.95 7188.07 331952.80
Off-peak 45681.41 149590.10 9844.97 22533.73 34855.45 262505.66
Peak 132888.01 221252.16 37757.22 163231.89 41393.43 596522.71
ADR Valley 67624.62 140740.60 23151.34 103621.78 7740.88 342879.22
Off-peak 52412.90 170027.95 11615.31 36415.08 37137.38 307608.62

The aggregated class-wise bills under BDR, NDR and ADR conditions using DPEM are
summarized in Table 3.2. It can be observed from the table that all customers class’ bills
are increased with zero DR contribution under NDR state. It is in the varying magnitude
under the different customers classes. On the other hand, the customers bills are signifi-
cantly reduces under ADR state relative to NDR condition. It is further verified with the
relative change in the bill, which indicates the change in the bill ADR relative to BDR
and NDR in the former and latter parts.

Table 3.2: Aggregated class-wise customer electricity bills using the proposed DPEM

Class R LI MI C A Total Bill

BDR (¢) 244358.56 552397.07 69182.46 286844.48 88075.63 1240858.20
NDR (¢) 260571.12 555233.90 74534.61 319035.11 88863.49 1298238.21
ADR (¢) 252925.53 532020.72 72523.87 303268.76 86271.69 1247010.56
Change (%) 6.63/-2.93 0.51/-4.18 7.74/-2.70 11.22/-4.94 0.89/-2.92 4.62/-3.95

In addition, the aggregated class-wise bills using the SPEM are also tabulated in Table 3.3.
The summarized results display similar pattern to the DPEM approach. Though, there is
slight variation in the change in the bills due to feasible load recovery in DR.

Table 3.3: Aggregated class-wise customer electricity bills using the proposed SPEM

Class R LI MI C A Total Bill

BDR (¢) 244358.56 552397.07 69182.46 286844.49 88075.63 1240858.20
NDR (¢) 260571.12 555233.90 74534.61 319035.11 88863.49 1298238.21
ADR (¢) 255822.79 534970.97 72648.70 303232.66 88164.51 1254839.63
Change (%) 6.63/-1.82 0.51/-3.65 7.74/-2.53 11.22/-4.95 0.89/-0.79 4.62/-3.34
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 35

The proposed PEM-based DR method is also investigated under the different pricing in
the subsequent section.

3.5 Proposed Incentive-based DR using PEM

The standard PEM based IBDR model is extension of PBDR model. It is obtained by
including an incentive/ penalty term to the standard PBDR model and is expressed as
follows:

(ρt − ρt,o + ρIN C)


t
dt = dt,o 1 + εP E (t) (3.26)
ρt,o

where, dt,o and dt are demands BDR and ADR, respectively. εP E (t), ρt,o and ρt denote
price elasticity, FR, and dynamic rate, respectively. ρIN
t
C is an incentive rate. Equation

(3.26) indicates that higher demand will be induced ADR with the addition of incen-
tive/penalty term in the numerator. However, elasticity value in the case of the standard
IBDR model remains same even after the inclusion of incentive term compared to the
PBDR model as described in the literature [9–12,112]. Since the incentive/penalty term is
equivalent to price [19, 21]; hence its addition will modify the overall elasticity compared
to price elasticity in IBDR model. Hence, two explicit elasticities, price and incentive elas-
ticity are proposed based on their price and incentive term. The proposed IBDR models
are interchangeably addressed as IBDR or augmented DR or mix DR (PBDR and IBDR)
model in the proposed study.
IBDR programs are usually categorized into the voluntary and mandatory programs [10,
11]. The voluntary programs offer incentives to the customers for load adjustment dur-
ing peak hours, whereas mandatory programs devise incentives in the form of a binding
contract along with a penalty provision for the agreement violation. In IBDR, voluntary
programs are primarily targeted to the small-capacity customers such as residential, com-
mercial, and irrigation customers [113], and mandatory programs are designed for large
industrial (or commercial) customers [7]. The complete incentive/penalty function for the
participating in IBDR is proposed as follows:

ICF (∆dki,t ) = ICFV P (∆dki,t ) × ξt + ICFM P (∆dki,t ) × (1 − ξt ); ξt ∈ {0, 1} (3.27)

where, ICF (.) denotes the customer incentive/penalty cost function. It consists of two
components, namely, voluntary and mandatory-based cost functions. A binary variable ξt
is defined to associate the customer class with one of the functions only. The expanded
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 36

forms of both incentive functions are described in (3.28) and (3.29).

ICFV P (∆dki,t ) = ρk,IN C k k

t,V P × (di,t,o − di,t ) (3.28)

Equation (3.28) defines the incentive cost function associated with the voluntary-based
program. It is the product of incentive cost coefficient and curtailed demand under vol-
untary IBDR programs.
h
ICFM P (∆dki,t ) = ρk,IN
t,M P
C
× (dk
i,t,o − dk ) × δ + (ρk,IN C + ρk,P EN ) × (dk
i,t t t,M P
k
i,t,o − di,t )
it,M P
−ρk,P EN k
t,M P × di,t,exp ) × (1 − δt ), δt ∈ {0, 1}

(3.29)

(
1 if (dki,t,o − dki,t ) ≥ dki,t,exp
δt = (3.30)
0 if (dki,t,o − dki,t ) < dki,t,exp

where, ICFM P (.) is defined as incentive/penalty cost function related to mandatory IBDR
program. It consists of two terms, reward and penalty cost, where reward/incentive cost
for adhering to the obligation, whereas penalty cost for the agreement violation. The
applicability of function is defined through a binary variable (δt ). It is explained in (3.30),
which indicates the acceptance level of obligation with the states of δt . The proposed
(3.29) is a modified form of Ref. [11], which imposes the complete penalty even on a
partial obligation. However, it is not appropriate to apply the full penalty even on the
partial demand curtailment as the customers’ response might be vague. Hence, the total
cost is defined in the mixed form of incentive and penalty as illustrated in the second
term of (3.29). In addition, incentive and penalty rates are assumed to vary instead of the
standard convention of having pre-determined values of incentive and penalty. It is taken as
a fraction of the difference of dynamic price and FR. The expressions of incentive/penalty
rates are defined as follows:

ρk,IN C (∆dki,t ) = λk,IN C (ρkt − ρkt,o )

(3.31)
ρk,P EN (∆dki,t ) = λk,P EN (ρkt − ρkt,o )

where, incentive/penalty rates are multiplied by a positive parameter λk,IN C /λk,P EN .

These controlling parameters reflect the volatility of wholesale price fluctuation on the
incentive/penalty rate offered to the customers.
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 37

3.5.1 Proposed IBDR Model with Dynamic Rate Charging

The modeling of the standard IBDR using PEM is an extended form of price-based PEM,
where an additional term, incentive, or penalty function is introduced as described in
(3.27). Under this condition, the customer’ net benefit function is expressed as follows:

S(dki,t ) = B(dki,t ) − dki,t (t)ρkt + ICFV P (∆dki,t ) × ξt + ICFM P (∆dki,t ) × (1 − ξt ) (3.32)

Equation (3.32) denotes the customer’s net benefit under the dynamic rate with incen-
tives/penalty. The optimality condition of (3.32) gives:

∂B(dki,t ) n h i o
= ρkt + ρk,IN
t,V P
C
× ξ t + (1 − ξt ) × ρk,IN C
t,M P × δ t + ρk,IN C
t,M P + ρk,IN C
t,M P × (1 − δt )
∂dki,t
(3.33)

Substituting incentive terms in the compact form as follows:

n h i o
ρk,IN
t
C
= ρk,IN
t,V P
C
× ξ t + (1 − ξt ) × ρk,IN C
t,M P × δ t + ρk,IN C
t,M P + ρk,P EN
t,M P × (1 − δ t ) (3.34)

Differentiating (3.5) and equating to (3.33), gives the customer’ demand ADR at time t
as follows:
( )
(∆ρkt + ρk,IN C
)
dki,t = dki,t,o k
1 + εi (t) k
t
(3.35)
ρt,o

where, (ρkt − ρkt,o ) is substituted with ∆ρkt , and ρk,IN

t
C
is an incentive/penalty rate. This
defines the complete composite standard IBDR model as follows:
 
k k,IN C 
 X (∆ρτ + ρτ )
dki,t = dki,t,o 1 + εki (t, τ ) k
(3.36)
 ρτ,o 
τ ∈ΩT

k and elasticity ′
Let’s suppose an additional rate of incentive changes the demand by ∆di,t
by ∆εki (t), then (3.36) can be re-written as follows:
 
k k,IN C 
′
 X k,P E (∆ρτ + ρτ )
dki,t + ∆di,t
k
= dki,t,o 1 + (εi (t, τ ) + ∆εki (t, τ )) k
(3.37)
 ρτ,o 
τ ∈ΩT
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 38

Now subtract (3.37) from the standard price-based DR with λki = 0 in (3.7). This gives
k ) as follows: ′
the change in demand (∆di,t

′ ρk,IN
τ
C
(∆ρkτ + ρk,IN
τ
C
(τ ))
εk,P E
X X
k
∆di,t = dki,t,o i (t, τ ) + dk
i,t,o ∆ε k
i (t, τ ) (3.38)
ρkτ,o ρkτ,o
τ ∈ΩT τ ∈ΩT

Performing some manipulation to obtain the change in elasticity ∆εki (t) as follows:
′k
ρkt,o ∆di,t ρk,IN
τ
C
εk,P E
X X
∆εki (t, t) = − dki,t,o i (t, τ ) − dki,t,o ∆εki (t, τ )
dki,t,o (∆ρkt + ρk,IN
t
C
) τ ∈ΩT (∆ρkτ + ρk,IN
τ
C
) τ ∈ΩT

(3.39)

Neglecting the effect of cross-elasticity and further simplification gives ∆εki (t) as follows:

′
εk,P
i
E k
(t)∆ρkt ∆di,t εk,P
i,t
E
(t)ρk,IN
t
C
∆εki (t) = − (3.40)
∆dki,t (∆ρkt + ρk,IN
t
C
) (∆ρkt + ρk,IN
t
C
)

Now, re-writing (3.40) in terms of the relative ratio of incentive rate/price change and
demand ratio due to IBDR and PBDR. This gives ∆εki (t) as follows:
′
!
k
∆di,t ρk,IN C
  
k,IN C.
k k,P E
∆εi (t) = εi (t) × 1 ρ
1+ t × − t k (3.41)
∆ρk t ∆dki,t ∆ρt

where, ∆εki (t) is termed as incentive elasticity due to augmentation of incentive in PBDR.
Thus, it can be equivalently written as εk,IE
i (t) = ∆εki (t). This expresses the change in
elasticity due to incentive rate and enables the utility to quantify the impact of price and
incentives, independently.
 
k k,IN C 
 X k,P E (∆ρτ + ρτ )
dki,t = dki,t,o 1 + (εi (t, τ ) + εk,IE
i (t, τ )) k
(3.42)
 ρτ,o 
τ ∈ΩT

Equation (3.42) represents the proposed augmented DR model considering the effect of
price as well as incentive rate on the elasticity. It distinctively defines price and incentive-
based elasticity based on their respective terms, i.e., price and incentive rate. It denotes
that demand ADR evinces the conclusive effect of price and incentive rate in DR relatively
to the standard IBDR model (3.36)
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 39

3.5.2 Proposed IBDR Model with Modified Flat Rate Charging

The proposed IBDR model extends the standard IBDR model, where the customers are
offered incentives with FR charging rates. FR price strategy may seem to benefit the
customers with the additional incentives. However, it may affect the utility’s revenue,
which bore it as compensation cost to the customers [114]. Nevertheless, it can be one
of the reasons for using dynamic rates in addition to incentives to assimilate the impact
of the compensation cost. Though, it mitigates a true virtue of IBDR, which is seen
as a reward-wise program [9]. Hence, a modified flat rate (MFR) is suggested to model
IBDR using PEM. It may lessen the utility’s cost burden for providing incentives to the
customers. It is evaluated as follows:
 
′k ℘k X
ρt,o = ρkt,o + (ρkt − ρkt,o ) = ρkt,o + ∆ρ̂kt (3.43)
ΩTp t∈ΩTp

′
k is a MFR. It consists of two components; the first term is FR and the second
where, ρt,o
term is the mean of differential of dynamic price and flat rate multiplied by a propor-
tional coefficient ℘. This coefficient absorbs dynamic price volatility. Now, the equivalent
formulation of incentive-based PEM under MFR is expressed as follows:

′
S(dki,t ) = B(dki,t ) − dki,t ρt,o
k
+ ICFV P (∆dki,t ) × ξt + ICFM P (∆dki,t ) × (1 − ξt ) (3.44)

Equation (3.44) defines the customer net benefit function under DR with MFR charging.
The optimality condition is defined as follows:

∂B(dki,t ) ′k
n h i o
= ρt,o + ∆ρk,IN
t,V P
C
× ξ t + (1 − ξt ) × ρk,IN C
t,M P × δ t + ρk,IN C
t,M P + ρk,P EN
t,M P × (1 − δ t )
∂dki,t
(3.45)

Similarly, replacing incentive/penalty terms with ρk,IN

t
C
(t). On solving the customer’s
demand at time t using (3.5) and (3.45), gives the composite IBDR as follows:
 
k k,IN C 
 X (∆ρ̂τ + ρτ )
dki,t = dki,t,o 1 + εki (t, τ ) k
(3.46)
 ρτ,o 
τ ∈ΩT

where, εki (t) is elasticity due to MFR and incentive rate. It is to be noted that ∆ρ̂kt is price
differential term. Thus, it will be equivalent to ∆ρ̂kt = ∆ρkt . Now, to segregate individual
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 40

elasticity, (3.46) is re-written in reference to the standard price-based PEM as follows:

 
k k,IN C 
′
 X k,P E (∆ρ̂τ + ρτ )
dki,t (t) + ∆di,t
k
= dki,t,o 1 + (εi (t, τ ) + ∆εki (t, τ )) (3.47)
 ρkτ,o 
τ ∈ΩT

k ) due to incentive elasticity ′

Expanding (3.47) and evaluating the change in demand (∆di,t
gives the following expressions.
 
′k ∆dki,t (∆ρ̂kt + ρk,IN
t
C
)
εk,P E (t, t) + εk,P E
X
∆di,t = i i (t, τ ) +
εk,P
i
R
(t, t)∆ρ̂kt τ ∈ΩT
X (∆ρ̂kτ + ρk,IN
τ
C
)
dkt,o ∆εki (t, τ ) k
(3.48)
ρτ,o
τ ∈ΩT

Neglecting cross-elasticity gives the change in demand as follows:

′k ∆dki,t (t) ∆dki,t ∆εki (t)

∆di,t = (ρk,IN
t
C
+ ∆ρ̂kt ) + (ρk,IN
t
C
+ ∆ρ̂kt ) (3.49)
∆ρ̂kt εk,P
i
E
(t)

Re-writing (3.49) in terms of relative ratio gives incentive elasticity as follows:

′
!
k
∆di,t ρk,IN C
  
k,IN C.
k,IE k k,P E
εi (t) = ∆εi (t) = εi (t) × 1 ρ
1+ t × − t k (3.50)
∆ρ̂k t ∆dki,t ∆ρ̂t

Equation (3.50) denotes incentive elasticity in respect to incentive and MFR.

3.6 Economic Assessment of Utility and Customers under

DR

In DR, the expected load adjustment is usually performed against the quoted prices in
advance by the utility. It is performed to evaluate load adjustment with the available time
flexibility [14]. This brings the change in price followed by the change in demand put forth
by the utility. Similar observation will be drawn at the electricity market, where the pro-
cured market price will be changed due to the expected altered demand in the real-time
operation. These concatenations of events give rise to establish a feedback mechanism
to contemplate DR impacts on the different levels in the power system hierarchy. It is
illustrated using PEM by considering the effect of DR on the expected market clearing
price (MCP) at the wholesale market and its reflection on the utility’s dynamic price to
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 41

the customers. It is in continuation of price computation of the electricity market using

feedback mechanism in Ref. [14], where DR impact is observed in the electricity market
price. However, the variations in market price as well as on the utility’s price (i.e., retail
rate) is also considered in the proposed approach. It defines three different elasticities
to elucidate the variations in price at the customers, the utility, and the wholesale mar-
ket levels, respectively. In this context, first, the effect of DR on MCP variation in the
wholesale market is defined as follows:

∆dt ρM
t,o
εM (t) = (3.51)
∆ρM
t dt,o

where, εM (t) is defined as the market elasticity on account of variation in MCP and
demand at the market level and ∆d represents the expected sum of the change in demand
at the system level ADR. It is defined as follows:
X X
∆d = dki (3.52)
k∈ΩK i∈ΩIk

Similarly, the variation in the utility’s price due to the expected change in demand is
described through the utility elasticity (εU (t)). It is expressed as follows:

∆dt ρU
t,o
εU (t) = (3.53)
∆ρU
t dt,o

In a similar line, an aggregated customers elasticity at the utility level can be defined as
follows:

∆dt ρt,o
εAGG (t) = (3.54)
∆ρt dt,o

It is to be noted that the total expected change in demand in all the defined elasticities is
the same, whereas the change in price is different on account of the different base price in
the electrical hierarchy. This designates the distinct elasticity at each level. It is measured
relative to initial MCP, initial retail price, and FR corresponding to the wholesale market,
the utility, and the customer level. Now, the expected market price ADR can be evaluated
as follows:

∆dt ρM t,o
ρM M M M
t = ρt,o + ∆ρt = ρt,o + M
(3.55)
dt,o ε (t)

where, ρM
 M
t market price ADR. 1 ε (t) is called an inverse market elasticity. It is based on
the Ramsey rule of inverse elasticity for the optimal taxation in a monopolist market [115].
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 42

This inverse elasticity is only feasible when demand is regulated by its own price without
considering the cross-price effects. Similarly, the expected utility’s retail price ADR is
defined as following:

∆dt ρU t,o (t)

ρU U U U
t = ρt,o + ∆ρt = ρt,o + (3.56)
dt,o (t) εU (t)

where, 1 εU (t) is defined as an inverse utility elasticity. It is to be noted that the retail or
utility rate will only change for those pricing strategies, which directly reflect the wholesale
market price fluctuation on the retail rate in day-to-day operations such as RTP, VPP,
etc [116].

Figure 3.5: Effect of DR on MCP and utility price using PEM

Fig. 3.5 depicts a flow of the process to show the impact of DR on the market and the
utility level operation using PEM. The figure reveals that the system will follow up a
chain reaction from the wholesale market to the customers’ level under DR via a negative
feedback. Under DR, a sizable demand (d) will be expected to change. This modifies the
forecasted demand, giving a new expected demand, which will subsequently evince new
expected MCP at the market level. Consequently, a new dynamic price (i.e., expected
price) will be offered to the customers in response to the adjusted demand at the utility
level in the real-time operation.

3.7 Simulation Results and Discussions

In this section, a comprehensive assessment of PBDR and IBDR models is investigated

on the standard 33-bus test distribution system. The system’s peak demand capacity is
3715 kW [106]. The system’s data information and class-wise customers’ data are given
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 43

in Table A.1 of Appendix A and Table B.1 of Appendix B. In the proposed study, R, C,
and A classes are kept under the umbrella of the voluntary program, and LI & MI class
customers are under the mandatory program. The market clearing price, RTP, and TOU
are taken from Ontario Energy Board as shown in Fig. 3.6 and is scaled as per the analysis.

Market clearing price (¢/kWh)

30 MCP TOU RTP 25
Utility Price (¢/kWh)

20
25
15
20
10
15 5

10 0
1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour)

Figure 3.6: Market price, TOU and RTP

The time-span of the analysis is 24 hours in day-ahead with three price/demand segments
viz. off-peak (0:00-7:00, 23:00-24:00), valley: (12:00-17:00) and peak period: (8:00-11:00,
18:00-22:00), respectively [109]. Several cases have been deliberated based on pricing
offered to the customers. The considered pricing are RTP, TOU, variable peak pricing
(VPP) [116] and are tested on PBDR and augmented DR programs. The considered cases
are detailed in Table 3.4.
The class-wise customer price elasticity values are the same as conveyed in Section 3.5 [111].
Since incentive elasticity is not explicitly reported in the literature. Thus, it is obtained
using (3.41) and (3.50) based on the presumed relative ratio of demand and price of PB
 ′. 
and IB-based DR programs. The ratio ∆d ∆d is assumed to be 0.48, 0.4, and 0.7 for
case 6, case 5 & 7, and case 8 & 10, respectively in the analysis. The incentives coefficient
parameters λIN C IN C P EN
V P and λM P /λM P are set equal to 0.3 & 0.4/0.3 in case 5, 6, 7, 8 & 10
and 1.3 & 1.1/0.7 for case 4 & 9, respectively. The market elasticity is inferred based
on the expected price variation in the wholesale market price for the expected demand
curtailment from Ref. [97]. The utility elasticity is obtained using the wholesale and the
utility price mark-up rule. This mark-up ratio is kept between 1.4-1.5 [117]. The demand
responsiveness of the customers is set to be normally distributed between 0-30 %. The
elasticity values are calculated using the concept of dynamic elasticity, as explained in
Section 3.2.
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 44

Table 3.4: Case considered in PBDR and IBDR study

Program Pricing
Case Description
Type Strategy
C.0 BDR FR 20.79 ¢/kWh for all hours
C.1 RTP Shown in Fig. 3.6
PBDR 15.8, 20.7 and 25.8 ¢/kWh for
C.2 TOU
off-peak, valley and peak period respectively
TOU structure during valley and off-peak,
C.3 VPP
and RTP during peak periods.
C.4 Pure IBDR FR 20.79 ¢/kWh for all hours
C.5 RTP
C.6 TOU
TOU structure during valley and off-peak,
C.7 (PB+IB) DR VPP
and RTP during peak hours
MFR rate during peak, and FR during valley
C.8 MF+FR
and off-peak period
FR during peak, and RTP during valley
C.9 FR+RTP
and off-peak hours
FR during peak hours and RTP during valley
C.10 MFR+RTP
and off-peak hours

An aggregated price and incentive elasticity variation pattern for the various cases are
illustrated in Fig. 3.7 (a). It can be observed from the figure that each case exhibits a
varying degree of elasticity pattern over the time horizon. The price elasticity patterns
for cases 1, 2, 3, 5, 6 & 7 are almost alike. In case 8, price elasticity appears during
the peak period only and is nearly close to case 1. Case 9 exhibits low price elasticity
during the valley and off-peak periods. On the other hand, case 10 displays a similar
pattern to cases 1 and 2. In the case of incentive elasticity, cases 4 & 9 and cases 8 &
10 demonstrate a similar pattern to a greater extent. In case 6, a moderate incentive
elasticity is displayed. Cases 4 and 9 exhibit much higher incentive elasticity than cases
5 and 7. This designates that the effect of incentives will be somewhat smaller under the
augmented DR as observed for cases 5, 6 and 7. On the other hand, it becomes of much
significance for IBDR program as perceived from cases 4, 8, 9, and 10 in Fig. 3.7 (a). These
observations reveal a diminishing effect of incentive elasticity as a degree of dynamism in
the pricing rises.
A class-wise induced demand flexibility under the various considered cases is illustrated
in Fig. 3.7 (b). The figure illustrates that each class gives the different degree of demand
flexibility in DR. These diversification in the flexibility reveal the discrete perception of
the customer class towards DR. These variations further can be corroborated with the load
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 45

C.1 C.2 C.3 C.5 C.6 C.7 C.8 C.9 C.10

C.4 C.5 C.6 C.7 C.8 C.9 C.10
-2.0 -0.4 R LI MI C A

Demand flexbility (DF %)

30

Incentive Elasticty (εIE)

Price Elasticity (εPE)

-1.5 -0.3
20
-1.0 -0.2

10
-0.5 -0.1

0.0 0.0 0
1 3 5 7 9 11 13 15 17 19 21 23 1 2 3 4 5 6 7 8 9 10
Time (Hour) Cases
(a) (b)

Figure 3.7: (a) An aggregated price and incentive elasticity variations, (b) Class-wise
demand flexibility under the different cases

and price pattern along with elasticity. It is lowest for pure IBDR, moderate for PBDR,
and highest for mixed DR cases. It indicates that the cumulative effect of incentive causes
appreciable flexibility in the augmented DR relative to PBDR and pure IBDR.
The aggregated class-wise DR response for the various cases using DPEM is depicted in
Fig. 3.8 (a)-(e). It can be observed from the figures that all cases exhibit DR behaviour
effectively with the adaptiveness in load patterns. The degree of response under the
different DRs varies with the offered pricing strategies. It is highest for RTP and VPP,
moderate for TOU-based DR programs. Moreover, a mix of PB and IB-based DR displays
better response than PBDR. It is revealed from the figures that load patterns with nominal
ratio (i.e., LI, A) exhibit load shifting effortlessly relative to load patterns with high ratio
(i.e., R, MI, C).

C.0 C.1 C.2 C.3 C.4 C.5 C.0 C.1 C.2 C.3 C.4 C.5
C.6 C.7 C.8 C.9 C.10 C.6 C.7 C.8 C.9 C.10
1000 1500
Load Demand (kW)
Load Demand (kW)

800

600 1000

400

200 500
1 3 5 7 9 11 13 15 17 19 21 23 1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour) Time (Hour)
(a) (b)

The economic performance in terms of the total customers’ bills and the total profit under
the different cases are summarized in Table 3.5 and Table 3.6. Table 3.5 presents a class-
wise DR assessment under BDR, NDR and ADR condition. The results demonstrate that
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 46

C.0 C.1 C.2 C.3 C.4 C.5 C.0 C.1 C. 2 C.3 C.4 C.5
C.6 C.7 C.8 C.9 C.10 C.6 C.7 C.8 C.9 C.10
600
200
Load Demand (kW)

Load Demand (kW)

150 400

100
200
50

0 0
1 3 5 7 9 11 13 15 17 19 21 23 1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour) Time (Hour)
(c) (d)

C.0 C.1 C.2 C.3 C.4 C. 5

C.6 C.7 C.8 C.9 C.10
800
Load Demand (kW)

600

400

200

0
1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour)
(e)

Figure 3.8: Aggregated class-wise load patterns before and after DR under the various
cases using DPEM for: (a) R (b) LI, (c) MI, (d) C, (e) A

the class-wise customers and the overall bills are increased with zero DR contribution
under NDR condition relative to BDR state for all cases except the case 9. However, the
customers and total bills are significantly changed under ADR condition relative to NDR
condition for all cases. It benefits all customers classes once they start participating in DR
after DR implementation. In addition, the customers’ bills in overall exhibit remarkable
reduction in the subsequent cases except for cases 9 relative to NDR condition.
The profit gains from the different customers’ classes under the various cases are compiled
in Table 3.6. The results indicate that the profits are increased in most cases except for
cases 4, 7, and 9. However, it varies on the class-wise profit gains as observed from the
table. It is lowest for case 9 and highest for case 8. The incentive cost in the augmented
DR indicates that higher incentives rates are required for case 4 and case 9 relative to cases
5, 7, 8, and 10, when measuring DR at the same level. Overall, the utility’s profit remains
consistent with certain positive variations except in case 9. It is higher in PBDR and
moderates in mix DR due to the bearing of incentive cost as the compensation cost except
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 47

Table 3.5: Aggregated customer class bills under the different cases

Case(s) R (¢) LI (¢) MI (¢) C (¢) A (¢) Total (¢)

0 BDR 244358.56 552397.07 69182.46 286844.49 88075.63 1240858.20
NDR 260571.12 555233.90 74534.61 319035.11 88863.49 1298238.21
1
ADR 250827.12 539190.06 71502.98 298637.18 87259.23 1247416.58
NDR 256864.53 553120.75 73715.38 313662.58 85985.52 1283348.75
2
ADR 252630.05 540264.23 72601.74 304930.45 84550.06 1254976.53
NDR 263475.76 563110.40 75298.83 325934.46 87931.53 1315750.99
3
ADR 247653.97 526203.63 71006.55 299255.63 82852.77 1226972.56
NDR 244358.56 552397.07 69182.46 286844.49 88075.63 1240858.20
4
ADR 236983.31 534319.54 67242.81 271635.60 85575.47 1195756.73
NDR 260571.12 555233.90 74534.61 319035.11 88863.49 1298238.21
5
ADR 232711.94 492614.80 66633.79 266039.40 81346.09 1139346.02
NDR 256864.53 553120.75 73715.38 313662.58 85985.52 1283348.75
6
ADR 240445.21 502925.72 69397.21 279803.45 80419.48 1172991.08
NDR 263475.76 563110.40 75298.83 325934.46 87931.53 1315750.99
7
ADR 229538.79 479628.37 66137.36 266657.86 76939.62 1118902.00
NDR 267342.16 594373.71 75705.75 316561.73 95324.23 1349307.58
8
ADR 242527.32 518520.16 69179.60 265389.75 86912.18 1182529.01
NDR 226785.55 495613.79 65238.44 271627.48 78347.19 1137612.45
9
ADR 223838.27 491775.48 64351.86 263486.43 77814.76 1121266.80
NDR 249769.16 537590.43 71761.73 301344.73 85595.80 1246061.84
10
ADR 239136.69 507206.58 68627.69 273166.52 83405.64 1171543.12

for case 8. The utility’s dynamic price variation BDR and ADR are shown in Fig. 3.9. It

Table 3.6: Aggregated class-wise utility’s profit under the different cases

Case(s) R (¢) LI (¢) MI (¢) C (¢) A (¢) Profit (¢) Incentive (¢)
0 64982.38 170037.13 17961.44 67056.01 26847.08 346884.04 0
1 78061.54 167540.60 22343.56 92750.39 27094.17 387790.26 0
2 77300.83 162929.22 22753.49 94346.10 23297.76 380627.41 0
3 75429.45 156567.91 21950.54 94831.06 22605.06 371384.01 0
4 64648.34 168144.03 17907.33 64750.27 26657.37 342107.33 12848.68
5 72068.29 151628.24 20730.40 81889.63 25136.08 351452.63 3652.29
6 77095.73 156251.66 22804.49 90243.73 23089.18 369484.79 2617.17
7 69436.19 140655.55 20337.37 83970.30 20646.98 335046.38 3652.29
8 88770.99 202752.65 24921.69 90906.18 34099.80 441451.31 3900.40
9 48434.13 115725.82 14287.75 51705.83 17531.27 247684.80 12848.68
10 76086.63 160654.84 22160.26 83609.93 26432.77 368944.43 3900.40

indicates that RTP and VPP in real-time operation will change ADR. It decreases during
peak hours owing to demand curtailment and reflects during the valley and off-peak periods
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 48

due to increased demand. The average decrease in the price is around 4.56 % in PBDR
and 7.26 % in the augmented DR during the peak period. The average increase in the
off-peak price is 2.1 % in PBDR and 3.65 % in the augmented DR programs, respectively.

RTP C.1 C.2 C.3 C.4 C.5 C.6

30 C.7 C.8 C.9 C.10

25
Price (¢/kWh)

20

15

10
1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour)

Figure 3.9: Utility Price variation with DR application under the different cases

An individual DR contribution in PB and IB DR programs via peak demand curtailment

are illustrated in Fig. 3.10 (a). The figure indicates that PBDRs mostly contribute in DR.
It is also visible in the augmented DR programs, where PBDRs dominate over IBDRs.
Nevertheless, the contribution ratio may change subject to incentives offers.

PBDR IBDR Total

-0.6
Incentive elasticity (εiIE)
DR contribution (%)

20 -0.4

10 -0.2

0 0.0
1 2 3 4 5 6 7 8 9 10 0 5 10
Cases ( ∆ρ ’ ∆ρ )
(a) (b)

Figure 3.10: (a) Aggregated Contribution of PBDR and IBDR programs under the
different cases; (b) Variation in incentive elasticity with variation in relative ratio of
incentive/change in price

Fig. 3.10 (b) illustrates the effect of the augmented DR program on incentive elasticity.
The figure depicts that incentive elasticity increases with an increment in the relative ratio.
It shows a noteworthy increase in early part, but became stagnant in the latter part of
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 49

the curve. This indicates that the customers exhibit gradual variation in DR as incentives
increase beyond a certain limit in the augmented DR program.

3.7.1 Difference between the Existing and the Proposed IBDR Modeling
Approach

This section investigates the adequacy of the proposed IBDR model to deliberate an in-
centive elasticity over the existing standard IBDR model. It is adjudged through elasticity
variation, induced demand flexibility, the customers’ bills and the utility’s profit. It is
tested for the standard augmented DR models i.e., cases 5, 6, 7, 8, which are also ab-
breviated by DR programs’ names shown in Table 7. Case 4 is considered to exemplify
elasticity scale only, as it is pure IBDR program.
The aggregated elasticity for 20:00 hour as exemplifying case using the existing and pro-
posed IBDR method is described in Table 3.7. It exhibits a higher elasticity value than the
standard IBDR model. In addition, a percentage variation indicates the effect of incentive
elasticity inclusion in the proposed IBDR model relative to the standard IBDR model.

Table 3.7: Elasticity evaluation using the standard and proposed IBDR model at 20:00
Hour

Standard IBDR Model Proposed IBDR model Variation (%)

Program (εAGG −εP E
Price Incentive Total AGG )
Elasticity = εP E
elasticity elasticity elasticity AGG
(εPAGG
E )
(εPAGG
E ) (εIE
AGG ) (εAGG )
FR+IB 0 0 -0.200 -0.200 -
RTP+IB -0.456 -0.456 -0.018 -0.474 3.98
TOU+IB -0.449 -0.449 -0.054 -0.502 12.00
VPP+IB -0.457 -0.457 -0.018 -0.475 3.97
MFR+IB -0.477 -0.477 -0.072 -0.548 15.05

The induced demand flexibility during peak periods using both methods is shown in Ta-
ble 3.8. The results indicate that the inclusion of incentive elasticity gives improved peak
demand curtailment than the standard IBDR model. It is further confirmed with a pos-
itive percentage variation, indicating the importance of having incentive elasticity due to
incentive rate in the augmented DR. The results suggest that the standard IBDR approach
gives the underestimated demand curtailment relative to the proposed IBDR model.
Moreover, an economic comparison using both DR methods is summarized in Table 3.9.
The table shows that the customers’ bills as well as the utility’s profit decrease using the
proposed IBDR model. The variation in the profit is found to be 3-4 % and 2-6 % in the
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 50

Table 3.8: Aggregated impact on peak demand curtailment using the standard and the
proposed approach

Peak Demand curtailment (kWh) due to IBDR Variation

Program
Standard IBDR model Proposed IBDR model (%)
RTP+IB 1257.64 1442.00 25.57
TOU+IB 1040.86 1432.15 54.64
VPP+IB 1257.64 1442.00 25.57

customers’ bills. This demonstrates that the standard IBDR model gives overestimated
economic results compared to the proposed IBDR approach.

Table 3.9: An economic comparison of the standard and the proposed augmented DR
during peak periods

Standard IBDR model Proposed IBDR model Variation (%)

Program
Profit (¢) Bills (¢) Profit (¢) Bills (¢) Profit Bills
RTP+IB 145403.73 466590.37 140734.45 452627.70 -3.21 -2.99
TOU+IB 178939.82 529682.47 175172.85 506716.78 -2.11 -4.34
VPP+IB 145403.73 466590.37 140734.45 452627.70 -3.21 -2.99
MFR+IB 161767.86 513561.76 155780.31 484824.80 -3.70 -5.60

3.8 Summary

An adaptive economical PBDR and an improved IBDR frameworks using PEM are pro-
posed in DS. The proposed PBDR models presents a dynamic elasticity to imitate the
customer’s load behaviour in DR. It has been modeled using the deterministic and stochas-
tic approaches. Both approaches envisage dynamism in the elasticity to make a feasible
load recovery in the cross-periods for the curtailed demand. The proposed DR modeling
approaches are thoroughly investigated on the various class-wise customers’ load patterns
in 33-bus test DS. The results illustrate an appropriate and pragmatic DR modeling com-
pared to the standard PBDR model. It is also extended in the augmented DR model using
the various pricing.
An improved economic IBDR puts forward a mathematical model to consider the effect
of incentive inclusion in PEM. It envisions an incentive elasticity equivalent to the price
elasticity to distinguish the price and incentive term effect on elasticity variation in the
augmented DR model. The comprehensive analyses have been carried out under the var-
ious pricing strategies to assess the applicability of incentive elasticity in the augmented
DR on 33-bus test DS. These pricing manifest a discrete impact on incentive elasticity.
The numerical analyses demonstrate the significant difference between the existing and the
Chapter 3. Price and Incentive Based DR using Price Elasticity Model 51

proposed augmented DR. It induces higher demand flexibility, consequently affecting the
economic and technical aspects. The demand curtailment is observed to be 25-60 % higher
than the standard IBDR model. This shows the conclusive effect of incentive leading to a
change in elasticity in the augmented DR.
Chapter 4

Price Based Demand Response

using Cardinal and Ordinal Utility
Functions

4.1 Introduction

Demand response (DR) modeling using the price elasticity model (PEM) has been a ver-
satile and appealing method for realizing the DR behaviour since its inception. It of-
fers mathematical simplicity and straightforwardness to understand. However, estimating
elasticities is a tedious and rigorous approach over the multi-states. Most importantly,
PEM-based DR model falls short in attributing the customer’s preference or willingness
uniquely in DR. In such instances, the utility functions-based DR modeling methods can
be considered a sophisticated DR approach. It can illustrates the customer’s willingness,
risk-behaviour and adaptability in DR [27, 29].
The utility functions (UFs) have been mainly described through two classes namely, car-
dinal utility functions (CUFs) and ordinal utility functions (OUFs) in DR. Both functions
exhibit the preference or willingness through their coefficient parameters. Among, CUFs
have been extensively applied to represent the various types of load models. However, their
coefficient attributes are hardly highlighted in the literature [25, 26, 30, 60]. Moreover, the
impact of the different structural functional forms on DR under the different load patterns
and price variation have not been critically assessed. On the other hand, OUF has been
effectively applied in DR modeling with its feature of minimal parameters. Though, its
applications in DR have been moderately discussed in the literature [27–29].

53
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 54

This chapter presents constrained economic PBDR frameworks using the CUFs and OUF.
The CUFs are employed in a social welfare function considering an energy consumption
constraint to model PBDR under the dynamic price. It is illustrated for the various CUFs
such as quadratic, exponential, logarithmic, and power functions. The CUF’s parameters,
satisfaction, and risk-aversion coefficients are correlated with the customer’s behavioral
attributes. Further, the impact of the different functional forms on DR are critically eval-
uated using the risk-based indices, different load patterns, parameters and price variation.
The OUF is utilized in a micro economic behavior model, overlapping generation model
(OLG) for PBDR modeling. It is a multi-state, discrete time-variant model, where the
customer changes its consumption with the change in the interest rate and degree of risk-
aversion. This analogy in DR is envisioned by associating the interest rate to the change
in relative price of the different periods and a degree of risk-aversion to the customer’s
willingness. The adequacy and effectiveness of proposed PBDR modeling approaches are
assessed on the standard 33-bus test load data. It is performed on the class-wise customers
to determine the DR response on the different load patterns. Further, PBDR models are
comprehensively assessed and are compared with the existing DR modeling approaches.

4.2 Proposed Constrained Economic PBDR Model using

CUFs

Cardinal utility is a function to measure the customer’s utility or satisfaction on consuming

the products. It measures the utility in terms of magnitude. The CUFs have the following
properties:

(i) CUF is a non-decreasing function (i.e., ∂U (d, w, β)/∂d ≥ 0).

 2
(ii) The marginal value of the CUF is a non-increasing (i.e., ∂U 2 (d, w, β) ∂d ≤ 0).

(iii) CUF gives zero value for no consumption (i.e., U (0, w, β) = 0).

(iv) The high value of satisfaction coefficient gives a higher utility value (i.e., U1 (d, w1 ) ≥
U2 (d, w2 ) ≥ U3 (d, w3 ) ≥ 0; ∃w1 ≥ w2 ≥ w3 ) [118].

(v) High value of risk-aversion coefficient gives high curvature in the utility function
shape (i.e., U1 (d, w1 , β1 ) ≥ U2 (d, w2 , β2 ) ≥ U3 (d, w3 , β3 ) ≥ 0; ∃β1 ≥ β2 ≥ β3 ) [119].

The aforesaid properties are explained using a quadratic cardinal utility function shown in
Fig. 4.1 (a)-(b). It can be observed from Fig. 4.1 (a) that the quadratic CUF is monotonic
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 55

and concave. It exhibits an increase in the overall utility with the successive consumption.
But its marginal utility decreases with increased consumption as shown in Fig. 4.1 (b).
Further, the trajectories of the utility function’s curves are distinct at the same consump-
tion level. It is due to the relative difference in the satisfaction value, which yields the
higher utility values for the increasing satisfaction coefficient.

Figure 4.1: Illustration of a quadratic utility function for: (a) customers’ utility values
(b) customers’ marginal utilities

The other common forms of the CUFs are exponential, logarithmic, and power utility
functions shown in Fig 4.2 (a). These CUFs are also defined as a non-decreasing concave
function with the diminishing marginal utility values depicted in Fig 4.2 (b). This indi-
cates that the CUFs typically possess a risk-averse behaviour [120].

Quadratic Exp Log Power Quadratic Exp Log Power

20 6

15 4
10
U’(d)
U(d)

2
5
0
0

-5 -2
0.10 2.80 5.50 8.20 10.90 13.60 16.30 0.10 2.80 5.50 8.20 10.90 13.60 16.30
d d
(a) (b)

Figure 4.2: Illustrations of: (a) the various utility functions, (b) marginal utility curves

It is worth noting that the shapes of the utility functions and their marginal utility curve
vary with their coefficient parameters. These coefficient parameters are called the satis-
faction coefficient w and risk-aversion coefficient β. Since the customers exhibit a varying
degree of participation in DR. Hence, the CUFs can be utilized to imitate the customer’s
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 56

behaviour in DR. The different functional forms of CUF are described for PBDR modeling
in following section.

4.2.1 Quadratic CUF Modeling

The CUF based PBDR modeling is described using a disaggregated load modeling ap-
proach [1]. The CUF is utilized using the social welfare or pay-off function in PBDR
problem. It is expressed as the difference between the customer’s utility function and the
amount paid for that utility. It is defined as follows:

2
(dki,t )
S(dki,t , wi,t
k k
, βi,t ) = k k
wi,t di,t − k
βi,t − ρkt dki,t (4.1)
2

where, S(.) defines the customer’s social welfare/pay-off function. dki,t , , wi,t
k and β k are
i,t
demand, satisfaction value, and risk-aversion of ith customer class k at time t. Now to
determine the possible satisfaction coefficient, a first-order condition (FOC) of (4.1) is
calculated. This gives an optimality condition as follows:

∂Υ(dki,t , wi,t
k , βk )
i,t k k k
= wi,t − βi,t di,t − ρkt (4.2)
∂dki,t

k k k
wi,t ≥ βi,t di,t + ρkt (4.3)

From (4.3), it can be observed that the satisfaction coefficient is dependent upon the
product of demand and risk-aversion coefficient and the price offered at that period. In
addition, it will be distinct during each hour. This variation in the satisfaction coefficient
can be correlated with the price and demand patterns. As in PBDR, peak and off-peak
hour prices are increased and decreased simultaneously for adjusting the load demand.
Similarly, the satisfaction coefficient can be lowered and raised simultaneously during peak
and off-peak hours for the curtailment and shifting. A similar viewpoint is also presented
in Ref. [31], where the authors have assigned a lower satisfaction value during peak hours
and a higher value during off-peak hours. It shows that the comfort level can be lowered
by reducing the satisfaction coefficients during peak hours and vice-versa. It is usually
perceived that if the load is high, the price will also be high or vice-versa [121]. However, if
the customer’s load pattern and price pattern are not cohesive, DR becomes trivial. Thus,
two indexes, load deviation index (LDI), and price deviation index (PDI) are proposed to
extract the information of both the attributes for each customer. Both indices measure
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 57

the deviation from mean value and are expressed in per-unit value as follows:

dki,t,o − dbki,o ρkt − ρkt,o

LDI(i, t) = ; PDI(i, t) = (4.4)
dbk i,o
ρkt,o

where, dki,t,o and ρkt are the demand and dynamic price of ith customer of class k at
time t, respectively. dbki,o and ρkt,o are the average demand of ith class of class k and FR,
respectively. These indices are plugged into a satisfaction coefficient using an exponential
function. It is to keep the value positive, as the deviation index can be either positive or
negative value. This defines the customer’s modified satisfaction coefficient under DR as
follows:

k = w k exp(−γ k LDI(i, t)) × exp(−χk PDI(i, t))

ωi,t i,t i i
(4.5)
k

d −dbk ρk −ρk
k
= wi,t exp −γik i,t,o i,o k t
− χi ρk t,o
bk di,o t,o

where, γik and χki are the coefficients associated with LDI and PDI of class k, respectively.
Equation (4.5) indicates that the satisfaction coefficient will be lowered during peak hours
due to a negative value of LDI and PDI, and increases during off-peak hours. This gives
the customer’s pay-off function under PBDR as follows:

2!
X (dki,t ) X
S(dki,t , ωi,t
k k
, βi,t )= k k
ωi,t k
di,t − βi,t − ρkt dki,t (4.6)
2
t∈ΩT t∈ΩT

As the customers participating in DR optimistically can expect to keep the overall energy
the same. Thus, this rationality is achieved by imposing an energy consumption constraint
onto the pay-off function. Further, it is converted into an unconstrained optimization using
the Lagrange formulation. This expresses the PBDR problem as follows:
 
2!
X (dki,t ) X X X
L(dki,t , ωi,t
k k
, βi,t )= k k
ωi,t di,t − βi,t − ρkt dki,t + λki  dki,t − dki,t,o 
2
t∈ΩT t∈ΩT t∈ΩT t∈ΩT

(4.7)

The third term of (4.7) represents the energy consumption constraint. dki,t,o and dki,t are
the demand of customer i of class k at time t BDR and ADR. The optimal condition of
the Lagrange function is defined by the first-order condition (FOC).

∂L(dki,t , ωi,t
k , βk )
i,t k
= ωi,t − βi,t dki,t − ρkt + λki (4.8)
∂dki,t
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 58

This gives the demand consumption at hour t as follows:

k − ρk + λk
ωi,t t i
dki,t = k
(4.9)
βi,t

Putting (4.9) into the energy consumption constraint, gives

k − ρk + λk
X ωi,t t i
X
k
− dki,t,o = 0 (4.10)
t∈ΩT
βi,t t∈ΩT

Solving (4.10) for λki , we get

P ω k .  P ρk. 
dki,t,o −
P
i,t β k + t βk
t∈ΩT t∈ΩT i,t t∈ΩT i,t
λki = P 1.  (4.11)
k
βi,t
t∈ΩT

Substituting (4.11) into (4.9) gives

( !, )
k − ρk +
P ω k .  P ρk.  P 1. 
dki,t,o −
P
ωi,t t i,t β k + t βk k
βi,t
t∈ΩT t∈ΩT i,t i,t t∈ΩT t∈ΩT
di,t = k
βi,t
(4.12)

Equation (4.12) denotes the ith customer’s demand ADR at hour t.

4.2.2 Exponential CUF Modeling

The exponential form of the CUF utilized in the social welfare function is defined as
follows [122–124]:

k k
S(dki,t , wi,t
k k
, βi,t k
) = wi,t × (1 − e−βi,t ×di,t ) − ρkt dki,t (4.13)

k and β k are defined as the satisfaction coefficient and risk-aversion coefficient,

where, wi,t i,t
respectively. The satisfaction coefficient value is obtained using the FOC. This gives the
satisfaction coefficient as follows:

ρkt βi,t
k k
k
wi,t ≥ k
e di,t (4.14)
βi,t
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 59

The satisfaction coefficient indicates dependence upon both parameters, and it varies ex-
ponentially. Under PBDR, the satisfaction coefficient is modified as follows:

k = w k exp(−γ k LDI(i, t)) × exp(−χk PDI(i, t))

ωi,t i,t i i
(4.15)
k

d −dbk ρk −ρk
k
= wi,t exp −γik i,t,o i,o k t
− χi ρk t,o
bk di,o t,o

Using the modified satisfaction coefficient, the Lagrange function of the exponential CUF
considering the energy consumption constraint is formulated as follows:
 
X k ×dk
−βi,t
X X X
L(dki,t , ωi,t
k k
, βi,t )= k
ωi,t × (1 − e i,t )− ρkt dki,t + λki  dki,t − dki,t,o 
t∈ΩT t∈ΩT t∈ΩT t∈ΩT

(4.16)

Now, obtaining the optimality condition using FOC gives the optimal demand at each
state t as follows:
!
k βk
ωi,t
1 i,t
dki,t = k ln (4.17)
βi,t ρkt − λki

The Lagrange multiplier value is calculated by substituting (4.17) into energy consumption
constraint.
!
X 1 k βk
ωi,t i,t
X
k
ln − dki,t,o = 0 (4.18)
t∈Ω
βi,t ρkt − λki t∈ΩT
T

On simplifying constraint (4.18) gives λki as follows:

P 1.  k β k ρk ) −
P k
k
βi,t ln(ωi,t i,t t di,t,o
k t∈ΩT t∈ΩT
λi = P 1.  (4.19)
k k
βi,t ρt
t∈ΩT

Substituting (4.19) into (4.17) gives the optimal demand of ith customer, class k at time
t as follows:
 P 1.  P 1.  P k 
ρkt k k
ρt βi,t − k k
ρt βi,t ln(ω k β k ρk ) +
i,t i,t t di,t,o
1 k k 1  t∈ΩT t∈ΩT t∈ΩT 
di,t = k
ln(ωi,t βi,t ) − k ln  P 1.  
βi,t βi,t  k k

ρt βi,t
t∈ΩT

(4.20)
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 60

4.2.3 Logarithmic CUF Modeling

The logarithmic form of the CUF for individual customers is defined as follows [30,125,126]:

S(dki,t , wi,t
k k
, βi,t k
) = wi,t k
ln(βi,t + dki,t ) − ρkt dki,t (4.21)

k , and β k are the satisfaction value and risk-aversion of ith customer, class k at
where, wi,t i,t
k is set equal to 1 to avoid the utility value
time t. It is to be noted that the parameter βi,t
from going to infinity. Its satisfaction value BDR is estimated using a FOC of (4.21) as
expressed follows:

k
wi,t ≥ ρkt (βi,t
k
+ dki,t ) (4.22)

Under DR condition, the satisfaction value is updated similarly to (4.5). It is defined as

follows:

k = w k exp(−γ k LDI(i, t)) × exp(−χk PDI(i, t))

ωi,t i,t i i
(4.23)
k

d − bk
d ρ k −ρk
k
= wi,t exp −γik i,t,o i,o k t
− χi ρ k t,o
bk di,o t,o

Now, the Lagrange formulation of logarithmic CUF with the energy consumption con-
straint is expressed as follows:
 
X X X X
L(dki,t , ωi,t
k k
, βi,t )= k
ωi,t k
ln(βi,t + dki,t ) − ρkt dki,t + λki  dki,t − dki,t,o  (4.24)
t∈ΩT t∈ΩT t∈ΩT t∈ΩT

On solving the optimality condition using FOC; gives the ith customer’s demand of class
k at hour t as follows:
k
ωi,t
dki,t = k
− βi,t (4.25)
ρkt − λki

To calculate the value of λki , (4.25) is substituted in the energy consumption constraint.
!
X k
ωi,t X
k
− βi,t − dki,t,o = 0 (4.26)
t∈ΩT
ρkt − λki t∈ΩT
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 61

On solving (4.26), using some simplification gives λki

P ω k . 
dki,t,o + k −
P P
βi,t i,t ρk
t
t∈ΩT t∈ΩT t∈ΩT
λki = P ω k .  (4.27)
i,t (ρk )2
t
t∈ΩT

Substituting (4.27) into (4.25) gives the customer’s demand at time t as follows:

k
P ω k . 
ωi,t i,t (ρk )2
t
T t∈Ω k
di,t = P ω k .  P ω k .  − βi,t (4.28)
ρkt 2
P k P k
i,t (ρk ) − di,t,o + βi,t − i,t ρk
t t
t∈ΩT t∈ΩT t∈ΩT t∈ΩT T

4.2.4 Power CUF modeling

For the power CUF, the customer’s social welfare function is defined as follows:
 k

1−βi,t
(dki,t ) −1
S(dki,t , wi,t
k k
, βi,t k 
) = wi,t k
 − ρkt dki,t (4.29)
1 − βi,t

where, dki,t , wi,t

k , and β k are demand, the satisfaction value and risk-aversion of ith cus-
i,t
tomer, class k at time t. Now, the satisfaction value of the power utility function is
obtained using the FOC of (4.29). It is expressed as follows:

k
k
wi,t ≥ ρkt (dki,t )βi,t (4.30)

Further, the modified satisfaction value under PBDR is evaluated according to the relation
defined in (4.5). This gives the modified satisfaction coefficient as follows:

k = w k exp(−γ k LDI(i, t)) × exp(−χk PDI(i, t))

ωi,t i,t i i
(4.31)

k k dki,t,o −dbki,o k ρkt −ρkt,o
= wi,t exp −γi bk − χ i ρk
di,o t,o

The Lagrange formulation of power CUF under PBDR is defined as follows:

 k
  
1−βi,t
X (dki,t ) −1 X X X
L(dki,t , ωi,t
k k
, βi,t )= k 
ωi,t k
− ρkt dki,t + λki  dki,t − dki,t,o 
t∈ΩT
1 − βi,t t∈ΩT t∈ΩT t∈ΩT

(4.32)
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 62

The FOC gives the customer’s demand at time t as follows:

k )
!(1/βi,t
k
ωi,t
dki,t = (4.33)
ρkt − λki

Further, calculating the value of λki as follows:

k )
P ω k . (1/βi,t
dki,t,o −
P
i,t ρk
t
t∈ΩT t∈ΩT
λki = k
(4.34)
(1/βi,t )
 
P ωk
 . 
1
i,t ρk
t ρkt βi,t
k
t∈ΩT

Substituting (4.34) into (4.33), gives the customer’s demand at time t as follows:

k
(1/βi,t

P ω k . (1/βi,t k )  )
k 1
ωi,t i,t ρk
ρkt βi,t
k
 t 
t∈ΩT
dki,t = 
 
k ) k )
P ω k . (1/βi,t  k . (1/βi,t
 

ρkt i,t ρk
1
−
P
dki,t,o +
P ωi,t k 
t ρk β k t i,t
ρ t
t∈ΩT t∈ΩT t∈ΩT
(4.35)

4.2.5 Modeling of Customer’s Risk Behaviour

Risk is a human psychological behaviour measure of a certain or uncertain outcome [27,

127]. In utility theory, the shape of the utility function is an elementary approach to
describe the degree of risk of the customers as illustrated schematically in Fig. 4.3.

Figure 4.3: Customer’s behaviour under risk

Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 63

The figure indicates that the degree of risk varies from risk-averse to risk-seeker as the
CUF’s curve changes from concave to convex. The centerline between these two represents
the customer’s risk-neutral behaviour. However, the value of the risk-aversion coefficient
and the curvature of CUF are insufficient to determine the risk associated with the cus-
tomers discretely [127]. It is better assessed using the risk-based indices suggested by
Arrow-Pratt. These indices are described as absolute risk-aversion (ARA) and relative
risk-aversion (RRA). ARA estimates the change in the individual absolute level of initial
consumption, whereas RRA measures the percentage change in the individual absolute
level of initial consumption. ARA is defined as the ratio of the second derivative to the
first derivative of the utility function, and a negative sign is used to make it a positive
value. It is also called as local risk-aversion due to the discrete values of ARA for the
different levels of d [127]. It is expressed as follows:
′′
U (dk )
ARA(dki ) = − ′ ki (4.36)
U (di )

On the other hand, RRA is defined as the risk-aversion proportional to the demand con-
sumed. It is also called global risk-aversion and is expressed as follows:
′′
dki U (dki )
RRA(dki ) = − = dki ARA(dki ) (4.37)
U ′ (dki )

Based on the formulated indices, the risk-indices expressions for the different CUFs are
summarized in Table 4.1.
Table 4.1: ARA and RRA indices for the different CUFs

Cardinal utility functions

Index
Quadratic Exponential Logarithmic Power
βik 1 βik
ARA ωi −βik dki
k βik βik +dki dki

dki βik dki

RRA ωik −βik dki
βik dki βik +dki
βik

The described expressions of indices reveal the discrete response under CUFs. It exhibits
the increasing value of ARA for the rising demand for quadratic CUF. In Exponential
CUF, ARA is constant and is equal to the risk-aversion coefficient. For logarithmic and
power CUFs, ARA values decrease with increase in demand. In terms of RRA, quadratic
and exponential CUFs exhibit increasing RRA, whereas decreasing value for logarithmic
CUF with increased demand. The RRA value in the case of power CUF is directly related
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 64

to the risk-aversion coefficient and is independent of demand.

A probabilistic model has been adopted for each customer to illustrate their varying degree
of risk-averseness. The risk-aversion coefficient is drawn randomly from a truncated normal
distribution and is defined as follows [105]:

βik ∼ N (µkβ , (σβk )2 ) (4.38)

k
β ki ≤ βik ≤ βi (4.39)

where, βik is a generated risk-aversion coefficient with its mean µkβ and variance (σβk )2 . In
addition, the risk-aversion coefficient is bounded within limits.

4.3 Simulation Results and Discussions

The simulation results are performed on the standard 33-bus test distribution load data
given in Table A.1 of Appendix A. The proposed CUFs based PBDR models are developed
®
in MATLAB . It is assessed on the class-wise customers described in section 3.6 of chap-
ter 3. The class-wise information such as customer nodes, total allocated demand, number
of customers, and S/CF are given in Table B.1 in Appendix B. Further, load modeling
is performed as described in section 3.3 of chapter 3. The suggested PBDR models are
investigated on the RTP price as shown in Fig. 4.4. RTP price is referred from Ontario
Energy Board [109]. The time spans of the considered price are segmented into three peri-
ods: off-peak (0:00-7:00, 23:00-24:00), valley: (11:00-17:00) and peak period: (8:00-11:00,
18:00-22:00) for the analysis purpose.

40
Price (¢/kWh)

30

20

10

0
1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour)

Figure 4.4: Dynamic price profile under RTP

The CUF’s satisfaction parameter BDR and ADR for each customer is evaluated using
the derived approach based on their demand level and price offered. The risk-aversion
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 65

coefficient parameter is generated randomly using a mean value of 0.4 and a variance of
0.6. The class-wise (R, LI, MI, C and A) values of coefficients γ k and χk are set in the
range of (0.1–1.4) appropriately.
The risk-based indices, ARA and RRA against the varying value of β at the same de-
mand level are shown in Fig. 4.5 (a)-(b). These indices are denoted by ARA1/2/3/4
and RRA1/2/3/4 for quadratic, exponential, logarithmic, and power CUF, respectively.
Fig. 4.5 (a) indicates that ARA increases for quadratic, exponential, and power CUF,
whereas it remains constant for the logarithmic CUF. A similar response is also exhibited
under RRA as shown in Fig. 4.5 (b). The figures indicate that both ARA and RRA indices
follow risk-aversion linearly with few exceptions. But it does not entirely corroborate the
general assumption that risk-averseness grows with an increase in the risk-aversion coeffi-
cient.

ARA1 ARA2 ARA3 ARA4 RRA1 RRA2 RRA3 RRA4

0.15 2.5 2.5 25
2.0 20
2.0
ARA1/3/4

0.10
RRA1/3/4
ARA2

1.5

RRA2
1.5 15
1.0
0.05 1.0 10
0.5
0.5 5
0.00 0.0
0.43 0.94 1.00 1.38 1.78 2.26 0.0 0
0.43 0.94 1.00 1.38 1.78 2.26
β
β
(a) (b)

Figure 4.5: Variation in (a) ARA; (b) RRA with the change in β for the different CUFs

ARA and RRA values for R class customers organized in increasing demand are illus-
trated in Fig. 4.6 (a)-(b). Both figures indicate a distinct and different values, showing the
customers’ discrete perception in DR. The customers with higher demand exhibit lower
values of ARA for logarithmic and power CUFs. On the other hand, the quadratic and
exponential CUFs exhibit increasing and randomized values on the higher demands. The
risk comparison using RRA shows that the customers with higher demand possess high
RRA values for quadratic, logarithmic, and exponential CUFs. It means that the cus-
tomers with higher demands respond less in DR. On the contrary, the power CUF exhibit
randomized value of RRA indicates that the customers’ risk-behaviour in DR remain un-
affected by their demand levels.
Now, the impact of the degree of risk behaviour via ARA on DR participation based on
peak load curtailment is illustrated in Fig. 4.7 (a)-(d). The figure indicates that quadratic
(QUAD), exponential (EXP), and power (POW) CUFs show diminishing curtailment as
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 66

ARA1 ARA2 ARA3 ARA4 RRA1 RRA2 RRA3 RRA4

0.8 3 3
30
0.6

RRA1/3/4
ARA1/3/4

2 2

RRA2
ARA2
20
0.4
1 1 10
0.2

0.0 0 0 0
1.97 3.97 5.93 7.94 8.99 11.86 14.90 1.97 3.97 5.93 7.94 8.99 11.86 14.90
Demand (d) Demand (d)
(a) (b)

Figure 4.6: Variation in (a) ARA; (b) RRA with increasing demand for R customer
class

their ARAs patterns follow upward shift in a randomized way for R class customers. It
suggests that the customers tend to exhibit more risk-averse behaviour as ARA value
increases. This yields less DR participation from the customers. On the contrary, loga-
rithmic (LOG) CUF displays opposite behaviour, which gives the same patterns for ARA
and curtailment. These observations indicates that the customers with higher ARA values
usually respond less in DR or another way around except the logarithmic CUF.

15 0.20 3
QUAD ARA1 EXP ARA2
Curtialment (%)

Curtialment (%)

15
0.15
2
ARA1

ARA2

10 0.10 10
1
0.05
5
5 0.00 0
1 21 41 61 81 101 121 1 21 41 61 81 101 121
No. of Customers No. of Customers
(a) (b)
11 0.4 0.8
LOG ARA3 POW ARA4
12
Curtialment (%)
Curtialment (%)

10 0.3 0.6
10
ARA4
ARA3

0.2 0.4
9 8
0.1 6 0.2
8
0.0 4 0.0
1 21 41 61 81 101 121 1 21 41 61 81 101 121
No. of Customers No. of Customers
(c) (d)

Figure 4.7: R class customer participation in DR under (a) Quadratic, (b) Exponential,
(c) Logarithmic, (d) Power CUF
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 67

Fig. 4.8 (a)-(e) present the aggregated class-wise load demand patterns, BDR and ADR
under the different CUFs based DR models. The sub-figures indicate that each customer
class effectively exhibit load adjustment with the variations in the CUF’s parameters. It
curtailed the load during peak periods, and simultaneously adjusted it to the valley and
off-peak periods.

BDR QUAD EXP LOG POW BDR QUAD EXP LOG POW
1000 1400
Load Demand (kW)

Load Demand (kW)

1300
800
1200
600 1100

400 1000
900
200
800
1 3 5 7 9 11 13 15 17 19 21 23 1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour) Time (Hour)
(a) (b)

BDR QUAD EXP LOG POW BDR QUAD EXP LOG POW
600
200
Load Demand (kW)

Load Demand (kW)

150 400

100
200
50

0 0
1 3 5 7 9 11 13 15 17 19 21 23 1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour) Time (Hour)
(c) (d)

BDR QUAD EXP LOG POW

800
Load Demand (kW)

600

400

200

0
1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour)
(e)

Figure 4.8: Aggregated class-wise load patterns BDR and ADR using the various CUFs
for: (a) R, (b) LI, (c) MI, (d) C, (e) A

The impact of DR on peak load curtailment is described using demand flexibility (DF ).
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 68

It is depicted in Fig. 4.9 (a). The figure shows that the CUFs give the different degrees of
DF for the considered customer classes. The induced DF shows the distinct application
of the CUF. In addition, DF is also affected by load patterns of the different customer
classes. Furthermore, DR impact is also measured on the system level in terms of load
factor shown in Fig. 4.9 (b). The figure shows that load factor is considerably reduced
during peak hours due to peak load curtailment and is increased on account of the adjusted
peak load to the valley/ off-peak hours.

QUAD EXP LOG POW BDR QUAD EXP LOG POW

Demand flexibility (DF %)

1.0
10 Load Factor (LF) 0.9

0.8
5 0.7

0.6

0 0.5
R LI MI C A 1 3 5 7 9 11 13 15 17 19 21 23
Classes Time (Hour)
(a) (b)

Figure 4.9: (a) Demand flexibility under the various customer classes using the different
CUFs; (b) Aggregated load factor

An economic comparison of the customer bills under the different CUFs is summarized in
Table 4.2. BASE case represents the BDR state at the flat rate and NDR state at RTP. It
is worth noting that NDR is condition of zero DR participation after its implementation.
The results show that the customer classes’ bills are increased with the transition of BDR
to NDR states. The variation in the class-wise bills are discrete and different. It is lower
for LI and A class customers, whereas higher for R, MI and C class customers. However,
the customers’ bills are appeared to reducing after their participation under ADR relative
to NDR state. Moreover the reduction in the class-wise and total bills are almost homo-
geneous. The analysis shows that R, MI and C are the highly benefited customer classes
in terms of bill increase/decrease under NDR and ADR states compared to LI and A class
customers.
Table 4.3 presents the profit obtained under the different CUFs based PBDR models. The
results demonstrate that the profit gained from the class-wise customers show similar vari-
ation under all PBDR models. It increases for R, MI & C class customers and decreases
for LI and A class customers.
A sensitivity analysis of the considered CUFs is performed by varying the satisfaction
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 69

Table 4.2: Aggregated customer class bills under the different CUFs based PBDR mod-
els

Methods Class R (¢) LI (¢) MI (¢) C (¢) A (¢) Total (¢)

BDR 244358.56 552397.07 69182.46 286844.48 88075.63 1240858.20
BASE NDR 260578.32 555234.20 74534.57 319037.28 88864.43 1298248.79
(%) 6.64 0.51 7.74 11.22 0.90 4.62
ADR 254358.31 541886.18 72716.74 312890.77 87844.02 1269696.03
QUAD
(%) -2.39 -2.40 -2.44 -1.93 -1.15 -2.20
ADR 255643.25 545871.17 72647.65 311846.54 85977.04 1271985.66
EXP
(%) -1.89 -1.69 -2.53 -2.25 -3.25 -2.02
ADR 252954.69 543961.34 73053.85 312581.26 87306.83 1269857.97
LOG
(%) -2.92 -2.03 -1.99 -2.02 -1.75 -2.19
ADR 256888.94 539660.35 73069.17 311314.78 86825.60 1267758.85
POW
(%) -1.41 -2.80 -1.97 -2.42 -2.29 -2.35

Table 4.3: Utility profit under the different CUFs based PBDR models

Class R (¢) LI (¢) MI C (¢) A (¢) Total (¢)

Methods
BDR 70858.74 183321.23 19625.15 73954.08 28965.13 376724.32
ADR 79273.29 168619.87 22739.02 97403.83 27316.42 395352.42
QUAD
(%) 11.88 -8.02 15.87 31.71 -5.69 4.94
ADR 79645.04 169875.96 22717.52 97113.37 26734.58 396086.47
EXP
(%) 12.40 -7.33 15.76 31.32 -7.70 5.14
ADR 78831.54 169262.72 22839.24 97261.81 27142.78 395338.10
LOG
(%) 11.25 -7.67 16.38 31.52 -6.29 4.94
ADR 80043.48 167878.74 22842.31 96949.17 26991.53 394705.23
POW
(%) 12.96 -8.42 16.39 31.09 -6.81 4.77

coefficient and price. The satisfaction coefficient is varied by proportionally changing its
PDI & LDI factors and price through price multiplying factors (PMF). The effect of satis-
faction coefficient and price variation on induced average DF is shown in Fig. 4.10 (a)-(b).
Fig. 4.10 (a) depicts the variation in the average DF induced at the system level against
the scaling factors of LDI & PDI coefficients. The figure shows that all CUFs exhibit in-
creasing DF value, as the scaling factor increases. In addition, the variations in all CUFs
are almost alike. Though, the exponential CUF shows higher variation as the scaling fac-
tor increases. On the other hand, quadratic, logarithmic and power CUFs show similar
and small variation.
Similarly, Fig. 4.10 (b) illustrates the variation in induced DF with the increasing dy-
namic price. The figure indicates that all the CUFs exhibit almost same response on base
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 70

QUAD EXP LOG POW QUAD EXP LOG POW

Average flexibility (DF %)

Average flexibility (DF %)

20
30
15
20
10

5 10

0 0
1.0 1.2 1.4 1.6 1.8 1.0 1.2 1.4 1.6 1.8
Scaling factor of LDI/PDI coefficients PMF
(a) (b)

Figure 4.10: Average demand flexibility with variation in : (a) Satisfaction coefficient;
(b) price

dynamic price. However, average DF variations become noteworthy for quadratic and ex-
ponential CUF as PMF increases. On the other hand, logarithmic and power CUF exhibit
gradual variation in DF as PMF increases.

4.4 Proposed Overlapping Generation based PBDR Model

using OUF

The overlapping generation (OLG) explains how a person spends its net wealth for con-
sumption in the present and for saving in the future states with the change in the interest
rate [65, 128]. It is a discrete time-variant state model to study human life cycle behavior,
named after American noble laureate Peter A. Diamond. It employs an OUF to measure
the customer’s utility in each period. The OUF measures the customers utility through
the order or ranking and has precise preference. In OLG, Romer’s CRRA is employed to
evaluate the customer’s utility. The general illustration of OLG model can be described
through the customer’s economic behaviour model for two states. Let’s suppose that the
customer’s lifetime is divided into the present t and future state t + 1. It consumes C1t
in the present state and saves C2t+1 amount for the consumption in the future state. The
amount C2t+1 earns interest rate r on the saved amount C2t+1 = (1 + r) × (T − C1t ). Then,
the customer’s lifetime utility in the present state t can be expressed as follows [65]

1−θ
C1t C 1−θ
U (Ct ) = + η 2t+1 (4.40)
1−θ 1−θ
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 71

where, U (Ct ) and θ are the customer’s total utility and a degree of risk-aversion, respec-
tively. The utility first term represents the present utility and the second term is the
future utility measured at the present state through the discount factor η. Parameter θ
measures the customer’s willingness or preference to achieve the desired output. It allows
the redistribution of the consumption level. The customer’s total utility is constrained by
a fixed budget. It is defined as follows:

C2t+1
C1t + =T (4.41)
1+r

where, T is the total budget constraint and r is relative interest rate. Equation (4.41) in-
dicates that the consumer consumption level varies with the change in the relative interest
rate for a fixed budget constraint.
Similarly, PBDR has also distinctive time states with the different demand consumption
and alters its demand with the change in relative price of the different time states. This
establishes a fair analogy between the OLG and PBDR model. It models consumption
equivalent to demand and interest rate to the change in the relative price of the different
time states, respectively. In addition, parameter θ defines the customer’s willingness or
degree of risk in DR. Hence, OLG framework is adapted for the designing of a three-tier
TOU price in PBDR. The time segments are defined as peak, valley, and off-peak, broadly
based on price level duration of TOU program and are assigned the period t, t+1, and t+2,
respectively. Since, the energy consumption in these three periods is distinct. This will
also give a distinct utility for demand consumption in each segment. Hence, the customer’s
utility at time t with overlapping in valley and off-peak period will be the summation of
the obtained utility in each segmented period. It is defined as follows [127]:

1−θk 1−θk 1−θk

di,pt i i
di,vt+1 i
di,opt+2
U (dki,t ) = u(dki,pt ) + u(dki,vt+1 ) + u(dki,opt+2 ) = +ηi,pv + ηi,pv ηi,vop
1 − θik 1 − θik 1 − θik
(4.42)

where, U (dki,t ) and u(dki,p/v/op ) denote the overall utility, and distinct utility in each period,
respectively. ηi,pv and ηi,vop are the proportion demand factors between peak to valley and
valley to off-peak period consumption, respectively. It is to be noted that the formulated
problem in (4.42) represents the utility function of the customer i of class k obtained using
the dis-aggregated load modeling approach [1].
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 72

4.4.1 Load Recovery in DR using Altruistic Behaviour

When PBDR is initiated with the price changes, the customer’s response to the price
variation can be exhibited in the following two rational ways:

ˆ The customer shifts its demand from peak hours to a valley or off-peak hours.

ˆ The customer shifts its demand from valley hours to off-peak hours.

The aforesaid two statements indicate that each period is linked with its subsequent periods
in PBDR. This DR behaviour is termed as the load recovery [129] or fading effect [48]. It
allows the customer to shift or recover its curtailed demand in the nearest hours. A similar
analogy is also exhibited in the OLG model, where the generations are inter-linked due
to the altruistic behavior. The altruism is a human behaviour which makes the consumer
care about its decedents, i.e., each generation cares about its next generations as long
as it lives. This behavior is called the forward altruistic motives [130, 131]. Hence, the
generations/states and altruistic behaviour of the OLG model can be made equivalent
to the segmented periods and load recovery in DR, respectively. It is mathematically
expressed as a weighted sum of the current and the future generations and is defined as
follows [131]:

∞  
X υ
Γt (dki ) = λ̄ki U (dki,t+υ ) (4.43)
υ=0

Equation (4.43) represents one-side altruism, where the generation cares about its own and
its later generations, not the preceding generations. (λ̄ki )υ is a non-negative coefficient. It is
referred as a degree of inter-generational altruism and has a value of less than 1. Moreover,
it decreases with the subsequent generations due to the finite lifetime of each generation.
It indicates that altruism diminishes over distance or time. Likewise, load recovery in DR
is diminished with time elongation, as the period goes far away from the reference period,
i.e., the curtailed demand time state [129]. The expanded form of the total utility function
in terms of the individual utility function is expressed as follows:

Γt (dki ) = [u(dki,pt ) + ηi,pv u(dki,vt+1 ) + ηi,pv ηi,vop u(dki,opt+2 )] + λ̄[u(dki,pt+1 ) + ηi,pv u(dki,vt+2 )
+ηi,pv ηi,vop u(dki,opt+3 )] + λ̄2 [u(dki,pt+2 ) + ηi,pv u(dki,vt+3 ) + ηi,pv ηi,vop u(dki,opt+4 )]
(4.44)

It is to be noted that i and k is dropped from the coefficient λ̄ν for clarity and represent
each customer. Since the customer’s utility is segmented into three periods t, t + 1, and
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 73

t + 2 only. Thus, the customer’s utility will be zero for the subsequent periods t + 3 and
t + 4. This gives the reduced form of the total utility function as follows:

Γt (dki ) = [u(dki,pt ) + ηi,pv u(dki,vt+1 ) + ηi,pv ηi,vop u(dki,opt+2 )] + λ̄[u(dki,pt+1 ) + ηi,pv u(dki,vt+2 )]
+ λ̄2 [u(dki,pt+2 )] (4.45)

As, t, t + 1, and t + 2 are pertaining to peak, valley, and off-peak periods respectively, so
the above expression can be written in the following way:

Γt (dki ) = [u(dki,pp ) + ηi,pv u(dki,vv ) + ηi,pv ηi,vop u(dki,opop )] + λ̄[u(dki,pv ) + ηi,pv u(dki,vop )] + λ̄2 [u(dki,pop )]
(4.46)

where, the subscript pp indicates the customer’s transferred utility from the peak-to-
peak period and is abbreviated p. Similarly, the subscript pv indicates the transferred
utility from peak to valley period and is denoted using a subscript v in (4.46). Then,
the customer’s utility is optimized, subjected to the budget constraint. It bounds the
customer’s consumption and is defined as follows:

Tki = dki,p ρkp + dki,v ρkv + dki,op ρkop (4.47)

Expressing (4.47) analogous to (4.41) using the some manipulation as follows:

Tki 1 1
= dki,p + dki,v + dk (4.48)
k
ρp k
(1 + r1 ) (1 + r1 )(1 + r2k ) i,op
k

Tki
where, ρkp
is equivalent to T. r1k and r2k are the change in relative price of peak to valley
and valley to off-peak price, respectively. These ratios are expressed as follows:

ρkp − ρkv ρkv − ρkop

r1k = ; r2k = (4.49)
ρkv ρkop

This formulates the proposed problem into a constrained optimization. It is solved using
the Lagrange formulation [96] and is expressed as follows:

1−θik 1−θik 1−θ k 1−θik 1−θik

" # " #
di,p ηi,pv di,v ηi,pv ηi,vop di,op i di,p ηi,pv di,v
L(dki,p , dki,v , dki,op ) = 1−θik
+ 1−θik
+ 1−θik
+ λ̄ 1−θik
+ 1−θik
"
1−θi k
#   (4.50)
di,p Tki dki,v dki,op
+λ̄2 1−θik
+ λki ρkp
− dki,p − (1+r1k )
− (1+r1k )(1+r2k )
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 74

The optimal condition of (4.50) is obtained by taking the first derivative as follows:

∂L −θik
dki,p : = d 2 k
i,p (1 + λ̄ + λ̄ ) − λi = 0 (4.51)
∂dki,p

∂L −θk 1
dki,v : k
= di,v i ηi,pv (1 + λ̄) − λki =0 (4.52)
∂di,v (1 + r1k )

∂L −θk 1
dki,op : k
= ηi,pv ηi,vop di,opi − λki =0 (4.53)
∂di,op (1 + r1 )(1 + r2k )
k

On solving (4.51)-(4.53), the energy consumption during the peak, valley, and off-peak
period is obtained as follows:

Tki
  
1
dki,p = (4.54)
ρkp Den

1/ k 1/ k 1/θk
Tki (1 + λ̄) i (1 + r1 ) θi ηi,pvi
θ k
dki,v = (4.55)
ρkp Den

1/ k 1/ k
Tki ((1 + r1k )(1 + r2k )) θi (ηi,pv ηi,vop ) θi
dki,op = k (4.56)
ρp Den

where, Den is defined as follows:

1 − θik/ k
"
1/ k
Den = (ηi,pv ηi,vop ) θi ((1 + r1k )(1 + r2k )) θi

‘1 − θik/ k
! #
1/ k 1/ k 1/ k
+(1 + λ̄) θi η i,pv
θi (1 + rk )
2
θi + (1 + λ̄ + λ̄2 ) θi (4.57)

The derived equations (4.54)–(4.56) give the customer’s chunk of the energy consumption
at the segmented time states. Now, the demand consumption in each period is obtained
using the direct proportionate relations under the different periods BDR and ADR as
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 75

stated below:
 . 
dki,p,tp = dki,p,tp ,o × dki,p dki,p,o (4.58)
 . 
dki,v,tv = dki,v,tv ,o × dki,v dki,v,o (4.59)
 . 
dki,op,top = dki,op,top ,o × dki,op dki,op,o (4.60)

Equations (4.58)-(4.60) define the demand ADR at time t in the respective time states.
n o
dki,t,o = dki,p,tp ,o , dki,v,tv ,o , dki,op,top ,o (4.61)
n o
dki,t = dki,p,tp , dki,v,tv , dki,op,top (4.62)

Equations (4.61)-(4.62) represents the set of demand of i customer of class k constituted

from for all time states BDR and ADR.

4.4.2 Modeling of Substitution Effect between Products

In microeconomics, the substitution effect is defined as a trade-off between two products,

where one is willing to give up a few units of products by consuming more units of the other
products due to the relative change in the price. It is measured through the marginal rate
of substitution (MRS) [23] and is expressed as the rate at which one product substitutes
the other product while keeping all other products constant. In the CRRA function, θ is
a coefficient to control the substitution effect. The MRS between peak and valley periods
consumption is defined as follows [24].
 
∂u −θk !θik
ddki,p ∂dki,v di,v i dki,p
MRSki,pv = − = −  = −ηi,pv = −ηi,pv (4.63)
ddki,v ∂u
−θk
di,p i dki,v
∂dki,p

Similarly, MRS between peak to off-peak, and between valley to off-peak periods can be
evaluated. It is further extended to calculate the elasticity of substitution, which measures
the change in the input ratio to the change in MRS. On evaluating elasticity of substitution
between peak to valley periods, gives the following expression:
h  . i  . 
d ln dki,p dki,v %∆ dki,p dki,v 1
k
σi,pv =  k
 = k

= k (4.64)
d ln(MRSi,pv ) %∆ MRSi,pv θi
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 76

The above relation indicates the rate of change consumption between peak and valley
periods to the change of MRS is a constant value and it depends t upon θ. Similarly,
relations for σpop and σvop can be confirmed. This indicates that θik is a control parameter
to redistribute the curtailed peak demand across a valley or off-peak periods. If θik < 1,
then the substitution effect dominates, i.e., peak load will be substituted in the cross-
periods and vice-versa. Since the elasticity of substitution has an inverse relationship with
the customer’s willingness factor in the OLG. Thus, the characterization of willingness will
quantify the response in DR.

4.5 Proposed Fuzzy Framework for Customer Participation

in DR

In DR, each customer can show a varying response from least to maximum level based
on their willingness and the available flexible demand at the same demand capacity level.
Thus, an illustration of such a vague response can be better illustrated through a fuzzy
inference system (FIS), which is better suited for representing human reasoning and per-
ception in an inextricable problem [132]. It maps input to output using a pre-defined
set of rules based on the expert knowledge. The flexible demand (FD) and customer de-
mand level (CDL) are assumed as the input variables and willingness parameter as output
variable. FD is assumed to be normally distributed within the limits

F Dik ∼ N (µkF D , (σFk D )2 ) (4.65)

k
F Dki ≤ F Dik ≤ F Di (4.66)

In the Fuzzy system, inputs are modeled as linguistic variables through each MFs. The cus-
tomer’s demand level is classed between the three linguistic variables L: Low, M: Medium,
H: High, and the flexible demand in seven linguistic variables VL: Very Low, L: Low, M:
Medium, MH: Medium-High, H: High, VH: Very High. The output variable is indexed in
eight linguistic variables EL: Extremely Low, VL: Very Low, L: Low, M: Medium, MH:
Medium-High, H: High, VH: Very High, EH: Extremely High. Once the input and output
variables are set in the ranges of linguistic variables/MFs, then the relationship between
input and output is mapped based on the intuitive behavior of the desired result being
investigated in the problem. The fuzzy mappings between input and output variables are
tabulated in Table 4.4.
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 77

Table 4.4: FIS Evaluation based on the customer demand and flexible demand

FD/CDL VL L M MH H VH
L EH VH H MH M L
M VH H MH M L VL
H H MH M L VL EL

The relationship shown in Table 4.4 indicates the output variable variation with the vary-
ing demand flexibility at the different demand levels. It is inversely related to the elasticity
of substitution. As the elasticity of substitution is a measure of the customers’ sensitiv-
ity. Hence, a relationship between load adjustment and the varying value of a control
parameter is devised as follows:

ζk
 
P C(θik ) k
= π exp (4.67)
θik

Equation (4.67) shows that load adjustment diminishes with the increasing value of θ.

4.6 Procedure of PBDR implementation using OLG model

In the proposed OLG based DR model, the customer’s energy consumption is assumed to
be constrained by its budget. It is set equivalent to the customer bill over a day at the flat
rate BDR for the analysis purpose. In addition, the customer’s total energy consumption
BDR and ADR is assumed to be the same. Though, this constraint may or may not satisfy
for the presumed budget constraint under the dynamic price. This may need to increase
the customer’s budget to keep the same energy level. Therefore, the customer’s budget
line has to be adjusted to maintain the same energy consumption level ADR. The process
of implementation is illustrated through an algorithm 4.1.

Algorithm 4.1 Procedure for the implementation of DR using OLG model

Input: ρp/v/o , κk , θik , dki,t,o , inital, ηi,pv , ηi,vo , ϵ
Output: dki , dki,p,tp , dki,v,tv , dki,op,top , Tki , ηi,pv , ηi,vo
for i∈ I do
Load customer demand profile dki,t,o
while |sum(dki,t ) − sum(dki,t,o )| ≤ ϵ do
Compute load dki,p/v/op under TOU according to (4.54), (4.55), and (4.56) for given
budget Tki
if Shifting≤Curtailment then
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 78

Update ηi,pv = ηi,pv + 0.001

Update ηi,vo = ηi,vo + 0.001
else if Shifting≥Curtailment then
Update ηi,pv = ηi,pv − 0.001
Update ηi,vo = ηi,vo − 0.001
end if
if sum(dki ) ≤ sum(dki,o ) then
Tki = Tki + 0.001
else if sum(dki ) ≥ sum(dki,o ) then
Tki = Tki − 0.001
end if
end while
Update ηi,pv , ηi,vo , Tki
Calculate dki,p,tp , dki,v,tv , dki,op,top according to (4.58)-(4.60)
end for
return Output

4.7 Simulation Results and Discussions

The effectiveness of the proposed OLG (Pro OLG) based DR model is performed on 33-bus
test distribution load data as given in Table A.1 of Appendix A. The class-wise informa-
tion is given in Table B.1 in Appendix B. The proposed OLG model is compared with
the standard PEM, CES and existing two-state OLG models [27]. The two-tier (peak and
off-peak periods) TOU pricing is applied on the existing two-state OLG model as em-
ployed in the Ref. [27]. The three-tier TOU pricing during the winter months in Ontario
Energy Board is utilized as illustrated in Fig. 4.11 (a) [109]. The load demand pattern is
segmented into three periods: off-peak (0:00-7:00, 23:00-24:00), valley: (11:00-17:00) and
peak period: (8:00-11:00, 18:00-22:00) for the analysis purpose. Now, to demonstrate the
effect of DR response on the different class customers, first willingness of each customer is
evaluated through FIS. The DR participation level is assumed to vary around in a range of
(0-30) % for the sake of simplicity. The output factor, preference, or willingness parameter
is supposed to vary between (0.8-1.2). Using (4.67), the willingness values are obtained
for the distinct demand levels. A sample of willingness values for the demand capacity of
2 kW of the R customers is shown in Fig. 4.11 (b).
The figure shows that willingness values are distinct for each customer at the same de-
mand level. This manifests that each customer will react to DR differently, revealing the
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 79

1.05
20
Price (¢/kWh)

1.00

θ
15
0.95
10
0.90
1 3 5 7 9 11 13 15 17 19 21 23 1 2 3 4 5 6 7 8 9
Time (Hour) No. of Customers
(a) (b)

Figure 4.11: (a) Three-tier TOU price framework, (b) A sample of willingness or control
parameter variation for demand level 2 kW

heterogeneity in their flexible loads [133]. Once willingness value is evaluated for each cus-
tomer class, the corresponding load curtailment during peak hour is evaluated as shown
in Fig. 4.12 (a)-(e). The figures reveal that peak load adjustment/shifting decreases as a
willingness parameter increases. It is due to the diminishing feature of the substitution
effect [23]. It indicates that the customers with the high CRRA coefficient or willingness
value respond less in DR. The values of a nonlinear exponential function are judged based
on error and trial approach. The value of the degree of altruistic behavior is set to λ̄ = 0.6
arbitrarily for the study purpose.

θ Curtailment Shifting θ Curtailment Shifting

30 1.2 30 1.3
Flexible Demand (%)

Flexible Demand (%)

1.1 1.2
20 20
θ

1.0 1.1
10 10
0.9 1.0

0 0.8 0 0.9
1 16 31 46 61 76 91 106 1 3 5 7 9 11 13 15 17
No. of Customers No. of Customers
(a) (b)

Fig. 4.13 (a)-(e) show the aggregated class-wise customer load patterns obtained using the
PEM, CES, OLG and the proposed OLG (Pro OLG) model relative to load demand of
BDR. The sub-figures indicate that the existing and proposed models effectively exhibit
peak demand curtailment. However, the shifting of curtailed load demand varies under the
different PBDR models. Moreover, it is also subjected to the customer’s load patterns.
Among PBDR models, PEM and CES based DR models shift their curtailed demand
more in the off-peak period than the valley period. On the other hand, the OLG and the
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 80

θ Curtailment Shifting θ Curtailment Shifting

1.2 20 1.3
Flexible Demand (%)

Flexible Demand (%)

20 1.2
15
1.1
1.1

θ
10
10 1.0
1.0
5
0.9

0 0.9 0 0.8
1 2 3 4 5 6 1 5 9 13 17 21 25
No. of Customers No. of Customers
(c) (d)

θ Curtailment Shifting
30 1.2
Flexible Demand (%)

1.1
20

θ
1.0
10
0.9

0 0.8
1 11 21 31 41 51 61
No. of Customers
(e)

Figure 4.12: Customer-wise induced demand flexibility with variation in willingness

parameter for (a) R, (b) LI, (c) MI, (d) C, (e) A

proposed OLG shifts their load demand more in the valley period than off-peak periods.
However, it is partially true for the existing OLG model, where the shifted demand is
more in the off-peak periods such as A class customers. Moreover, it exhibits a varying
response under the different load patterns. On the contrary, the proposed OLG model ap-
propriately exhibits load shifting with altruism behaviour for all class load patterns. This
makes to shift the curtailed demand in the nearest hours. Though, it can be argued that
the response in DR depends upon the relative price of peak, valley, and off-peak hours.
High the relative price ratio means more DR response. It means that most of curtailed
demand should be shifted in the off-peak period ideally. But it is not feasible due to the
customer or appliance’s flexibility in deferring the loads [66, 129].
The class-wise induced DF using PEM, CES, OLG, and the proposed OLG model is
shown in Fig. 4.14 (a)-(d). The sub-figures demonstrate that each PBDR modeling ap-
proach gives distinct demand flexibility under the different customer classes. Fig 4.14
(a) exhibits the partial shifting of the curtailed demand under PEM modeling. It is due
to staticness in the elasticity value, which is usually kept constant for each time period
as considered in the studies [10, 14]. Moreover, it shows higher demand shifting in the
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 81

BDR PEM CES OLG Pro_OLG BDR PEM CES OLG Pro_OLG
1000
Load Demand (kW)

Load Demand (kW)

1400
800
1200
600
1000
400
800
200
600
1 3 5 7 9 11 13 15 17 19 21 23 1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour) Time (Hour)
(a) (b)

BDR PEM CES OLG Pro_OLG

BDR PEM CES OLG Pro_OLG
200
Load Demand (kW)

Load Demand (kW)

150 400

100
200
50

0 0
1 3 5 7 9 11 13 15 17 19 21 23 1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour) Time (Hour)
(c) (d)

BDR PEM CES OLG Pro_OLG

Load Demand (kW)

600

400

200

0
1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour)
(e)

Figure 4.13: Aggregated class-wise load patterns BDR and ADR under the different
PBDR models for: (a) R, (b) LI, (c) MI, (d) C, (e) A

off-peak period than the valley period. It is result of the high relative ratio between off-
peak price to FR price compared to valley to FR price. In Fig. 4.14 (b), the curtailed
demand is completely shifted to the valley and off-peak periods. It maintains the same
level of energy consumption BDR and ADR. However, CES method also does not follow
the customer’s feasible load behavior in DR as observed from the figure. It displays least
or even negative shifting (MI class show decrease in demand) in the valley period. In case
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 82

of the existing two-tier OLG model, the curtailed demand is reflected in the valley period
for almost all customers except A, shown in Fig. 4.14 (c). It is better illustrated using
the proposed OLG model in Fig. 4.14 (d), which shifts the curtailed demand to the valley
and off-peak periods. Further, the shifted demand is relatively higher in the valley period
than off-peak period due to the intertemporal constraint flexibility via altruism behaviour.
Besides, it explicitly imitates the customer’s utility behavior, where the utility diminishes
in the successive time periods [129].

Peak Valley Off-peak Peak Valley Off-peak

Demand flexibility (DF %)

Demand flexibility (DF %)

25 30

20 20
15
10
10
0
5

0 -10
R LI MI C A R LI MI C A
Classes Classes
(a) (b)

Peak Valley Off-peak Peak Valley Off-peak

Demand flexibility (DF %)

Demand flexibility (DF %)

20 20

15 15

10 10

5 5

0 0
R LI MI C A R LI MI C A
Classes Classes
(c) (d)

Figure 4.14: Demand flexibility using (a) PEM; (b) CES; (c) OLG; (d) Proposed OLG
model

The effect of DR on the aggregated load patterns at the system level is shown in Fig. 4.15
(a). It is illustrated using PEM, CES, OLG and the proposed OLG methods. The figure
reveals that load patterns are considerably changed with PBDR applications. The de-
mand is sufficiently reduced during the peak period. However, its reflection on the valley
and peak periods varies in magnitude. It increases during the off-peak period using PEM
and CES, but increase in the valley period is below par. On the other hand, the existing
and proposed OLG manifests appropriate load pattern with slightly better load behaviour
in the latter method. One important observation of the load pattern obtained using the
proposed OLG is that load ADR is increased higher than load BDR at 7: 00 hour. This
reveals a case of rebounding effect in DR, where demand is increased in post DR relative
to BDR load [103].
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 83

BDR PEM CES OLG Pro_OLG LFP LFV LFOP P/V P/OP
4000 1.0 1.5

Load Factor (LF)

0.8
Demand (kW)
3000
1.0

P/V/OPR
0.6
2000
0.4
0.5
1000
0.2

0 0.0 0.0
1 3 5 7 9 11 13 15 17 19 21 23 BDR PEM CES OLGPro_OLG
Time (Hour) Classes
(a) (b)

Figure 4.15: (a) Aggregated load pattern, (b) load factor, and relative ratios under the
different DR models

The impact of DR is further assessed using load factor and the relative ratio between peak
to valley (P/V) and peak to off-peak (P/OP) as shown in Fig. 4.15 (b). Comparing the
relative performance under the different DR methods gives a closer load factor for all DR
models. However, the relative ratios reveal the notable difference between the existing
and proposed DR approaches. The application of DR tries to minimize peak to valley or
off-peak ratio compared to the BDR situation. The highest decrease is observed for the
proposed OLG model. Though, the relative ratios value optimally should be closer to 1.
If it is higher than one even after DR, it indicates the less efficacy of DR application. On
the other hand, a value lower than 1 indicates a rebound effect [103]. Thus, the values of
relative ratios exemplify under-performing DR (i.e., P/V /OP > 1 ) and over-performing
DR (i.e., P/V /OP < 1 ).
The aggregated class-wise customers’ bills BDR, NDR and ADR under the different PBDR
models are summarized in Table 4.5. Here BASE case indicates the BDR and NDR state.
The percentage variations in the subsequent PBDR models are performed relative to NDR
state. The result indicates that all class-wise customers’ bills are increased under BDR to
NDR state except A class customers. It reveals that the shifting from FR to dynamic will
most likely increase the customers bills if the customers do not participate in DR after
its implementation. The bill increase is higher for R, MI and C class customers, whereas
marginal increase for LI customers. If the customers start responding in DR, then their
bills start to decease as observed under ADR state relative to NDR condition. The bill
reduction is higher for PEM and lower for the CES, OLG, and proposed OLG based PBDR
models. The bill reduction in PEM is due to the overall reduction in the energy consump-
tion, whereas it is maintained under CES, OLG and proposed OLG model. Besides, the
overall reduction in the bills under CES, OLG and proposed OLG are uniform.
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 84

Table 4.5: Aggregated customer class bills under different PBDR Models

Methods Class R (¢) LI (¢) MI (¢) C (¢) A (¢) Total (¢)

BDR 178644.38 402724.89 51331.03 209351.07 64345.29 906396.66
BASE NDR 193188.11 406607.77 56434.50 238715.67 63127.91 958073.96
(%) 8.14 0.96 9.94 14.03 -1.89 5.70
ADR 182911.89 383062.16 53484.32 217074.24 61542.07 898074.67
PEM
(%) -5.32 -5.79 -5.23 -9.07 -2.51 -6.26
ADR 187443.06 385862.14 55779.81 235549.37 62101.83 926736.20
CES
(%) -2.97 -5.10 -1.16 -1.33 -1.63 -3.27
BDR 178644.38 402724.89 51331.03 209351.07 64345.29 906396.66
NDR 178234.41 376282.83 51193.17 216044.44 61523.60 883278.45
OLG
ADR 168790.39 363639.55 49234.83 212162.19 58449.79 852276.75
(%) 7.37/-5.30 0.55/-3.36 7.33/-3.83 11.06/-1.80 2.90/-5.00 4.87/-3.51
ADR 186027.17 391478.44 54806.57 232858.80 60808.77 925979.76
Pro OLG
(%) -3.71 -3.72 -2.88 -2.45 -3.67 -3.35

The utility’s profit obtained from the class-wise customers under the different PBDR mod-
els is summarized in Table 4.6. The presented results indicate that utility gains profit from
R, MI, C class customers and losses in LI and A class customers for all DR models except
the two-state OLG model. Further, the existing two-state OLG gives the overall reduction
in the profits. However, the total profit gain is increased ADR for other DR models. The
positive profits demonstrate the optimistic effect of load diversity and price cross-subsidy
among the different customers classes.

Table 4.6: Utility profit under the different PBDR models

Class R (¢) LI (¢) MI (¢) C (¢) A (¢) Total (¢)

Methods
BDR 54143.24 137887.05 15222.46 56943.46 21905.34 286101.56
ADR 64148.46 129723.88 19036.40 77902.08 19750.68 310561.50
PEM
(%) 18.48 -5.92 25.05 36.81 -9.84 8.55
ADR 65476.53 130764.52 19864.51 84689.48 20177.31 320972.35
CES
(%) 20.93 -5.17 30.49 48.73 -7.89 12.19
ADR 47945.74 103895.46 13813.92 61265.46 17344.21 244264.78
OLG
(%) -11.45 -24.65 -9.25 7.59 -20.82 -14.62
ADR 65241.59 133741.50 19486.41 83636.97 19598.73 321705.20
Pro OLG
(%) 20.50 -3.01 28.01 46.88 -10.53 12.44

A comparison of the various existing and the proposed PBDR models based on PEM,
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 85

CUFs and OUFs is summarized in Fig. 4.16 (a)-(b). It is illustrated through the aggre-
gated curtailment/ shifting and is evaluated on the RTP and TOU pricing. The RTP
pricing is applied on the PEM and CUFs, whereas TOU is utilized for all DR functional
forms. The results demonstrate the varying induced demand flexibility via curtailment
and shifting in the different periods. The proposed PEM based PBDR models exhibit the
effective demand adjustment with an adaptive elasticity, which was lacked in the standard
PEM. Though, DPEM lack in representing intertemporal load flexibility in DR. Similar
DR responses are observed for the CUFs based PBDR models. These functions effectively
demonstrate the demand adjustment in the valley and off-peak period with the variations
in their coefficient parameters. However, these functions also exhibit more demand shift-
ing in the off-peak period than the valley periods. In case of OUFs, CES based PBDR
model exhibits alike response to DPEM and CUFs based PBDR models. On the contrary,
the existing and proposed OLG based models exhibit the appropriate load behaviour with
uniform load shifting in the valley and off-peak period. Among, the proposed OLG based
PBDR model displays better load behaviour with adaptive load adjustment and intertem-
poral constraint of load recovery/load flexibility.

Off-peak Valley Peak Off-peak Valley Peak

15
Curtailment/Shifting (%)

15
Curtailment/Shifting (%)

10 10

5 5

0 0
1 2 3 4 5 6 7 1 2 3 4 5 6 7 8 9 10
DR Models DR Models
(a) (b)
1: PEM, 2: DPEM, 3: SPEM, 4: CUF_QUAD, 5: CUF_EXP, 6: CUF_LOG, 7: CUF_POW, 8: CES, 9: OLG, 10: Pro_OLG

Figure 4.16: Comparison of curtailment/shifting under the various PBDR models using
(a) Real-time pricing (RTP), (b) Time-of-use (TOU)

4.8 Summary

An economic PBDR framework using the CUF is proposed incorporating the customer’s
willingness or preference via a satisfaction level and degree of risk. It is modeled for the
various CUFs. Its coefficient attributes are exemplified with the customer’s behavioral
Chapter 4. Price Based DR using Cardinal and Ordinal Utility Functions 86

traits in DR. The developed DR models are investigated on the class-wise customers in
33-bus test distribution load data. The results manifest that all CUFs display appropriate
DR with the customer’s behavioral features. Further, the impact of different CUFs func-
tional forms on DR are assessed under the different load patterns, parameters, price and
risk-based indices. The analyses demonstrate that the different CUFs can be effectively
utilized to model the load behaviour of the diversified customer classes.
An OUF based overlapping generation (OLG) is proposed for PBDR modeling considering
the customer’s willingness and intertemporal load flexibility constraint. It envisions the
customer’s load behavior in DR through the relative change in the price and the risk be-
havior in the form of willingness. It also amalgamates intertemporal flexibility constraint
using human altruistic behavior to imitate load recovery in DR. The comprehensive anal-
yses on the different types of customer class load patterns demonstrate the usefulness of
the proposed DR approach. It indicates that the proposed DR model exhibits feasible DR
behavior. Further, it evinces a uniform load shifting compared to PEM and CES based
DR approach. Most importantly, the proposed OLG approach realizes the customer’s load
behavior using the minimal parameters in DR.
Chapter 5

Hierarchical Price Based Demand

Response Framework in
Distribution systems

5.1 Introduction

DR modeling is an elementary and constructive tool for measuring an individual or

an aggregated response. It enables the system to perform the techno-economic analy-
sis adequately and efficaciously in the post-DR state. Such procedural methods have
been widely utilized in the wholesale electricity market and the distribution systems
(DS) [10, 12, 14, 20, 27]. However, DR has also been establishing itself in a prominent
position with the rapid proliferation of information and communication infrastructures
(ICT) in recent times. This is capacitating DR integration in the system’s operation and
management to enhance its tecno-economic benefits at the wholesale market. It has also
been strengthened on the distribution level with the emergence of the smart grid [18] and
bi-directional advanced metering infrastructure (AMI) [133].
Several works on price-based DR strategies on the optimal decision frameworks have been
widely described in the literature [32, 35]. These PBDR frameworks are performed in
a real-time/day-ahead operation and are primarily explored for improving the system’s
performance without much emphasis on DR integration approach, operational structure,
valuation, etc. Though, DR has also been evolving as a strategic entity or player via
demand response providers (DRPs)/demand response aggregators (DRAs) and large cus-
tomers. This is creating more opportunities for all contributing participants in DR. It

87
Chapter 5. Hierarchical Price Based Demand Response Framework 88

has resulted in the multi-level economic decision model comprising of distribution system
operator (DSO)/ load serving entity (LSE), DRP, customers, etc. This yields a bi-level
problem or Stackelberg game (SG) and a tri-level problem. Generally in DR, bi-level
programming has been broadly demonstrated between LSE/DSO and DRP [35, 36, 70] or
between DSO/DRP and customers [34, 36, 40, 71, 134]. However, these strategic models
have been postulated under the assumptions such as a limited number of customers, fixed
bounds of demands [34], fixed types of customers, either residential or industrial, etc [135].
Similarly, tri-level decision frameworks have also been developed in literature [72–75, 77,
136]. However, these models have also had the shortcoming of accounting for the techni-
cal constraints and representing a broader perspective of customers and DRPs scenarios.
Further, the optimization of bi-level [35, 36, 76] or tri-level problems [40, 73, 75] has usu-
ally been solved using the single equivalence modeling via mathematical programming
or heuristic optimization. However, the equivalence modeling results into the numerous
decision variables in a single level, whereas heuristic optimization lags due to its computa-
tional complexity. This mitigates their viability for optimizing the multi-level large-scale
practical system.
This chapter presents a hierarchical PBDR framework in DS consisting of the LSE, DRPs,
and customers spanning over three levels considering ac network operating constraints.
The formulated problem is modeled as a tri-level, two-stage nonlinear optimization prob-
lem nested with bi-level form in each stage. It comprises LSE and DRP in the first stage
and DRP and customers in the second stage. The interaction among players in each stage
is established in a non-cooperative game-theoretic framework using a sequential Stackel-
berg game. Moreover, a hybrid pricing model comprising dynamic and static pricing is
proposed for flexible and non-flexible loads, respectively. For the solution process, a re-
formulation & decomposition (R&D) method based on the column constraint generation
(CCG) is employed to solve the problem. The proposed problem is investigated on the
standard 33-bus test and a real 108-bus Indian distribution system. Further, the scalabil-
ity and tractability of the algorithm are thoroughly assessed on the small to large-scale
problem and is compared with the existing solution approach.

5.2 Proposed Hierarchical PBDR Framework

The proposed hierarchical framework is formulated as a tri-level, two-stage problem within

the sphere of DS. It encompasses three entities, LSE, DRPs and the customers. LSE
oversees the operation and management of the network, and DRPs acts as an intermediary
between LSE and the customer for DR interaction. In the problem, LSE and multiple
Chapter 5. Hierarchical Price Based Demand Response Framework 89

DRPs operate in the first stage; DRPs and customers are in the second stage. Each DRP
is assumed to negotiate with LSE on behalf of the customers to set the dynamic price
signal.

5.2.1 Upper-level Problem: LSE’s Profit

LSE is at the top of the hierarchy, procuring the wholesale market’s power and selling it
to customers. It is assumed that LSE levied the electricity charges to the customers in two
ways based on the types of loads. Fixed retail price for inflexible load and dynamic price
for the flexible load. In view of this, LSE’s objective subject to ac power flow constraints
is expressed as follows:
X
∆t ρFt R PtIF + ρDR F M CP
 
max ZLSE = t Pt − ρt PG,t (5.1)
ρDR
t ,PG,t ,e,f
t∈ΩT

subject to:

PtIF + PtF = Pt ∀t (5.2)

X X k,F X X
dk,F
X X
PtF = dFn,t = dn,t = i,n,t ∀t (5.3)
n∈ΩN k∈ΩK n∈ΩNk k∈ΩK n∈ΩNk i∈ΩIn

ρDR
t
≤ ρDR
t ≤ ρDR
t , : ϑt , ϑt ∀t (5.4)
P
PG,n,t − PL,n,t = {en,t (gnm em,t − bnm fm,t ) +fn,t (bnm em,t + gnm fm,t )} ,
m (5.5)
: λn,p,t ∀n ∈ ΩN , ∀t ∈ ΩT
P
QG,n,t − QL,n,t = {fn,t (gnm em,t − bnm fm,t ) −en,t (bnm em,t + gnm fm,t )} ,
m (5.6)
: λn,q,t ∀n ∈ ΩN , ∀t ∈ ΩT
IF F
PL,n,t = Pn,t + Pn,t , ∀n ∈ ΩN , ∀t ∈ ΩT (5.7)
QL,n,t = QIF F
n,t + Qn,t , ∀n ∈ ΩN , ∀t ∈ ΩT (5.8)
2
|V n |2 ≤ Vn,t
2
(en , fn ) ≤ V n , :ν n,t , ν n,t ∀n ∈ ΩN , ∀t ∈ ΩT (5.9)
2 2
max
Sbnm,t (e, f ) ≤ Sbnm,t (e, f ) , :µnm,t ∀(n, m) ∈ ΩL , ∀t ∈ ΩT (5.10)

P G,n,t ≤ PG,n,t ≤ P G,n,t , : π n,p,t ,π n,p,t ∀t ∈ ΩT (5.11)

QG,n,t ≤ QG,n,t ≤ QG,n,t , : π n,q,t ,π n,q,t ∀t ∈ ΩT (5.12)

The objective function (5.1) comprises inflexible and flexible loads charged at the fixed
retail price and dynamic rate, respectively, against the power transaction from the grid
at the wholesale market price. Equation (5.2) denotes the composition of total active
Chapter 5. Hierarchical Price Based Demand Response Framework 90

demand. Expression (5.3) represents the total flexible demand occurring at the utility
level. It is elucidated using three different expressions for clarity. Similarly, reactive
demand is derived. The dynamic prices are constrained by limits against the available
flexible demand (5.4). Distribution power flow equations and its constraints are described
in the rectangular coordinate. It is based on easiness in the formulation and inexpensive
computation in the rectangular coordinate instead of polar coordinate [83,137]. The active
and reactive power flow equations are stated in constraints (5.5)-(5.6). Active and reactive
demand at bus n is elaborated in (5.7)-(5.8) to define the flexible and inflexible demand.
Constraints (5.9)-( 5.10) define the bus voltage limits and feeders’ line flow limits in either
direction. The active and reactive power generation limits are accounted in constraints
(5.11)-(5.12). The dual variables are written after the colon in each constraint.

5.2.2 First Lower-Level Problem: (DRP k) Model

DRPs communicate the information to the customers passed on by LSE. Each customer
class is assumed to be overseen by an independent DRP for DR interaction. The interac-
tions among the DRPs are not considered in work. The DRP acquires the flexible loads
from LSE at the dynamic price and sells them to the customers by adding its benefit to
that price. Since, in the proposed formulation, the different customer classes such as resi-
dential (R), large industry (LI), medium industry (MI), commercial (C), and agricultural
(A) have been assumed. Therefore, each class has set the procurement and selling price
distinct. It reflects the practical pricing rates for the different customer classes in DS. The
different pricing rates are based on the activity usage, demand capacity, and cross-subsidy
between the various customer classes [34]. Suppose κk is subsidizing (negative value)/
overcharging (positive value) factor is the power transactions from LSE to the different
DRPs; then, the price perceived by the customer classes is obtained using an expression
k,DR/F R DR/F R
ρt = ρt × (1 + κk ). Then, kth DRP’s objective that maximizes its profit
against the set dynamic price by LSE over the time horizon is defined as follows:

′
ρtk,DR dk,F ρk,DR dk,F
X X X X
k
max
′
ZDRP = ∆t i,n,t − ∆t t i,n,t ∀k ∈ ΩK
ρk,DR
t ,dk,F
i,t t∈ΩT i∈ΩI k t∈ΩT i∈ΩI k

(5.13)
k,DR′ k,DR′ k,DR′ k,DR′ k,DR′
s.t. ρt ≤ ρt ≤ ρt , :µt , µt ∀t, ∀k (5.14)

Equation (5.13) defines DRP’s profit against the payment made to LSE. dk,F
i,n,t is the flexible
load of ith customer on bus n at time t under DRP k and ΩIk is set of the enrolled
Chapter 5. Hierarchical Price Based Demand Response Framework 91

customers under kth DRP. The dynamic price offered to the customers is bounded by
(5.14). The summation of the flexible load for the enrolled customers under DRP k at
time t is expressed in (5.15).

dk,F dk,F dk,F

X X X X
t = i,n,t = i,n,t (5.15)
i∈ΩI k k∈ΩK n∈ΩNk i∈ΩIn

5.2.3 Second Lower-Level problem: (Customer i) Model

Once the dynamic price is availed to the customers by their respective DRPs, the enrolled
customers start adjusting their flexible loads in response to set out dynamic prices. Then,
the customers redistribute its load demand through its utility against the dynamic price
while maintaining the total energy consumption over time horizon constant before DR
(BDR) and after DR (ADR) [138]. The objective of each customer is expressed as follows:

k,DR′ k,F
n o
∆t U (dk,F
X
max Zik = i,n,t ) − ρt di,n,t , ∀i ∈ ΩI (5.16)
dk,F
i,n,t t∈ΩT
k,F k
s.t. dk,F k,F k
i,n,t ≤ di,n,k,t ≤ di,n,t , : λi,n,t ,λi,n,t ∀i, ∀t, ∀k (5.17)
dk,F
i,n,t − dk,F
i,n,t−1 ≤ k,U
Ri,n,t , :λk,U
i,n,t ∀i, ∀t, ∀k (5.18)
dk,F k,F k,D k,D
i,n,t−1 − di,n,t ≤ Ri,n,t , :λi,n,t ∀i, ∀t, ∀k (5.19)
X k,F k,F
∆t di,n,t = Ei,n,o , : λki ∀i, ∀t, ∀k (5.20)
t∈ΩT
k,F
dk,F
X
Ei,n,o = ∆t i,n,t,o (5.21)
t∈ΩT

The customer’s objective (5.16) consists of two terms, the utility’s satisfaction function
and customer payment to DRP. Constraint (5.17) denotes the upper and lower bound
of the flexible load. Constraints (5.18)-(5.19) represent the ramp-up/down limit of load
adjustment between two consecutive time-states as an intertemporal constraint. Power
balance constraint is set equal to the energy consumption BDR in (5.20). It is deliberated
in (5.21). The dual variables of constraints (5.17)-(5.20) are also mentioned after the colon.

2
(dk,F
i,n,t )
U (dk,F
i,n,t ) = k
αi,n,t dk,F
i,n,t − k
βi,n,t ∀i, ∀t, ∀k, ∀n (5.22)
2

Equation (5.22) states the customer’s utility as a quadratic, concave function with a di-
k
minishing marginal utility. In the utility function, coefficient αi,n,t indicates satisfaction
k
gain from consumption, and βi,n,t defines the degree of risk-aversion [23].
Chapter 5. Hierarchical Price Based Demand Response Framework 92

The proposed problem formulation reveals that each competing player trade-off its op-
timization variables with the subsequent levels. This leads to an interaction mechanism
among the participating players through a sequential process at the different levels. The
architecture of the proposed hierarchical PBDR framework is described in Fig. 5.1.

Figure 5.1: Hierarchical architecture of the proposed PBDR framework

The proposed study presents a two-stage Stackelberg game (SG) model to capture in-
teractivity among the participating players. In the first stage, LSE operates as a leader
and DRPs as followers; in the second stage, DRP acts as a leader and the customers as
the followers. The desired outcome of these interactions gives rise to Stackelberg equi-
librium (SE). The outcome of SE depends upon the strategies revealed by the various
∗ ∗ ∗ ∗
participating players. Let PFT = [P1F , P2F , ......., PΩFT ] the set of flexible demand strate-
∗ ∗ ∗ ∗
gies of LSE, the set of dynamic prices put forth by LSE, ρDR
T = [ρDR
1 , ρDR
2 , ......., ρDR
ΩT ]
′∗ ′∗ ′∗ ′∗
and ρk,DR
T = [ρk,DR
1 , ρk,DR
2
k,DR
, ...., ρΩ T
] set of dynamic prices broadcasted by DRP to
∗ ′∗ ∗
the customers. Then the amalgamation of set of an optimal strategy (PFT , ρDR
T , ρDR
T )
Chapter 5. Hierarchical Price Based Demand Response Framework 93

constitutes SE if the following inequalities at each level hold.

∗ ′∗ ∗ ′∗
Zik (dk,F k,DR
i,T , ρT ) ≥ Zik (dk,F k,F k,DR
i,T , d−i,T ρT ) i∀ΩIk (5.23)
∗ ′∗ ∗ ′ ∗
k
ZDRP (dk,F k,DR
T , ρT
k,DR
, ρT k
) ≥ ZDRP (dk,F k,DR
T , ρT , ρk,DR
T ) k∀ΩK (5.24)
F∗ DR′∗ DR∗ DR∗
ZLSE (PT , ρT , ρT ) ≥ ZLSE (PFT , ρT ,ρDR
T ) (5.25)

∗ k,F ∗ k,F ∗ k,F ∗ k,F ∗ k,F ∗

h i
where, dk,F
−i,T = d 1,T , d 2,T , ......, di−1,T , d i,T , d Ik ,T denotes the set of strategies of all
the customers enrolled under kth DRP except ith customer. Similarly, the set of strategies
formed at each DRP level will be the aggregated demand from each customer under DRP k
∗ ∗ ∗ ∗
defined as dk,F
T = [d1k,F , dk,F
2 , ....., dk,F
ΩT ]. The following inequalities (5.23)-(5.25) reveal
that all the participating players cannot get better pay-off if they deviate unilaterally from
their equilibrium points. If they do, then SE does not constitute [39].

5.3 Stackelberg Equilibrium

To deduce SE, backward induction is utilized based on looking ahead and working back-
ward with a sequential rationality [136]. In this, the first optimal response of the customers
′
is obtained against the revealed strategy of DRP to customers (i.e., ρDR
k,T ) at the lower
level. Once the best response of each customer is known to DRP, DRP set its optimal
′
DR ). Then, DRP reveals its strategy to LSE, which set its optimal strat-
strategy (i.e., ρk,T
egy in response. A proposition is put forwarded to verify SE’s existence and uniqueness.
Besides, bus index n is dropped for brevity.
Proposition: In a two-stage SG, three participant players with a set of their strategies
′
(PFT , ρT
DR , ρDR ) will constitute a SE, if the following conditions are satisfied.
T

(i) The set of the strategy of each participating player is non-empty

′∗ ′
(ii) Zik is continuous and concave over dk,F k,DR
i,T . For each dT ∈ dk,DR
T , a unique solution
∗
dk,F k,F
i,T ∈ di,T exists.
′ ∗ ∗
k
(iii) ZDRP is continuous and concave over dk,DR
T . For each (ρk,DR
T , dk,F k,DR
T ) ∈ ρT ×
′∗ ′
dk,F k,DR
T , a unique solution ρT ∈ ρk,DR
T exists.
′∗ ∗ ′
(iv) ZLSE is continuous and concave over ρDR DR , PF ) ∈ ρDR × PF , a
T . For each (ρT T T T
∗
unique solution ρDR
T ∈ ρDR
T exists.

Proof: a) Obtain the best response of each customer in response to DRP revealed strategy
′
(i.e., dynamic price ρk,DR
T )
Chapter 5. Hierarchical Price Based Demand Response Framework 94

In order to find the optimal response of each customer, first (5.16) is reformulated as
unconstrained optimization using the Lagrange multipliers to accommodate all equality
and inequality constraints. It gives the customer’s objective function as follows [96]:
 
 
k,DR′ k,F
n o
∆t U (dk,F k,F k,F k,F k,F
X X
Zik = i,t ) − ρt di,t + λk
i ∆tdi,t − E k
i,o  − λi,t (di,t − di,t )

t∈ΩT t∈ΩT
k k,F
+ λi,t (dk,F k,U k,F k,F k,U k,D k,F k,F k,D
i,t − di,t ) + λi,t (di,t − di,t−1 − Ri,t ) + λi,t (di,t−1 − di,t − Ri,t ), ∀i ∈ ΩIk
(5.26)

The Karush-Kuhn Tucker (KKT) condition for the optimal demand, gives the demand of
ith customer of class k at time t as follows:
′ k
k − ρk,DR − λk − λk + λ
αi,t t i i,t i,t λk,U k,U k,D k,D
i,t − λi,t+1 + λi,t+1 − λi,t
dk,F
i,t = + , ∀i ∈ ΩIk (5.27)
βik βik

Furthermore, a second derivative is obtained to prove the convexity of function as below:

∂ 2 Zik
2 = −βik , ∀i ∈ ΩIk , ∀t ∈ ΩT (5.28)
∂dki,k
.
Since, βi,k > 0, i.e., ∂ 2 Zi,k ∂d2i,k,t < 0, then the customer’s objective function (i.e., Zik ) is
a strictly concave function over the feasible region of dk,F
i,T . This ensures the existence and
uniqueness of the solution. Hence, the customer’s objective function corresponding to a
∗ ′
unique solution ρk,DR
T will also be unique as per the proposition (ii).
b) Obtain the best strategy in response to the customer’s optimal solution
Once the customer’s best response is calculated, DRP’s best strategy can be obtained by
substituting the customer’s best response (5.27) into (5.13) according to the backward-
induction principle and a binding constraint of DRPs price. This reformulation gives kth
DRP objective as follows:
 ′
k − ρk,DR − λk − λk + λk λk,U k,U k,D k,D
)
i,t − λi,t+1 + λi,t+1 − λi,t
X X  αi,t t i i,t i,t
k
ZDRP = ∆t +
 βi,k βi,k
t∈ΩT i∈ΩIk
′ ′ ′ ′ ′ ′
  ′
× ρk,DR
t − ρ k,DR
t − µk,DR
t (ρk,DR
t − ρk,DR
t
) + µk,DR
t (ρk,DR
t − ρk,DR
t ) (5.29)
Chapter 5. Hierarchical Price Based Demand Response Framework 95

The best strategy of DRP is obtained by taking the first derivative of (5.29) as follows:

k
P αki,t −λki −λki,t +λi,t P λk,U k,U k,D k,D
i,t −λi,t+1 +λi,t+1 −λi,t
βik βik
′ 1 i∈ΩIk 1 i∈ΩIk
ρtk,DR = P 1 + P 1
2 2
βik βik
i∈ΩIk i∈ΩIk
′
k,DR′
− µk,DR
ρk,DR 1 µt t
+ t − P 1 (5.30)
2 2
βik
i∈ΩIk

Now to verify the convexity of DRP’s best strategy, a second derivative of (5.29) is calcu-
lated:

∂ 2 ZDRP
k X 1
′2 = −2 < 0, ∀i ∈ ΩK , ∀t ∈ ΩT (5.31)
∂ρk,DR
t i∈Ω
βik
Ik

P 1 ∂ 2 ZDRP
k
Since, βik
> 0 i.e., ′2 < 0, then DRP’ solution will also be unique and optimal.
i∈ΩIk ∂ρk,DR
t

This makes DRP’s function optimal to a unique solution as per the proposition (iii).
c) To verify the uniqueness and optimal solution of LSE
Once the best strategies of both customer and DRP are revealed, then the best strategy
of LSE can be evaluated through the backward-induction principle. To obtain this, first,
the following sub-steps are performed.
The first DRP’s best response is substituted into the customer’s best response function
(5.27) to calculate the flexible demand. This gives optimal flexible demand of the customer
enrolled under DRP k at time t as:

k
k,F ∗
k − λk − λk + λ
αi,t i i,t i,t λk,U k,U k,D k,D
i,t − λi,t+1 + λi,t+1 − λi,t
di,t = +
βik βik
k
P αki,t −λki −λki,t +λi,t P λk,U k,U k,D k,D
i,t −λi,t+1 +λi,t+1 −λi,t
βik βik
1 i∈ΩIk 1 1 i∈ΩIk 1
− 1 − 1
βik
P P
2 βi,k 2
βik βik
i∈ΩIk i∈ΩIk
′ ′
1 ρk,DR
t 1 µk,DR
t − µk,DR
t
− + 1 (5.32)
2 βik
P
2 k βi
i∈ΩIk
Chapter 5. Hierarchical Price Based Demand Response Framework 96

Next, aggregate the customer’s flexible demand under each DRP and summation of all
DRP’s demand gives the total flexible demand at LSE as follows:

′ ′
∗
X ∗ 1 X k,DR X 1 1 X X µk,DR − µk,DR
PtF = dkDRP,t =− ρt + t
P 1 t

k∈ΩK
2
k∈Ω i∈Ω
βik 2
k∈Ω i∈Ω βk
K Ik K Ik i
i∈ΩIk

k − λk − λk + λk k,U k,U k,D k,D

1 X X αi,t i i,t i,t 1 X X λi,t − λi,t+1 + λi,t+1 − λi,t
+ + (5.33)
2 βik
k∈ΩK i∈ΩIk
2 βik k∈ΩK i∈ΩIk

Expanding ρk,DR
t in terms of LSE dynamic price (ρDR
t ) as follows:

ρk,DR
X X X
t = ρDR
t × (1 + κk ) = ρDR
t (1 + κk ) (5.34)
k∈ΩK k∈ΩK k∈ΩK

Now, substituting (5.34) into (5.33) gives the complete form of flexible demand at the LSE
level in terms of price as follows:

∗ ρDR X X (1 + κk ) 1 X X µk,DR′ − µk,DR′

t
PtF =− + t
P 1t
2 βi,k 2 β
k∈ΩK i∈ΩIk k∈ΩK i∈ΩIk i,k
k∈ΩK
k − λk − λk + λk k,U k,U k,D k,D
1 X X αi,t i i,t i,t 1 X X λi,t − λi,t+1 + λi,t+1 − λi,t
+ + (5.35)
2 βik
k∈ΩK i∈ΩIk
2 βik k∈ΩK i∈ΩIk

The Lagrange multipliers related to equality and inequality constraints are attributed to
the constants and are written in the form of notation for clarity. Similarly, the customers’
utility coefficients are summarized in the compact form using the following notations:

k
X X αi,t X X (1 + κk ) X X µk,DR′ − µk,DR′
t
Γ= > 0;Λ= > 0; Ξ = P 1t
k∈ΩK i∈ΩIk
βik k∈Ω i∈Ω
β k
i k∈Ω i∈Ω βk
K Ik K Ik i
i∈ΩIk
k
λk,U k,U k,D k,D
( )
X X −λki − λki,t + λi,t i,t − λi,t+1 + λi,t+1 − λi,t
Υ= + (5.36)
k∈ΩK i∈ΩIk
βik k
βi,t

This gives a compact form of the total flexible demand occurring at LSE at each hour t
and is expressed as follows:

1
PtF = −ρDR


t Λ+Γ+Υ+Ξ (5.37)
2

Equation (5.37) reveals that the flexible demand is a function of dynamic price. It indi-
cates that a high dynamic price will cause a reduction (curtailment) in the flexible demand,
Chapter 5. Hierarchical Price Based Demand Response Framework 97

and a low dynamic price will induce increment (shifting) flexible demand. Now, substitut-
ing (5.37) into (5.1) provides reformulated LSE’s objective function. This defines LSE’s
objective function as follows:
  
DR 1
X
DR
+ ρFt R PtIF + −ρM CP


ZLSE = ∆t ρt −ρt Λ + Γ + Υ + Ξ t Pt
2
t∈ΩT

− ϑt (ρDR
t − ρDR
t
) + ϑt (ρDR
t − ρDR
t ) (5.38)

The convexity condition is confirmed by calculating the second derivative of (5.38) as

follows:

∂ 2 ZLSE
2 = −Λ < 0, ∀t ∈ ΩT (5.39)
∂ρDRt

∂ 2 ZLSE
Since, Λ > 0 i.e., 2 < 0, thus, the optimal solution of LSE is unique and strictly con-
∂ρDR
t
cave function. This completes the proof of proposition (iv) and satisfies all SE conditions.
The existence and uniqueness of optimal global solution of LSE, DRP, and the customers
indicate that the proposed interactive SG will cater SE. However, the steps mentioned
earlier for SE deduction are based on when all participants have global knowledge of one
another and are performed centrally. Though, it is not feasible in the practical operation.
Therefore, a sequential process is proposed to achieve the optimality condition.

5.4 Solution Approach

In this section, the tri-level problem is formulated using the decomposition-based R&D
approach. It is adaptively extended to the tri-level optimization using a nested R&D
algorithm. To implement, first, the proposed model is decoupled as the master-slave
problem, where the first stage is defined as a master problem (MP) and the second stage
as the slave problem (SP). Likewise, two levels of the second stage are transformed into
an inner-master problem (IMP) and inner-slave problem (ISP). Then, the problem is
optimized using an iterative procedure until the convergence tolerance is satisfied. The
complete model of a tri-level two-stage problem is expressed as follows:
X
∆t ρFt R PtIF + ρDR F M CP

Φ= max t Pt − ρt PG,t (5.40)
x={ρDR
t ,PG,t ,e,f }
t∈ΩT

s.t. [Constraints (5.4) − (5.12)] (5.41)

Chapter 5. Hierarchical Price Based Demand Response Framework 98

′
ρk,DR dk,F ρk,DR dk,F
X X X X
F (y, z) = max
′
∆t t i,n,t − ∆t t i,n,t ∀k ∈ ΩK
y={ρk,DR
t },z={dk,F
i,n,t } t∈ΩT i∈ΩI k t∈ΩT i∈ΩI k

(5.42)
s.t. [Constraint (5.14)] (5.43)
k,DR′ k,F
n o
dk,F ∆t U (dk,F
X
i,n,t ∈ arg max i,n,t ) − ρt di,n,t , ∀i ∈ ΩI (5.44)
z={dk,F
i,n,t } t∈ΩT

s.t. [Constraints (5.17) − (5.20)] (5.45)

The LSE’s objective is handled as MP, and DRPs and customers combined are marked as
SP. In SP, two different entities DRPs and customers, are re-casted as IMP and ISP, sequen-
tially. The optimization decision variables of MP (i.e., LSE) are x = {ρDR
T , PG,t , QG,n,t , e, f }.
In SP, IMP’s (i.e., DRPs) decision variable is y = {ρk,DR
1 , ρk,DR
2 , ...., ρk,DR
T }, and ISP’s (i.e.,
the customer) decision variable is z = {dk,F k,F k,F
i,n,t , di,n,t , ...., di,n,T }, respectively. The program
is first initiated by the slave problem based on the received initial dynamic price signal
from LSE.

5.4.1 Slave Problem

∗
To solve a tri-level, the first presumed optimal result x∗ = {ρDR
T } of MP is substituted
∗
into SP by transforming it into DRP’s price (i.e.,ρk,DR
T ) as follows:

′ ∗
ρk,DR dk,F ρk,DR dk,F
X X X X
F (y, z) = max
′
∆t k,t i,n,t − ∆t t i,n,t ∀k ∈ ΩK
ρk,DR
t ,dk,F
i,n,t t∈ΩT i∈ΩI k t∈ΩT i∈ΩI k

(5.46)
s.t. [Constraint (5.14)] (5.47)
k,DR′ k,F
n o
dk,F ∆t U (dk,F
X
i,n,t ∈ arg max i,n,t ) − ρt d i,n,t , ∀i ∈ ΩI (5.48)
dk,F
i,n,t t∈ΩT

s.t. [Constraints (5.17) − (5.20)] (5.49)

Chapter 5. Hierarchical Price Based Demand Response Framework 99

5.4.1.1 Inner-Master Problem

Now, the equivalent formulation of SP as IMP is reformulated by duplicating the decision

variables and constraints in ISP as follows [79]:

′ ∗ k,F
ρk,DR dek,F ρk,DR
X X X X
IMP :Ψ(x∗ , y, z) = max
′
∆t t i,n,t − ∆t t dei,n,t
ρk,DR
t ,dek,F
i,n,t t∈ΩT i∈ΩI k t∈ΩT i∈ΩI k

(5.50)
′ ′ ′
s.t. ρtk,DR ≤ ρk,DR
t ≤ ρk,DR
t ∀t, ∀k (5.51)

ek,F 2

ek,F − β k (di,n,t ) −ρk,DR′ dek,F ≥ −λk,(l) E k,F − λk,(l) dk,F
o
k d
P
∆t αi,n,t i,n,t i,n 2 t i,n,t i i,n,o i,k,t i,n,t
t∈ΩT (5.52)
k,(l) F k,U,(l) k,U k,D,(l) k,D
+λi,t di,n,t − λi,t Ri,n,t − λi,t Ri,n,t ,1 < l < q, i ∈ ΩI
k − β k dk,F − ρ k,DR′ k,(l) k,(l) k,(l) k,U,(l) k,U,(l) k,D,(l)
αi,t i i,t t − λi − λi,t + λi,t + λi,t − λi,t+1 + λi,t+1
k,D,(l) (5.53)
−λi,t = 0, 1 < l < q
k,(l) k,(l) k,(l) k,U,(l) k,D,(l)
λi ≥ 0, λi,t ≥ 0, λi,t ≥ 0, λi,t ≥ 0, λi,t ≥ 01 < l < q (5.54)
[Constraints (5.17) − (5.20)] (5.55)

where, dek,F
i,n,t is a duplicate decision variable. The constraints (5.52)-(5.54) are the equiva-
lent formulation of (5.16)-(5.20) and are obtained by the primal-dual formulation to over-
come the computational intensiveness of KKT based formulation [79]. It is to be noted
that duplicate decision variable dek,F is not only feasible to IMP but will also be optimal
i,n,t
∗
to ISP for the given x∗ = DR
{ρT } [79].

5.4.1.2 Inner-Slave Problem

′∗
Now, the optimized result y ∗ = {ρk,DR
T } obtained in IMP is substituted in ISP. ISP is
further decomposed into two sub-problems using the CCG method. The first inner-slave
problem (ISP1) is defined as follows:

k,F 2
n (di,n,t ) ′∗

dk,F k,DR k,F
X
ISP1 :φ(x∗ , y ∗ , z) = max k
∆t αi,n,t i,n,t −β k
i,n − ρt di,n,t  (5.56)
dk,F
i,n,t t∈ΩT
2

s.t. [Constraints (5.17) − (5.20)] (5.57)

The results obtained through ISP1 may have many solutions, which may not be feasible to
IMP. Hence, one feasible solution that satisfies IMP is obtained by the second inner-slave
Chapter 5. Hierarchical Price Based Demand Response Framework 100

problem (ISP2) expressed as follows:

′∗ ∗
ρk,DR dk,F ρk,DR dk,F
X X X X
ISP2 :ϕ(x∗ , y ∗ , z) = max ∆t t i,n,t − ∆t t i,n,t ∀k ∈ ΩK
dk,F
i,n,t t∈ΩT i∈ΩII k t∈ΩT i∈ΩI k

(5.58)

2

k,F (dk,F
i,n,t ) ′∗ k,F
o
−ρk,DR
X
k k ∗ ∗
∆t αi,n,t di,n,t − βi,n t di,n,t ≥ φ(x , y , z) (5.59)
 2
t∈ΩT

[Constraints (5.17) − (5.20)] (5.60)

This gives the optimal result z ∗ = {dk,F

i,n,T } feasible to ISP1 and ISP2. The obtain result
ensures that it will also be optimal for the given y ∗ in IMP based on the optimistic
consideration [38]. This demand now will be reflected as the aggregated demand at buses
or under DRPs and finally at the system level as follows:

∗ ∗ ∗ ∗
dk,F dk,F
X X X X X X
PtF = dFn,t = n,t = i,n,t ∀t (5.61)
n∈ΩN k∈ΩK n∈ΩNk k∈ΩK n∈ΩNk i∈ΩIn

∗ ∗ ∗
Hence, dk,F k,F
i,n,T , dt and PtF are bracketed in z ∗ interchangeably based on the stage of
the problem. Now, substitute optimal tuple (y ∗ , z ∗ ) of SP into MP, gives the following
expression:
n o
F∗
X
Φ = max ∆t ρFt R PtIF + ρDR
t P t − ρM CP
t P G,t (5.62)
t∈ΩT

[Constraints (5.4) − (5.12)] (5.63)

The optimal solution (x∗ , y ∗ , z ∗ ) is obtained on solving MP, satisfying the SP. This process
continues iteratively between MP and SPs until the convergence condition is satisfied. The
complete procedure of nested R&D is described in algorithm 5.1:
Notes that in the proposed model, the objectives are nonlinear functions as (5.13) and
(5.16) are in bi-linear and quadratic form, respectively. In addition, distribution power flow
incorporated in the system is formulated as a nonlinear programming (NLP). Therefore,
the commercial off-the-shelf NLP solver, interior-point optimizer (IPOPT), is leveraged to
solve the proposed tri-level optimization problem.
Chapter 5. Hierarchical Price Based Demand Response Framework 101

Algorithm 5.1 Procedure of Nested R&D algorithm for tri-level problem

1) Step 1. Set LB = −∞, U B = ∞, iteration counter v and ϵ, initialize x(0) and Φ(0)
as feasible solution to MP.
2) Step 2. Set lb = −∞, ub = ∞, iteration counter q and ε.
∗ ∗
3) Step 3. Solve IMP for the optimal value x(v) and derive its optimal solution y (q+1)
and update lb = Ψ(q+1) .
∗
4) Step 4. With y (q+1) from step 5, solve ISP1 to obtain φ(q+1) .
∗ ∗
5) Step 5. With y (q+1) and φ(q+1) , solve ISP2 to obtain z (q+1) and ϕ(q+1) , update
ub = min ub, ϕ(q+1) .

6) Step 6. if ub−lb ub ≤ ε , return U B = Φ(q) + ub and corresponding optimal solution

∗ ∗
y (q) , z (q) of SP, otherwise q = q + 1 and go back to step 3.


∗ ∗
∗
7) Step 7. With the solutions y (q) , z (q) , solve MP to obtain optimal solution x(v+1)
and Φ(v+1) and update LB = Φ(v+1) + ub.
8) Step 8. if U B−LB LB ≤ ϵ , return U B and corresponding optimal solution
∗ ∗ ∗

x(q) , y (q) , z (q) of MP, otherwise v = v + 1 and go back to step 2.

5.5 Simulation Results and Discussions

In this section, the proposed framework and the solution method are conducted on the
standard 33-bus test system [106] and a large real 108-bus Indian distribution system [139].
The considered systems’ data information is given in Table A.1 and Table A.2 of Ap-
pendix A. The class-wise details such as customer nodes, total allocated demand, number
of customers, and S/CF for the considered test systems are given in Table B.1 and Ta-
ble B.2 in Appendix B. The power flow studies parameters are described in Table B.4 of
Appendix B. The load pattern is segmented into three periods: off-peak (0:00-7:00, 23:00-
24:00), valley: (11:00-17:00) and peak period: (8:00-11:00, 18:00-22:00) for the analysis
purpose.
The flexible demand of customers is obtained probabilistically using the truncated normal
distribution to reflect the diversity in the customer’s behaviour in DR participation. It
k ∼ N (µ , σ ), where ϖ k is flexible demand fac-
is evaluated using the expression ϖi,n ϖ ϖ i,n
tor of ith customer of nth bus under DRP k. The DR participation rate is assumed to
vary in the range of 0-30 %. This gives flexible demand and is conveyed to DRPs. The
remaining inflexible demand is obtained using the expression (dk,IF k k,F
i,n,t = di,n,t − di,n,t ). The
LSE’s offered price is supposed to vary within the interval (ρDR,min T , ρDR,max
T ). It is set
DR′ ,min ′
to (ρT
RT P , 1.2ρ
T
RT P ). Similarly, DRP’s price limits set to lie within (ρ T , ρDR ,max ) ∼
T =
RT P RT P ′
(0.8ρT , ρT ) with the condition of (ρDR
T ≥ ρDR
T ) for allocating monetary benefits
to DRPs in lieu of brokering between LSE and customers. Each customer class’s retail
rate/flat rate (ρFTR ) is set above the average electricity rate to accommodate the retailer’s
compensation for buying electricity at a variable rate and selling at the fixed rate [14]. It
Chapter 5. Hierarchical Price Based Demand Response Framework 102

is taken as 1.1 times the average of obtained optimal price schedule. The ramp-up/down
limits of the flexible demand are set proportionally to the fraction of its maximum demand.

5.5.1 Standard 33-bus test system

The standard 33-bus test distribution system is rated at 12.66 kV with active demand
of 3.715 MW and reactive demand of 2.30 MVAr. The optimal pricing schedules of LSE
and DRA over the scheduling horizon obtained using the proposed nested algorithm are
shown in Fig. 5.2. It can be observed from the figure that DRA follows up the LSE price
profile synchronously. Further, the price during peak hours is realized to be high, while
low during off-peak/valley hours. This high price will induce load curtailment during peak
hours, and a low price will accommodate the curtailed demand in off-peak/valley hours.

LSE RATE DRP RATE

30
Price (¢/kWh)

20

10

1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour)

Figure 5.2: Pricing schedule of LSE and DRP for flexible loads

5.5.1.1 Effect of PBDR on Class-wise Load Pattern

Fig. 5.3 (a)-(e) displays the aggregated class-wise load patterns BDR and ADR. The ag-
gregated load represents the summation of flexible and inflexible demand. It can be shown
from the figures that each customer class responds to high prices by curtailing the load
and shifting the load during low prices. Moreover, a notable observation on load pattern
ADR is that all customer classes shift their demands to valley period then off-peak period
excluding class A customers. These load patterns also reflect a feature of intertemporal
constraint, modeled by the ramp-up/down constraints. This makes customers shift most
of the curtailed demand to the nearest hours (i.e., valley period). A better visualization
Chapter 5. Hierarchical Price Based Demand Response Framework 103

of class-wise contribution in load curtailment and shifting in the segmented periods is

illustrated in Fig. 5.3 (f). It shows that all classes actively respond in DR. In addition,
all customer classes except A class customers exhibit almost uniform load shifting in the
valley and off-peak period.

BDR_R ADR_R 1600 BDR_LI ADR_LI

Load Demand (kW)

Load Demand (kW)

1000 1400

1200

500 1000

800
1 3 5 7 9 11 13 15 17 19 21 23 1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour) Time (Hour)
(a) (b)

BDR_MI ADR_MI BDR_C ADR_C

Load Demand (kW)

Load Demand (kW)

600
200

400
100
200

0 0
1 3 5 7 9 11 13 15 17 19 21 23 1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour) Time (Hour)
(c) (d)
Curtailment/Shifting (%)

BDR_A ADR_A Off-peak Valley Peak

Load Demand (kW)

800 10
600

400 5

200

0 0
1 3 5 7 9 11 13 15 17 19 21 23 R LI MI C A
Time (Hour) Customer Classes
(e) (f)

Figure 5.3: Aggregated class-wise load patterns BDR and ADR of (a) R, (b) LI, (c)
MI, (d) C, (e) A, (f) Class wise curtailment and shifting
Chapter 5. Hierarchical Price Based Demand Response Framework 104

5.5.1.2 Effect of PBDR on Network Operation Conditions

The effect of DR on the system’s load factor and the active power loss is illustrated in
Fig. 5.4 (a)-(b). The figures display that load factor, and the active power loss are notably
improved during peak hours. Further, their reflections reappear on the smaller scale during
the valley and off-peak periods due to load shifting. The percentage increase and decrease
in the active power loss are found to be 19.63 and 19.28 during peak and off-peak-valley
hours combined relative to BDR state, respectively.

1.0 LF_BDR LF_ADR 200 PLoss_BDR PLoss_ADR

Active Power Loss (kW)

150
Load Factor

0.8
100
0.6
50

0.4 0
1 3 5 7 9 11 13 15 17 19 21 23 1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour) Time (Hour)
(a) (b)

Figure 5.4: Effect of DR on (a) System’ load factor and (b) Active power loss

5.5.1.3 Effect of PBDR on the Economic Prospects of the Different Entities

The effect of varying demand flexibility on the economic prospects of the participating
players is illustrated in Fig. 5.5 (a)-(c). It is described through the term demand response
flexibility (DRF). It is defined as the percentage ratio of the committed flexible demand
.
F
to the total available flexible demand expressed as DRF = (PiF P i ) × 100. Fig. 5.5
(a) displays variation in the LSE profit and the aggregated customer pay-off with an
increasing factor of DRF. It can be seen from the figure that LSE profit tends to decrease
as DRF increases, while at the same time, the aggregated customers’ pay-off increases. It
shows that as demand flexibility increases, demand requirement from the market reduces
counter-wise more demand is drawn during the valley and off-peak period on account of the
shifted load. This gives the overall reduction in LSE profit. It reveals that high flexibility
provision may lead to profit loss to LSE. A similar observation is also drawn for DRPs that
exhibit decline in their profit with increasing DRF, as shown in Fig. 5.5 (b). A similar
response is also observed in the case of the total customers’ billing, where bills exhibit
Chapter 5. Hierarchical Price Based Demand Response Framework 105

Agg. Customer pay off (¢)

40000

Agg. Customer pay off (¢)

LSE Agg. Cust pay off DRP Agg. Cust pay off
620000 800000 800000

DRP profit (¢)

LSE profit (¢)

600000 600000 600000

35000
580000 400000 400000

560000
200000 200000
30000
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
DRF (%) DRF (%)
(a) (b)

1600000 BDR ADR Change

Total Customer Bill (¢)

100000

Change in Bill (¢)

1550000

50000
1500000

1450000 0
0 10 20 30 40 50 60 70 80 90 100
DRF (%)
(c)

Figure 5.5: Effect of DRF variation on (a) LSE profit; (b) DRP profit, and (c) customers’
bills

downward movement as DRF increases, as shown in Fig. 5.5 (c). It can be confirmed
with a positive change in the billing ADR, which indicates that the participating in DR
accompanied by a higher flexibility margin can lower the customers’ bills.

5.5.1.4 Effect of PBDR on the Customer Bill and LSE Profit under Different
Pricing Strategies

In this subsection, the effect of the proposed pricing strategy is evaluated on LSE, DRP, and
the customers’ economic performances with the varying demand flexibility. It is compared
with the existing pricing approach using RTP as depicted in Fig. 5.6 (a)-(b). The figure
shows that LSE profit is relatively low under the RTP signal compared to the proposed
pricing. Further, the profit under both pricing strategies follows a declining pattern with
increasing DRF. Likewise, response is also perceived in the case of customer billing. It
indicates that the proposed pricing mechanism may result in a win-win situation from
both LSE and customers’ perspectives. It hedges customers with less flexible demands
against the high RTP price and LSE profit against RTP volatility [140].
Chapter 5. Hierarchical Price Based Demand Response Framework 106

700000 RTP Proposed Pricing RTP Proposed Pricing

Total Customer Bill (¢)

1550000
LSE profit (¢)

600000
1500000

500000 1450000

400000 1400000
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
DRF (%) DRF (%)
(a) (b)

Figure 5.6: Effect of DRF on the total billing and LSE profit under RTP and the
proposed pricing

5.5.1.5 Effect of PBDR on Economic Performance

Table 5.1 presents a comparative economic performance of the proposed framework under
the flat rate, RTP, and the proposed pricing strategy. The analysis is conducted for DRF
setting at 40 %. The results indicate that LSE’s profit and the total customers’ bills are
significantly lowered by 37.07 % and 6.63 %, with demand flexibility (DF) of 10.13 %
under RTP pricing. In contrast to this, the proposed pricing strategy exhibits a moderate
decrease of 15.00 % in LSE’s profit and a moderate increase of 3.34 % in the customers’
bills, respectively. The moderate changes under the proposed pricing indicates the benefits
of having different pricing for the flexible and inflexible loads. The total DR contribution
in terms of demand flexibility (DF ) is the same for both pricing approaches. These results
demonstrate that the proposed pricing strategy gives balanced results to all the involved
participants.

Table 5.1: Economic performance on the various pricing methods

Total Energy LSE Profit DRP Profit Total Bill

Pricing Method DF (%)
(kWh) (¢) (¢) (¢)
Flat Rate 68526.82 770455.35 - 1570017.90 -
RTP 68526.82 484811.72 36552.56 1465896.41 10.13
Proposed Pricing 68526.82 654869.81 36552.56 1517578.30 10.13

The aggregated class-wise customers’ bills are summarized in Table 5.2. It can be observed
that all customer class bills are reduced ADR. The bill reduction is highest in LI and lowest
in C class customers. It indicates that customer classes (i.e., LI, A) with a nominal ratio
between peak to valley/off-peak exhibit DR effectively compared to customers’ classes
(i.e., C, MI, R) with relatively higher ratios.
Chapter 5. Hierarchical Price Based Demand Response Framework 107

Table 5.2: Class-wise bills BDR and ADR using the proposed pricing strategy

Class R LI MI C A Total Bill

BDR (¢) 308373.74 698373.05 88053.69 363873.94 111341.65 1570016.07
ADR (¢) 299909.40 667218.67 85953.42 357556.82 106938.23 1517576.54
Change (¢) -8464.35 -31154.38 -2100.27 -6317.12 -4403.42 -52439.53
Change (%) -2.74 -4.46 -2.39 -1.74 -3.95 -3.34

5.5.2 108-bus Indian Distribution System

The proposed framework and algorithm are also examined on a real 108-bus Indian dis-
tribution system comprising five feeders in an urban area. The system and line data are
taken from Ref. [139]. It is rated at 11 kV with the active power of 12.132 MW and
reactive power of 9.099 MVAr. The detail of considered customers for each class/DRP is
summarized in Table B.2 in Appendix B. The implementation of the proposed framework
and algorithm in 108-bus distribution system result in similar characteristics to the test
system. It gives identical optimal pricing profiles to test the system for LSE and DRP,
when evaluating on the same bound limits of the prices.

5.5.2.1 Effect of PBDR on Class-wise Load Pattern

The class-wise load patterns also display a similar pattern to the test system with a higher
magnitude in the large 108-bus system. An aggregated class-wise load patterns is shown
in Fig. 5.7 (a)-(e). It demonstrates that customers classes’ shift their curtailed demand
to the valley and off-peak hours dynamically. The class-wise contributions in DR are also
illustrated in Fig. 5.7 (f).

5.5.2.2 Effect of PBDR on Network Operation Conditions

Fig. 5.8 (a)-(b) shows the system’s load factor and active power loss. Both attributes
reveal a notable improvement during peak hours. The percentage increase and decrease
in the active power loss were found to be 19.63 and 19.28 during peak and off-peak/valley
hours, respectively.

5.5.2.3 Effect of DR on Economic Performance

A comparative economic assessment of the proposed model under the different pricing
strategies is presented in Table 5.3. The results indicate that LSE’s profit is greatly affected
Chapter 5. Hierarchical Price Based Demand Response Framework 108

4000 BDR_R ADR_R 6000 BDR_LI ADR_LI

Load Demand (kW)

Load Demand (kW)

5500
3000
5000
2000 4500
4000
1000
3500
0 3000
1 3 5 7 9 11 13 15 17 19 21 23 1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour) Time (Hour)
(a) (b)
BDR_C ADR_C 2000 BDR_MI ADR_MI
Load Demand (kW)

Load Demand (kW)

1000
1500

1000
500
500

0 0
1 3 5 7 9 11 13 15 17 19 21 23 1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour) Time (Hour)
(c) (d)
Off-peak
Curtailment/Shifting (%)

BDR_A ADR_A Valley Peak

Load Demand (kW)

2000 10

1500

1000 5

500

0 0
1 3 5 7 9 11 13 15 17 19 21 23 R LI MI C A
Time (Hour) Customer Classes

(e) (f)

Figure 5.7: Aggregated class-wise load patterns BDR and ADR of (a) R, (b) LI, (c)
MI, (d) C, (e) A, (f) Class wise curtailment and shifting

under RTP, whereas the customers get much higher benefits in bills. The reduction in the
profit and bills are 35.62 % and 6.98 %, respectively. However, the proposed pricing
strategy gives both entities a moderate loss of 11.07 % and a moderate benefit of 3.14 %.
Thus, the proposed strategy offers a hedging option for both entities against the potential
loss or benefit.
The class-wise customers’ bills are described in Table 5.4. The results demonstrate that
all customer classes’ bills are reduced with DR participation. The decrease in the bills lies
in the range of (1-5) %. Further, the overall reduction in the bills was found to be 3.14 %.
Chapter 5. Hierarchical Price Based Demand Response Framework 109

1.0 LF_BDR LF_ADR 800 PLoss_BDR PLoss_ADR

Active Power Loss (kW)

600
Load Factor

0.8
400
0.6
200

0.4 0
1 3 5 7 9 11 13 15 17 19 21 23 1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour) Time (Hour)
(a) (b)

Figure 5.8: Effect of DR on (a) System’s load factor and (b) Active power loss

Table 5.3: Economic performance on the various pricing methods

DR
Total Energy LSE Costs DRP Cost Total Bill
Pricing Method
(kWh) (¢) (¢) (¢)
(%)
Flat Rate 230820.69 2473695.31 0.00 5148279.32 0
RTP 230820.69 1602999.43 109690.76 4789175.65 10.13
Proposed Pricing 230820.69 2199928.34 109690.45 4986375.32 10.13

Table 5.4: Class-wise bills BDR and ADR using the proposed pricing strategy

Class R LI MI C A Total Bill

BDR (¢) 1000152.33 2468861.10 723430.58 648298.92 307184.94 5147927.87
ADR (¢) 973030.34 2374153.51 706290.87 638340.83 294219.85 4986035.40
Change (¢) -27121.99 -94707.59 -17139.70 -9958.09 -12965.09 -161892.47
Change (%) -2.71 -3.84 -2.37 -1.54 -4.22 -3.14

5.5.3 Comparison of the Proposed Algorithm

In this subsection, the effectiveness of the proposed method is compared with the conven-
tional equivalent single-level approach for both systems. It is provided in Table 5.5. It can
be observed from the table that the proposed method produces equivalent results to the
conventional method for both systems. It gives the better LSE’s profit for the same level
of peak demand curtailment, showing its feasibility to obtain better optimal results.

Table 5.5: Comparison of methods on 33 and 108-bus systems

33-bus test system Indian 108-bus

Method
LSE Peak LSE Peak
Profit (¢) Curtailment Profit (¢) Curtailment
Single Level 619116.60 10.07 2087836.11 9.30
Proposed approach 654869.81 10.13 2199928.34 9.43
Chapter 5. Hierarchical Price Based Demand Response Framework 110

Moreover, the proposed algorithm caters to an appropriate approach for solving the prob-
lem without disclosing the customers’ information to all levels. It preserves their privacy.
Though, it is not the case with the conventional method, where all the information from
the lower level is passed to the upper level from the optimization perspective.

5.5.4 Computational Issues

The proposed model and algorithm are implemented in MATLAB ® toolbox YALMIP [141]
and IPOPT [142] as a solver on Window 10 based personal computer Intel(R) Core(TM)
i3-8130U CPU @ 2.20GHz, 4GB RAM. The convergence threshold values (ϵ and ε) of the
nested decomposition algorithm are set to be 0.01 each. The algorithm converges within
2-3 iterations for both considered systems. The computational burden of the proposed
algorithm compared to the traditional single-level approach for both systems is presented
in Table 5.6. It can be observed from the table that the computation burden of the
algorithms has less time gap for the small test system. However, in the case of the 108-bus
system, the computation burden becomes high with an increase in the problem size. The
proposed algorithm gives much better time due to the decomposition approach relative
to the traditional equivalent single-level approach. It lowers the impact of constraint
enumeration. Moreover, it is easy to implement and is computationally tractable.

Table 5.6: Computational comparison of the methods

System 33-bus test system Indian 108-bus

Single Proposed Single Proposed
Index/Approach
level Approach level Approach
# Variables in MP 30571 1656 92493 5256
# Variables in IMP 28555 86395
# Variables in ISP1 5640 17208
# Variables in ISP2 5640 17208
# Equality constraints in MP 1867 1632 5949 5232
# Inequality constraints in MP 65328 2424 197952 7800
# Equality constraints in IMP 235 717
# Inequality constraints in IMP 33960 103368
# Equality constraints in ISP1 235 717
# Inequality constraints in ISP1 11280 34416
# Equality constraints in ISP2 235 717
# Inequality constraints in ISP2 11515 35133
CPU Time (sec) 1299 792 7000 5125
Chapter 5. Hierarchical Price Based Demand Response Framework 111

5.6 Summary

A tri-level, two-stage, price-based DR framework composed of LSE, DRPs, and customers

is proposed in distribution systems. It is structured in a non-cooperative game-theoretic
framework using the Stackelberg game and is validated theoretically using SE conditions.
The proposed problem is represented in the comprehensive form considering multiple
DRPs, large number of customers, different customer classes, and practical ac network
constraints. A mix pricing based on the types of loads is designed. Numerical analyses are
performed on the 33-bus test and 108-bus practical system, and a decomposition-based
approach, a nested reformulation & decomposition (R&D) method, is employed as the
solution algorithm. The results demonstrate that the proposed model and the solution
method effectively enumerate the economic benefits among the participating actors with
minimum cooperation while maintaining the system’s operational security. The suggested
algorithm performs adequately on the small to large scale problems with a reasonable
computational time. Besides, it manifests a sophisticated DR mechanism through DRPs
with minimal sharing of the customers’ information at each level from the optimization
viewpoint. The comparative assessment of the mix-pricing with the existing RTP strategy
indicates balanced outputs for all participants with a non-dominating feature. It lowers
the risk of loss in the LSE profit by securing its margin through mix-pricing. Further, it
renders an optimistic attitude among the customers with a low flexibility margin/income
level, who face high dynamic price to a fraction of the total demand only.
Chapter 6

Bi-level Incentive Based Demand

Response Framework in
Distribution Systems

6.1 Introduction

Incentive-based demand response (IBDR) is considered promising prospects in the demand

response (DR). It offers incentives for shaving the load during peak periods to alleviate
sudden peak demand rise, wholesale market price volatility, generators monopolies etc.
IBDR programs have had been a wide range of subjects in the wholesale market [10, 114].
It also has been proliferating in distribution systems (DS) with the advent of smart grid
technologies in recent times [14]. This empowers the large to small retail customers for
IBDR expansion [61, 92]. These developments lead to the emergence of multiple resource
players via DR, which may influence economic and technical activities in DS [89]. Under
such circumstances, the behaviour of the DR participants can be viewed as a strategic
interaction.
The strategic interactions occurs when multiple players compete non-cooperatively in the
system. In the strategic interactions, individual’ action is also influenced by rival players’
actions [86]. It has been widely employed in the electricity market to model the strate-
gic behaviour of generators, suppliers, customers, demand response providers (DRPs),
etc [83–85]. It has also been expanded in DS in the form of retail competition. Moreover,
DR has also been materializing in the sizable amount to influence the system’ performance.
It has been part of many applications in DS described in the literature [38, 42–44, 94, 120].

113
Chapter 6. Bi-level Incentive Based Demand Response Framework 114

It envisions DR to improve the system’s techno-economic performances through demand

response providers (DRPs)/ demand response aggregators (DRAs)/ load aggregators (LA).
These DR enablers coordinate with the distribution system operator (DSO) or load-serving
entity (LSE) on behalf of the customers or large customers. Though, DRPs’ interactions
in the competitive environment have rarely been part of the described studies. It is
usually investigated for a single DRP [38]. In other instances, the strategic interactions
have been devised between LSE and DRPs [42, 94], not among the DRPs. Furthermore,
IBDR economic decision models in DS have usually been investigated on without power
flow [42, 44, 94] or linear ac power flow [38], or dc power flow [42].
This chapter presents an IBDR framework under the ambit of multiple DRPs in a com-
petitive environment to imitate a retail electricity market in DS. The multiple DRPs are
assumed to act as light asset resources and are interacting competitively to determine
incentive rates. The problem is formulated as a bi-level, nonlinear problem (NLP) con-
sidering AC network constraints. The LSE with DRPs operates at the upper-level (UL)
and DRPs as a large customer or an aggregator in the lower-level (LL). It is illustrated
using the multi-leader-multi-follower (MLMF) game. The strategic behaviour of DRPs is
exemplified in a game-theoretic framework using a generalized Stackelberg game (GSG).
The equivalent mathematical problem is modeled as the EPEC and is solved using the
diagonalization method [86, 87]. The proposed framework is investigated on the stan-
dard 33-bus test and a practical 108-bus Indian distribution system. The algorithm is
thoroughly tested on the small to large-scale problems and is compared to the existing
method.

6.2 Proposed Bi-level IBDR Framework

The proposed IBDR framework consists of LSE and DRPs in distribution systems (DS).
DSO operates and manages the network security using a security-constrained AC optimal
power flow (AC-OPF) while coordinating with both entities. In the proposed framework,
DRPs operate as the retail participant in DS and are assumed to be equivalent to DRPs’
role in the electricity market. It is formulated as a bi-level problem in relation to the
MLMF game [86, 90]. The structure of MLMF game is illustrated in Fig. 6.1.
In UL, multiple DRP as the leaders are strategically interacting to optimize the LSE’s
cost and DRPs as the followers are optimizing their objective at the LL. It is worth noting
that DRPs roles are different at the UL and LL. It serves as dual purpose. DRPs at the
UL indicate the DRP’s incentive cost, whereas DRPs at the LL represent the aggregated
customers/ large customers’ objective function. The problem formulation is described in
Chapter 6. Bi-level Incentive Based Demand Response Framework 115

Figure 6.1: Bi-level Structure of the Proposed IBDR Framework

the following section. The dual variables associated with each constraint are written after
each colon.

6.2.1 Upper-level Problem: LSE’s Operating Cost

The objective of LSE is to optimize its operating cost against the wholesale market price
along with distinctive incentive offers to the respective DPPs during peak periods while
satisfying ac network constraints. It is expressed as follows:
 
X  X X k,F X k,IN C k,F 
min ZLSE = ρRT P (PG,t + PLoss,t ) − ρFt R ( PL,n,t − Pt ) + ρt Pt
PG,t, ρIN
t
C  t 
t∈ΩT n∈ΩN k∈ΩK k∈ΩK

(6.1)

s.t. ρk,IN
t
C
≤ ρk,IN
t
C
≤ ρk,IN
t
C k
:µρ,t ,µkρ,t ∀t ∈ ΩT , k ∈ ΩK (6.2)

ρk,IN
t
C
≥ ρk,F
t
R
: µkρ,t (6.3)
IF F
PL,n,t = Pn,t + Pn,t (6.4)
QL,n,t = QIF F
n,t + Qn,t (6.5)
k,F
PG,n,t − PL,n,t + ϑkn Pn,t
P
= {en,t (gnm em,t − bnm fm,t )
m (6.6)
+fn,t (bnm em,t + gnm fm,t )} : λp,n,t , ∀n ∈ ΩN
Chapter 6. Bi-level Incentive Based Demand Response Framework 116

QG,n,t − QL,n,t + ϑkn Qk,F

P
n,t = {fn,t (gnm em,t − bnm fm,t )
m (6.7)
−en,t (bnm em,t + gnm fm,t )} : λq,n,t ,, ∀n ∈ ΩN
2
|V n |2 ≤ Vn,t
2
(en , fn ) ≤ V n : π v,n,t , π v,n,t ∀n ∈ ΩN (6.8)
2 2
Sbnm,t (e, f ) ≤ Sbnm,t,max :πs,nm,t ∀(n, m) ∈ ΩL (6.9)

P G,n,t ≤ PG,n,t ≤ P G,n,t : π p,n,t , π p,n,t (6.10)

QG,n,t ≤ QG,n,t ≤ QG,n,t : π q,n,t , π q,n,t (6.11)

Equation (6.1) stands for the LSE’s operating cost in conjunction with DRP’s incentive
cost. The overall LSE cost is defined as the summation of purchasing and incentive costs
against the transacted power’s revenue cost. Constraint (6.2) denotes the offered incentive
rates’ upper and lower limits. Constraint (6.3) defines incentive rate constraint. It is
set to greater than or equal to the proportionality factor of FR/retail rate for inducing
demand curtailment during the peak period [43]. Constraints (6.4)-(6.5) describe the
total composition of active and reactive demands. It is the combination of flexible and
inflexible loads. The active and reactive power flow constraints are stated in (6.6)-(6.7).
DRP’s status at the particular bus in DS is defined through a binary variable ϑ with
ϑkn = {0, 1}, k ∈ ΩK in (6.6)-(6.7). Bus voltage limits and power flow limits in the lines
in either direction are defined in the constraints (6.8)-(6.9). The active and reactive
generations limits are designated in the constraints (6.10)-(6.11).

6.2.2 Lower-level Problem: (DRP k) Model

In LL, individual DRP induces demand flexibility based on the received incentives rates
offers. It represents the cluster of the participated customers in DR to increase its market
demand [24]. Therefore, DRP’s function denotes the aggregated or large customer con-
sidering its behavioral aspects at the LL. In addition, a discrete set of customer classes
such as residential (R), large industrial (LI), medium industrial (MI), commercial (C),
and agricultural (A) are considered. It is assumed that a single DRP oversees a particular
customer class for IBDR illustration. This conceives a distinct incentives rate and selling
price for the considered customer classes. It is illustrated through a factor κk termed as
subsidy (negative value)/ overcharging (positive value) factor. This defines the following
relation.

k,F R/IN C F R/IN C

ρt = ρt × (1 + κk ) (6.12)
Chapter 6. Bi-level Incentive Based Demand Response Framework 117

The DRP’s objective is formulated as the combination of customers’ bill saving and the
disutility function. It is illustrated along with DRP constraints as follows:
X n k,F o
k
min ZDRP = (Pt,o − Ptk,F )ρk,F
t
R
− Pt
k,F k,IN C
ρt + U k (P k,F
t ) (6.13)
Ptk,F t∈ΩT
k,F
s.t. P k,F
t ≤ Ptk,F ≤ P t :ν kP F ,t ,ν kP F ,t , ∀t ∈ ΩT , k ∈ ΩK (6.14)
X k,F
Pt ≤ Pt,max : νPk F ; ∀k ∈ ΩK , ∀t (6.15)
k∈ΩK

Ptk,F = Π(ρk,IN
t
C
) (6.16)
F
ℓk P k,t  
Π(ρk,IN
t
C
)= × ℏk ρk,IN
t
C
− ρk,IN C
(6.17)
(ρk,IN
t
C
− ρk,IN
t
C
) t

Equation (6.13) consists of two terms, saving in the customer bills and the disutility
function. The first term is bill saving due to the demand reduction and incentive cost. The
second term accounts for the customers’ discomfort for impending demand curtailment.
Constraint (6.14) defines upper and lower flexible demand limits under DRP k. The
total demand contributed in DR from all the DRPs is expected to be kept below a fixed
demand. It is defined in (6.15), which maintains a proportion of demand supplied by
the large customers or DRPs. This indicates that a shared resource bounds all DRPs’
responses. It is called the common-coupling or the shared constraint [143]. The incentive
rate offers are assumed to be dynamic to increase DR attractiveness among the different
customer classes as defined in (6.16) [120]. This makes induced demand dependent upon
the incentive rate offers and the types of the customer class. A linearized function is
illustrated in (6.17) to show the dependency on the concerned parameters [120, 144]. The
disutility function of the aggregated customers is defined as follows:

1
Uk (Ptk,F ) = ßk (Ptk,F )2 + ς k Ptk,F , ßk > 0, ς k > 0 (6.18)
2

Equation (6.18) describes the aggregated customer’s dissatisfaction cost/disutility cost.

Its coefficient parameters ßk and ς k denote the customers’ attributes towards DR partic-
ipation. The function measures dissatisfaction on sacrificing or giving up the services or
utility partially or entirely.

6.2.3 Equilibrium Model

The proposed problem reveals that the strategic interaction of DRPs over incentive rates
determines the possible demand curtailment. This is equivalent to Bertrand or price
Chapter 6. Bi-level Incentive Based Demand Response Framework 118

competition. However, Bertrand’s competition does not consider the capacity constraints
on the products [145]. Hence, an extended model, named as the Bertrand-Edgeworth
model also called price-demand competition, is deliberated [146]. It imposes a capacity
constraint analogous to individual DRP’s demand limits at the LL, as defined in (6.14).
The price-demand competition is conceived through an interaction mechanism between
the price and demand using the Stackelberg game (SG). It is formulated as a MLMF
to model DRPs in the competitive environment. As in the proposed formulation, each
strategic decision variable is affected by the decisions making of other strategic variables
due to the common-coupling constraint or shared constraint (6.15). Hence, a coupling
effect is established amongst the decision-makers, which deduces a trade-off among DRPs.
It is illustrated through a GSG rather than SG, which defines the equilibrium conditions
in the shared constraint [86, 87, 147]. The equilibrium conditions are termed GSE [147].

6.2.3.1 Generalized Stackelberg Game

In this section, the proposed problem is formally presented in GSG. It is described in ab-
stract form using the standard game notation as Υ = {ΩK ∪G, {Sk }k∈ΩK , {Uk }k∈ΩK , u(x)},
where ΩK = {1, 2, ....., ΩK } is set of players, S is set of the strategy space of all the players.
It consists set of incentive rates ρIN C at the UL and flexible demand P F at the LL. U
t t
k
represents the utility sets, which describe ZLSE and ZDRP objectives, respectively. u(x)
defines the shared constraint among DRPs. The aspects of GSG game are described in
the following section

(a) In the UL, DRP’s incentive rates are offered as decision variables. It acts as a leader
and optimizes the overall LSE cost in conjunction with DRP cost through the set of
incentive prices ρk,IN
t
C
.

(b) Individual DRP as the aggregated customers at the LL acts as a follower. It opti-
mizes the customer’s bill and disutility cost subject to constraints (6.14)-(6.17). The
shared-constraint u(x) is defined in (6.15). It denotes the limits for possible DRP
participation in DS.
k
(c) The objectives ZLSE and ZDRP are a continuously differentiable in ρIN C and P F ,
t t
respectively. In addition, each objective is convex for every ρk,IN
t
C
and Ptk,F . In
addition, the strategy space is bounded.

Definition 1: Let’s suppose the incentive rates offered to the various DRPs are denoted as
∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗
ρIN C = {ρ1,IN C , ρk,IN C , ...., ρΩK ,IN C } at the UL and P F = {P 1,F , P k,F , .., P ΩK ,F }
t t t t t t t t
Chapter 6. Bi-level Incentive Based Demand Response Framework 119

is a set of strategies of induced demand by DRPs at the LL. Then, an optimal strategy
∗ ∗
set (ρIN C , P F ) constitutes GSE [86] if the following inequalities hold.
t t

∗ ∗ ∗ ∗
k
ZDRP (ρk,IN
t
C∗
, Ptk,F , P −k,F
t
k
) ≤ ZDRP (ρk,IN
t
C
, Ptk,F , P −k,F
t ) k ∈ ΩK (6.19)
k,IN C ∗ k,F ∗ −k,IN C ∗ k,F ∗ −k,IN C ∗
ZLSE (ρt ,Pt , ρt ) ≤ ZLSE (ρk,IN
t
C
,Pt , ρt ) (6.20)
k,F ∗
X
Pt ≤ Ptmax (6.21)
k∈ΩK

∗ ∗ ∗ ∗
where ρ−k,IN
t
C
= (ρk,IN
t
C
)k∈ΩK \{k} and P −k,F
t = (P k,F
t )k∈ΩK \{k} denote the set of
strategies i.e., offered incentive price and induced demand of all DRPs except kth DRP.
The inequalities above (6.19)-(6.20) define the conditions that the players will not have
incentives to deviate unilaterally from their equilibrium points. If the players do change
their optimality points, then GSE will not constitute.

6.2.3.2 Existence of GSE

The formulated non-cooperative game includes mainly two types of constraints. First is
an individual constraint, which depends upon its variable (i.e., ρk,IN
t
C
and Ptk,F ). The
second is a shared or joint constraint, which depends upon all variables, (i.e., P Ft ). These
individual constraints are usually satisfied through a pre-defined set of feasible spaces.
Hence, only a shared or coupled constraint is illustrated for the sake of simplicity [143].
The GSE is demonstrated with variational inequalities (VI) under a shared constraint.
It is a refinement over GSE, which ensures that the existence of GSE will lead to the
existence of a variational equilibrium (VE) [143, 148]. Though it is not always held in
general. Hence, GSE is obtained by solving VI, which ensures the solution of GSE [25].
It is denoted by V I(X, F), where X ≡ P k,F ⊆ Rk defines a set and F : Rk → Rk is a
point-to-point mapping into itself. It states that if x is a solution of VI, then it will also
be a solution of GSE, and GSE should satisfy the following condition:

T
F(x) (z − x) ≥ 0, ∀z ∈ X(x) (6.22)

∗ ∗
F , .., P F } is the solution of VI and (x) ≡ (∇ φ(x)T , .., ∇ T T
where, x ≡ {P1,t k,t F x1 xΩK φ(x) ) ,
k
x ∈ X denotes the partial derivative of each DRP function, i.e., φ(x) = ZDRP (Ptk,F , Pt−k,F ).
Theorem 1: For the offered incentive price ρIN
t
C , a GSE will exist for the proposed game

Υ between LSE and DRPs.

Proof: To find GSE, the first DRP problem is reformulated as unconstrained optimiza-
tion using the Lagrange multipliers [96]. The DRP function with the shared constraint is
Chapter 6. Bi-level Incentive Based Demand Response Framework 120

formulated as follows:
 
X X n k,F o
(Pt,o − Ptk,F )ρk,F R
− Ptk,F ρk,IN C
+ Uk (Ptk,F ) + λk  Ptk,F − Pt,max 
X
k
ZDRP = t t
k∈ΩK t∈ΩT k∈ΩK

(6.23)

Now, the optimality condition for DRP is obtained through Karush-Kuhn-Tucker (KKT).
 

(Ptk,F , Pt−k,F , ρk,IN C

Ptk,F − Pt,max  λk = 0
X
k
−∇P k,F ZDRP t ) + ∇P k,F  (6.24)
k∈ΩK
 

Ptk,F − Pt,max  ≤ 0
X
0 ≤ λk ⊥  (6.25)
k∈ΩK

where, λk is the Lagrange multiplier associated with each DRP. The KKT condition states
∗
that GSE exists for the follower game for a given fixed incentive rate, only if (Ptk,F , λk )
satisfies KKT conditions (6.24)-(6.25). Further, GSE solution can be obtained through
VI. Since, (6.21) is coupling-constraint and if constraint qualification holds for all DRPs,
then the Lagrange multiplier will be common for all DRPs i.e., (λ1 = λ2 = .. = λΩK = λ).
Then comparing (6.24)-(6.25) with KKT conditions of VI (see [143]) guarantee that P k,F
is the solution of V I(X, F ). This defines the existence of GSE condition.

6.2.3.3 Uniqueness of GSE

The uniqueness of GSE determines the stable social outcome of the game Υ. It is obtained
through the existence of one GSE. It is derived through VI and the coupling constraints.
Theorem 1 For the defined game Υ, only one GSE exists.
Proof: To verify the uniqueness of GSE, the first function F is obtained as follows:
 
−ρ1,F
t
R
− ρ1,IN
t
C
+ ß1 P1,t
F + ς1

 −ρ2,F R − ρ2,IN C + ß2 P F + ς 2
 

t t 2,t
F= (6.26)
 
 .
..


 
−ρk,F
t
R
− ρk,IN
t
C
+ ßk Ptk,F + ς k
Chapter 6. Bi-level Incentive Based Demand Response Framework 121

Equation (6.26) is the partial derivative of the DRP function with respect to Ptk,F . Now,
Jacobian of F gives as follows:
 
ß1 0 ··· 0
 
 0 ß2 · · · 0 
JF =  (6.27)
 
.. 
 0
 0 . 0 
0 0 · · · ßk

where, (6.27) is the Jacobian matrix of F. It is positive definite matrix with all positive
diagonal elements as ßk > 0. This indicates that F is a strongly monotonic function. It
∗ ∗
verifies the uniqueness of the solution (P Ft , F(P Ft )) of VI, which is also equivalent to a
unique solution to GSG. This concludes the proof of Theorem 1 [143].

6.3 Solution Approach

The MLMF problem is the augmentation of the multiple Stackelberg games. It competes
non-cooperatively to reach GSE. Mathematically, it is formulated as EPEC to find the
equilibrium points, which simultaneously solves the various MPECs problems [86]. To
illustrate the formulated bi-level problem in the equivalent formulation, the LL problem is
incorporated in the UL using MPEC, obtained through KKT conditions. The formulations
are compactly written, and their sub or superscript notations are kept minimum for the
sake of clarity. The UL and LL problems are denoted using the subscripts u and l,
respectively. The complete equivalent formulation of a bi-level problem is expressed as
follows:

∗
min ZLSE (xk , x, x−k , y) (6.28)
xk ,x

s.t.gu,P E (x, y) = 0 : λP (6.29)

gu,QE (x, y) = 0 : λQ (6.30)
gu,ρIN C (xk ) ≥ 0 : µIN C (6.31)
gu,I (x) ≥ 0 : π (6.32)
hl,E (xk , y, υ k ) = ∇y ZDRP (y) − υ k ∇y hl,I (y) = 0 (6.33)
0 ≤ υ k ⊥hl,I (y) ≤ 0 (6.34)
Chapter 6. Bi-level Incentive Based Demand Response Framework 122

where (6.33)-(6.34) are the stationarity conditions of MPECs of the LL. It is in fact a
mathematical program with the complementarity constraints (MPCC) due to the com-
plementarity constraints (6.34) [149]. The UL problem is described through (6.28)-(6.32).
The state and control variables of LSE such as PG,n,t , QG,n,t , en,t and fn,t are confined in x
in the UL and DRPs variable in the UL is denoted by xk = {ρk,IN
t
C
}. The individual DRP
decision variable at the LL problem is denoted by y = {Ptk,F }. The LSE’s active equality,
reactive equality, and inequality constraints are comprised into gu,P E (x, y), gu,QE (x, y),
and gu,I respectively in the UL. Further, their dual variables associated with the active
and reactive equality constraints are enclosed in λP = {λp,n,t } and λQ = {λq,n,t }, re-
spectively. LSE’s inequality constraints (6.8)-(6.11) are abbreviated in π. The inequality
constraints (6.14)-(6.15) of the LL are denoted by hl,I (y), and their dual variables by
ν = {νPk F ,t , ν kP F ,t , ν kP F ,t }. The nonlinear complementarity constraints (6.34) of the LL can
be further converted into the equality constraints by introducing a slack variable s. This
gives the DRP problem in the reformulated form as follows:

hl,E (xk , y, υ k ) = 0 : ϕkE (6.35)

hl,I (y) − s = 0:ϕkI (6.36)
s ≥ 0:ϕk1 (6.37)
υ k ≥ 0:ϕk2 (6.38)
s ◦ υ k ≤ 0:ϕk3 (6.39)
ϕkE , ϕkI free (6.40)

where, the constraints (6.37)-(6.39) are obtained by transforming the complementarity con-
straints 0 ≤ c⊥d ≥ 0 into nonlinear constraints in the form as c ≥ 0, d ≥ 0, c ◦ d ≤ 0, where
◦ is a notation of Hadamard product. Since, the formulated problem is a non-convex due
to nonlinear constraints and complementarity constraints. Hence, KKT conditions of the
problem (6.29)-(6.33) and (6.36)-(6.39) are the strong stationarity conditions rather than
the equilibrium conditions [83]. The concatenation of the strong stationarity conditions
gives the complete nonlinear complementarity formulation (NCP) as follows:

∇x ZLSE − λP ∇x gu,P E − λQ ∇x gu,QE − π∇x gu,I = 0 (6.41)

∇xk ZLSE − µIN C ∇xk gu,ρIN C − ϕE,d ∇xk hl,E = 0 (6.42)
∇y ZLSE − λP ∇y gu,P E − λQ ∇y gu,QE − ϕE,k ∇y hl,E − ϕkI ∇y hl,I = 0 (6.43)
−ϕkE ∇υk hl,E − ϕk2 + s ◦ ϕk3 = 0 (6.44)
−ϕkI − ϕk1 + υ k ◦ ϕk3 = 0 (6.45)
Chapter 6. Bi-level Incentive Based Demand Response Framework 123

gu,P E (x, y) = 0 (6.46)

gu,QE (x, y) = 0 (6.47)
hl,I (y) − s = 0 (6.48)
hl,E (xk , y, υ k ) = 0 (6.49)
0 ≤ gu,ρIN C (xk )⊥µIN C ≥ 0 (6.50)
0 ≤ gu,I (xk )⊥π ≥ 0 (6.51)
0 ≤ s⊥ϕkI ≥ 0 (6.52)
0 ≤ υ k ⊥ϕk2 ≥ 0 (6.53)
0 ≤ s ◦ υ k ⊥ϕk3 ≥ 0 (6.54)

The above formulation is a non-square NCP problem as the constraints (6.44), (6.45), and
(6.48) are not uniquely matched to free variables y, s, υ k and ϕl,I . In addition, variables,
s, and υ k are associated with multiple inequalities (6.52), (6.53), and (6.54). Therefore,
the complementarity constraints (6.52)-(6.54) are represented with the different sets of
inequalities in (6.55)-(6.57) [87].

0 ≤ s⊥ϕkI ≥ 0 ∀k (6.55)
0 ≤ υ k ⊥ϕk2 ≥ 0 ∀k (6.56)
0 ≤ s⊥υ k ≥ 0 ∀k (6.57)
ϕk3 ≥ 0 ∀k (6.58)

Further, the complementarity constraints (6.50)-(6.51) are written as equality constraints

by introducing the slack variables ω k and ϖk .

gu,ρIN C (xk ) − ω k = 0 (6.59)

0 ≤ µIN C ⊥ω k ≥ 0 (6.60)
gu,I (xk ) − ϖk = 0 (6.61)
0 ≤ π⊥ϖk ≥ 0 (6.62)

NCP formulation has an inherent redundancy problem due to large multipliers and com-
plementarity constraints. Hence, it is reformulated as a nonlinear problem by transforming
the complementarity constraints into nonlinear constraints [86]. Further, the complemen-
tarity constraints are minimized as the objective function using a penalty method. This
Chapter 6. Bi-level Incentive Based Demand Response Framework 124

gives the complete NLP formulation as follows:

X
minimize CP en = ϕk1 s+ϕk2 υ k + sυ k + ω k µIN C + ϖk π (6.63)
k∈ΩK

s.t. ∇x ZLSE − λP ∇x gu,P E − λQ ∇x gu,QE − π∇x gu,I = 0 ∀k ∈ ΩK (6.64)

∇xk ZLSE − µIN C ∇xk gu,ρIN C − ϕkE ∇xk hl,E = 0 ∀k ∈ ΩK (6.65)
∇y ZLSE − λP ∇y gu,P E − λQ ∇y gu,QE − ϕkE ∇y hl,E − ϕkI ∇y hl,I = 0 ∀k ∈ ΩK (6.66)
−ϕkE ∇υk hl,E − ϕk2 + s ◦ ϕk3 = 0 ∀k ∈ ΩK (6.67)
−ϕkI − ϕk1 + υ k ◦ ϕk3 = 0 ∀k ∈ ΩK (6.68)
gu,P E (x, y) = 0 (6.69)
gu,QE (x, y) = 0 (6.70)
hl,I (y) − s = 0 (6.71)
hl,E (xk , y, υ k ) = 0 (6.72)
gu,ρIN C (xk ) − ω k = 0 ∀k ∈ ΩK (6.73)
gu,I (xk ) − ϖk = 0 ∀k ∈ ΩK (6.74)
hl,I (y) − s = 0 (6.75)
−ϕkI − ϕk1 + ϕk3 ◦ υ k = 0 ∀k ∈ ΩK (6.76)
−ϕk2 + ϕk3 + s ◦ υ k = 0 ∀k ∈ ΩK (6.77)
µIN C ≥ 0, π ≥ 0, ω k ≥ 0, ϖk ≥ 0 (6.78)
s ≥ 0, υ k ≥ 0, ϕk1 ≥ 0, ϕk2 ≥ 0, ϕk3 ≥ 0 ∀k ∈ ΩK (6.79)

The reformulated problem (6.63)-(6.79) gives the complete NLP model. It excludes all
complementarity constraints from the problem constraints by moving them into the objec-
tive function as a penalty function. The feasibility of the problem is evaluated by optimiz-
ing the complementarity conditions to zero. The optimality is confirmed through the strong
∗
stationarity conditions of the problem. It states that if solution set (x∗ , x∗k , y ∗ , s∗ , υ k , ϕ∗l,E ,
∗ ∗ ∗ ∗ ∗ ∗
ϕ∗l,I , ϕk1 , ϕk2 , ϕk3 , ϖk , ω k , π ∗ , µIN C ) is local optimal with Cpen = 0 of the problem. Then
the solution set will be strong stationary points. These strong stationary conditions are
equivalent to KKT conditions of the NLP problem, as proved in Ref. [149]. The diagonal-
ization method is adopted to solve the formulated NLP, EPEC problem [84, 86]. It is an
iterative approach for repetitively solving a sequence of MPECs problems to reach equilib-
rium points. A nonlinear Gauss-Seidel based diagonalization approach is considered, and
its procedure is illustrated in the algorithm 6.1 as follows:
Chapter 6. Bi-level Incentive Based Demand Response Framework 125

Algorithm 6.1 Procedure of diagonalization method based on Gauss-Seidel ap-

proach
1) Step 0. Initialization: Assuming a starting point (xk,(0) , y (0) ) =
(x 1,(0) ,xk,(0) , ...., x ΩK ,(0) (0)
, y ) , iteration count j and convergence tolerance ε. .

2) Step 1. Loop over MPEC problems: Solving a non-linear MPEC prob-

lem for current iteration j (xk,(j) , y (j) ) , while keeping constant recently up-
dated information of decision variables of others DRPs i.e., (x−k,(j) , y (j) ) =
(x1,(j) , ..., xk−1,(j+1) , xk+1,(j) ...., xΩK ,(j) , y (0) ) . Denoting the current loop solution as
X k,(j) = (xk,(j) , y (j) ).
3) Step 2. Check convergence: if X k,(j) − X k,(j−1) ≤ ε, return (xk,(j) , y (j) ), “equilib-
rium is reached”, otherwise j = j + 1 and go back to step 1.

6.4 Simulation Results and Discussions

This section tests the proposed IBDR framework and the solution method on the 33-bus
test [139] and Indian 108-bus practical system [106]. The systems’ load and line data are
given in Table A.1 and Table A.2 of Appendix A. The considered systems are sectionalized
into five different customer classes. The details related to the customers’ class, such as
class-wise bus, demand, and maximum flexible demand, are outlined in Table B.1 and
Table B.2 in Appendix A. The techno-economic parameters related to power flow studies
are given in Table B.4 of Appendix B. The DR participation level is set to be varied within
the range of 0-30 %. The utility offered RTP prices are taken from the Ontario energy
board as shown in Fig. 6.2 [109]. It is scaled according to the LMP rates of PJM [43] to
investigate the impact of IBDR during the peak periods.
Further, the considered peak periods are (8:00-11:00) and (18:00-22:00), respectively, to il-
lustrate multi-periods analysis in IBDR. The class-wise offered incentive rates are assumed
to be varied within the interval of (ρk,IN C , ρk,IN C ) ∼
t t= (ρ
k,F R k,RT P
,ρ
t t) with the condition
of (ρkt ≥ ρk,F
t
R
). The dis-utility parameters are set according to Ref. [39]. The class-wise
retail/flat rates are set above the average rate of electricity for hedging against the price
volatility [14]. It is taken as 1.05 times the average of the offered RTP rate. The power
factor is assumed to be the same before DR (BDR) and (after DR) ADR states. Hence,
the reactive power will also change according to active power changes under the IBDR
application.
Chapter 6. Bi-level Incentive Based Demand Response Framework 126

80

Price (¢/kWh)
60

40

20

0
1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour)

Figure 6.2: RTP profile

6.4.1 Standard 33-bus Test System

It is rated at 12.66 kV with peak active and reactive power demand of 3.715 MW and 2.30
MVAr, respectively [106]. To assess DRPs’ performance in the test system, first, DRPs’
locations are determined. The considered locations are pre-established based on optimal
siting of embedded DGs and ESS [150]. For the DR analysis, it is taken as 4, 14, 24, 28,
and 30. The schedules of DRPs’ demand and incentive prices in the different peak periods
obtained from the optimization are summarized in Table 6.1. It can be seen from the
table that each customer class’ DRP effectively induces demand curtailment during the
peak periods based on incentive rate offers. It makes the incentive rate appear distinct
and has the different values. If this aspect (i.e., equation (6.12)) is not considered, then
the incentive rates will be in close proximity for all the customer classes with a discrete
magnitude in the different periods. The incentive rate offers for the respective DRPs under
the different time periods are appeared to close to retail rate/flat rate. The variation in
incentive rate is the highest during peak periods of 19th and 20th. Further, the demand
curtailment also varies under the different peak time periods. It exhibits higher demand
curtailment during the peak period of the 19th and 20th periods. It is on account of
the highest peak RTP price; this causes the LSE to offer a higher price for the further
reduction in the operating cost.
The impact of DRPs participation on active power, LSE cost, and DRP cost is illustrated
in Table 6.2. The results indicate that DR causes notable change in the active power
drawn by the grid. It is decreased by 12.74 % on average during peak periods. Similarly,
LSE also faces a decrease in the operating cost with reduced demand except 8, 10, 11, and
18 hours. The highest cost decrease is exhibited during the peak period of 19th and 20th.
Chapter 6. Bi-level Incentive Based Demand Response Framework 127

Table 6.1: Schedule of class-wise incentive rates and induced flexible demand in the
different peak periods

Index t8 t9 t10 t11 t18 t19 t20 t21 t21

ρ1,IN
t
C
(¢/kWh) 25.30 25.30 25.30 25.30 25.30 35.16 35.32 30.91 25.30
ρ2,IN
t
C
(¢/kWh) 31.63 32.03 31.63 31.63 31.63 47.39 47.88 42.38 33.41
ρ3,IN
t
C
(¢/kWh) 37.95 37.95 37.95 37.95 37.95 53.72 53.84 47.01 37.95
ρ4,IN
t
C
(¢/kWh) 63.25 63.25 63.25 63.25 63.25 92.90 94.50 83.25 65.72
ρ5,IN
t
C
(¢/kWh) 15.81 15.94 15.81 15.81 15.81 23.58 24.08 21.28 16.87
Pt1,F (kW) 72.86 86.30 73.29 67.97 87.52 212.59 223.64 128.34 59.06
Pt2,F (kW) 104.16 109.39 101.85 108.27 107.56 262.14 264.26 193.88 100.62
Pt3,F (kW) 19.04 19.22 18.74 20.21 19.66 42.31 45.04 32.15 6.77
Pt4,F (kW) 14.40 28.07 30.48 33.88 53.15 135.35 138.67 106.90 50.16
Pt5,F (kW) 53.22 20.28 16.91 13.55 30.19 89.59 126.91 109.29 65.75

In addition, DRP cost indicates granular variation under the different periods. The DRP
cost is also highest during 19th and 20th of peak periods.

Table 6.2: Results in the different peak periods BDR and ADR

BDR ADR
Index
PG (kW) LSE cost (¢) PG (kW) LSE cost (¢) DRP cost (¢)
t8 3320.27 85617.47 3031.84 86087.11 7612.02
t9 3175.00 114813.11 2892.76 112633.73 8515.50
t10 2970.84 75456.36 2713.08 76232.66 7981.91
t11 2995.21 60037.01 2734.48 62206.91 8267.59
t18 3725.84 22519.38 3396.34 29379.47 10201.22
t19 3675.14 232819.78 2860.84 214195.82 36858.24
t20 3901.03 247859.54 3014.00 227251.19 39137.18
t21 3594.67 189903.78 2957.22 178606.88 24920.04
t22 3168.32 116193.02 2852.75 113322.26 9518.01

6.4.1.1 An aggregated load pattern ADR and DR contribution

Fig. 6.3 (a) illustrates an aggregated load pattern BDR and ADR at the system level. It
is obtained by summing up class-wise demand. The figure reveals that the peak demand
is curtailed appreciably during the peak period. The demand curtailment is discrete and
exhibits the highest curtailment during 20th peak period. The class-wise DR contribution
Chapter 6. Bi-level Incentive Based Demand Response Framework 128

is depicted in Fig. 6.3 (b). The figure illustrates that each customer class discretely
responds in IBDR. The aggregated IBDR contribution is obtained to be 12.62 %.

15

DR contribution (%)
4000 P_BDR P_ADR
Load Demand (kW)

3000 10

2000
5
1000

0 0
1 3 5 7 9 11 13 15 17 19 21 23 R LI MI C A Agg
Time (Hour) Customer Classes
(a) (b)

Figure 6.3: (a) Load pattern before and after DR and (b) Class-wise demand contribu-
tion

6.4.1.2 Effect of IBDR on Network Operation Conditions

The effect of IBDR on the network operating conditions in connection with load factor,
active power loss, and voltage variation is illustrated in Fig. 6.4 (a)-(c). It can be seen from
Fig. 6.4 (a) load factor is notably reduced during the peak period. The active power losses
exhibit decrement during peak period as observed in Fig. 6.4 (b). The average decrease
in power loss is found to be around 15.77 %. The voltage variations on the different buses
indicate slight improvement on the highly loaded buses 18-19, as shown in Fig. 6.4 (c).

6.4.1.3 Effect of IBDR on Economic performance

The impact of the proposed IBDR model on the economic performance is summarized in
Table 6.3. It presents a comparative assessment of the system’s different attributes BDR
and ADR state over a day. The results indicate that the total energy consumption and
LSE cost are lowered by 5.41 % and 5.47 %, respectively. Further, the LSE purchasing cost
is decreased by 10.29 % ADR. Similarly, the total customer bills and active power loss are
also show noteworthy reduction. Though, high incentives rates also increase DRP costs.
But it is a much better alternative to pay an exorbitant price during the peak period.
The aggregated class-wise customers’ bills BDR and ADR are summarized in Table 6.4.
Chapter 6. Bi-level Incentive Based Demand Response Framework 129

1.0 LF_BDR LF_ADR 200 PLoss_BDR PLoss_ADR

Active Power Loss (kW)

150
Load Factor

0.8
100
0.6
50

0.4 0
1 3 5 7 9 11 13 15 17 19 21 23 1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour) Time (Hour)
(a) (b)

V_BDR V_BDR
Mean voltage (p.u.)

1.00

0.98

0.96

0.94
1 4 7 10 13 16 19 22 25 28 31
Buses
(c)

Figure 6.4: Effect of DR on (a) system’s load factor and (b) Active power loss (c)
Voltage variation

Table 6.3: Optimization results of the system for a day

Index BDR ADR Variation Variation (%)

Total Energy (kWh) 68508.00 64804.44 3703.56 5.41
Average Peak Demand (kW) 3261.29 2849.78 411.51 12.62
Flexible Energy (kWh) 7293.40 3703.56 3589.84 49.22
LSE cost (¢/day) 828793.24 783489.83 45303.40 5.47
LSE purchasing cost (¢/day) 2995387.41 2687241.52 308145.89 10.29
DRP payment (¢) - 153011.69 - -
Customers Bills (¢) 2177261.53 1901730.19 275531.35 12.65
Disutility cost (¢) - 71976.54 - -
Total Power loss (kWh) 2345.72 1975.90 369.82 15.77

The results exhibit decrement in bills for all the customer classes ADR. The bill reduction
is highest in MI and lowest in A class customer. Further, the overall decrease in the bills
is around 12.65 %.
Chapter 6. Bi-level Incentive Based Demand Response Framework 130

Table 6.4: Class-wise bills BDR and ADR using the proposed pricing strategy.

Class R LI MI C A Total Bill

BDR (¢) 428909.48 968311.76 121461.53 504153.27 154425.49 2177261.53
ADR (¢) 372666.66 872051.82 102851.04 418776.00 135384.67 1901730.19
Change (¢) -56242.82 -96259.93 -18610.49 -85377.27 -19040.82 -275531.35
Change (%) -13.11 -9.94 -15.32 -16.93 -12.33 -12.65

6.4.2 Indian 108-bus system

The Indian 108-bus system is rated at 11 kV with an active demand of 12.132 MW and
reactive power of 9.099 MVAr. The system and line data are taken from Ref. [139]. The
DRPs’ locations are predetermined based on the optimal allocation of DGs and ESS [139].
DRPs locations for the different customer classes are as follows: R = {21, 24, 30}, LI =
{60, 63, 67} , MI = {85, 88} , C = {45, 49} and A = {97, 102}. The incentive rates and
DRPs’ induced demands for the different peak periods are summarized in Table 6.5 and
Table 6.6. The results indicate that the customer classes effectively exhibit the load shaving
based on the offered incentive rates. It demonstrates the higher incentive rates and higher
demand curtailment as peak periods prices rises. In addition, DRPs of the same customer
class exhibit discrete incentive rates and induced demands during the highest peak price
periods. The maximum variations in the incentive rate and induced demand are observed
during the peak periods of 19th and 20th.
Table 6.5: Schedules of class-wise incentive rates during the different peak periods

Index t8 t9 t10 t11 t18 t19 t20 t21 t21

25.30 25.30 25.30 25.30 25.30 35.75 35.80 31.12 25.30
ρ1,IN
t
C
(¢/kWh) 25.30 25.30 25.30 25.30 25.30 36.01 36.07 31.38 25.30
25.30 25.41 25.30 25.30 25.30 36.94 37.02 32.18 25.30
31.63 31.63 31.63 31.63 31.63 44.96 45.02 39.19 31.63
ρ2,IN
t
C
(¢/kWh) 31.63 31.63 31.63 31.63 31.63 45.28 45.35 39.49 31.63
31.63 35.23 31.63 31.63 31.63 49.17 49.29 43.06 33.98
37.95 37.95 37.95 37.95 37.95 48.54 54.61 47.57 37.95
ρ3,IN
t
C
(¢/kWh)
37.95 37.95 37.95 37.95 37.95 53.76 48.74 46.91 37.95
63.25 63.98 63.25 63.25 63.25 76.63 77.26 78.00 63.25
ρ4,IN
t
C
(¢/kWh)
63.25 63.25 63.25 63.25 63.25 76.63 77.26 77.19 63.25
15.81 15.81 15.81 15.81 15.81 22.69 20.36 19.30 16.79
ρ5,IN
t
C
(¢/kWh)
15.81 15.81 15.81 15.81 15.81 23.63 20.36 19.30 18.33

The variations in the power transacted from the grid, LSE cost, and DRP cost in the
different peak periods are presented in Table 6.7. The table indicates that the power
Chapter 6. Bi-level Incentive Based Demand Response Framework 131

Table 6.6: Schedules of induced demand during the different peak periods

Index t8 t9 t10 t11 t18 t19 t20 t21 t21

75.11 87.20 76.66 70.21 90.04 226.77 236.43 134.46 60.04
Pt1,F (kW) 75.11 87.20 76.66 70.21 90.04 230.15 239.97 137.08 60.04
75.11 88.40 76.66 70.21 90.04 241.82 252.40 145.09 60.04
124.68 125.14 123.16 130.02 130.37 280.14 284.93 198.19 101.82
Pt2,F (kW) 124.68 125.14 123.16 130.02 130.37 284.00 289.01 201.62 101.82
124.68 172.74 123.16 130.02 130.37 331.85 338.07 243.15 127.07
66.82 66.52 67.25 69.24 70.38 123.41 156.44 110.77 24.07
Pt3,F (kW)
66.82 66.52 67.25 69.24 70.38 152.71 123.71 107.29 24.07
11.97 23.89 25.82 27.64 43.98 75.93 76.30 76.75 36.80
Pt4,F (kW)
11.97 23.01 25.82 27.64 43.98 75.93 76.30 74.90 36.80
81.57 31.37 25.52 20.80 47.12 126.93 136.71 132.44 99.23
Pt5,F (kW)
81.57 31.37 25.52 20.80 47.12 137.20 136.71 132.44 125.93

drawn from the grid is appreciably reduced after the IBDR application. It is decreased by
11.92 % on average during the peak periods. Likewise, the overall LSE cost also appears
to reducing during peak periods except 8, 10, 11, and 18 hours.

Table 6.7: Results during the different peak periods BDR and ADR

BDR ADR
Index
PG (kW) LSE cost (¢) PG (kW) LSE cost (¢) DRP cost (¢)
t8 11644.44 303721.38 10628.65 304837.06 26695.45
t9 11106.30 406465.05 10092.02 397986.15 29684.59
t10 10586.84 273734.03 9680.60 275728.39 26681.73
t11 10634.54 218761.98 9727.11 225506.97 27073.82
t18 12508.24 78119.60 11423.69 100893.27 31598.79
t19 12104.37 767806.99 9604.46 709332.57 99060.96
t20 12738.53 809336.48 10153.60 746325.88 100994.04
t21 11603.45 612061.22 9741.12 578492.35 66535.41
t22 9744.86 356938.75 8791.69 348555.22 25771.44

6.4.2.1 An aggregated Load Pattern ADR and DR Contribution

The impact of IBDR on the system’s load profile is illustrated in Fig. 6.5 (a). It rep-
resents the summation of the class-wise demand at the system level. The figure reveals
a noteworthy reduction in the demand during peak periods. It is highest at 20th peak
period. The class-wise DRPs participation is also illustrated in Fig. 6.5 (b). The figure
Chapter 6. Bi-level Incentive Based Demand Response Framework 132

demonstrates that all customer classes discretely respond in IBDR. In addition, the overall
DR contribution is obtained to be 11.91 %.

15

Curtailment/Shifting (%)
12000 WDR ADR
Load Demand (kW)

10000 10

8000
5
6000
0
1 3 5 7 9 11 13 15 17 19 21 23 R LI MI C A Agg
Time (Hour) Customer Classes
(a) (b)

Figure 6.5: (a) Load pattern after DR and (b) Class-wise demand contribution

6.4.2.2 Effect of IBDR on Network Operation Conditions

The impact of DRPs on the technical parameters such as load factor, active power loss,
and voltage variation is exemplified in Fig. 6.6 (a)-(c). Fig. 6.6 (a) indicates that the
system’s load factor reduces during the peak period. Similarly, the active power losses are
also reduced, as shown in Fig. 6.6 (b). The average reduction in the active power losses
is obtained to be 10.87 %. The voltage variations after DR integration reveal a moderate
improvement on the load buses, as depicted in Fig. 6.6 (c).

6.4.2.3 Effect of IBDR on Economic Performance

The effect of the proposed IBDR decision model on the techno-economic performances
is described in Table 6.8. The results present the variations in the system’s different
parameters ADR relative to BDR status over a day. The table indicates that the total
energy consumption and the LSE operating cost are reduced by 5.07 % and 4.97 %,
respectively. This consequently lowers the LSE transaction cost from the grid by 9.46 %
and active power losses by 10.87 % ADR.
In addition, the class-wise bills BDR and ADR are presented in Table 6.9. The results
indicate the bill reduction for all the customer classes ADR. The overall bill reduction is
11.13 %.
Chapter 6. Bi-level Incentive Based Demand Response Framework 133

1.0 LF_BDR LF_ADR 800 PLoss_BDR PLoss_ADR

Active Power Loss (kW)

600
Load Factor

0.8
400
0.6
200

0.4 0
1 3 5 7 9 11 13 15 17 19 21 23 1 3 5 7 9 11 13 15 17 19 21 23
Time (Hour) Time (Hour)
(a) (b)
BDR_V ADR_V
Mean voltage (p.u.)

1.00

0.98

0.96

0.94
1 11 21 31 41 51 61 71 81 91 101
Buses
(c)

Figure 6.6: Effect of DR on (a) System’s load factor and (b) Active power loss (c)
Voltage variation

Table 6.8: Optimization results of the system for a day

Index BDR ADR Variation Variation (%)

Total Energy (kWh) 230804.96 219113.79 11691.17 5.07
Average Peak Demand (kW) 10899.35 9600.33 1299.02 11.92
Flexible Energy (kWh) 24178.70 11691.17 12487.53 51.65
LSE cost (¢/day) 2802835.81 2663548.18 139287.64 4.97
LSE purchasing cost (¢/day) 10102138.78 9146211.06 955927.73 9.46
DRP payment (¢) - 434096.23 - -
Customer Bills (¢) 7143196.95 6347862.61 795334.35 11.13
Discomfort cost (¢) - 140393.19 - -
Total Power loss (kWh) 10469.39 9331.78 1137.61 10.87

Table 6.9: Class-wise bills BDR and ADR

Class R LI MI C A Total Bill

BDR (¢) 1387798.21 3425759.17 1003822.74 899571.05 426245.78 7143196.95
ADR (¢) 1206408.00 3092642.86 880060.98 792583.77 376166.99 6347862.61
Change (¢) -181390.22 -333116.31 -123761.76 -106987.28 -50078.79 -795334.35
Change (%) -13.07 -9.72 -12.33 -11.89 -11.75 -11.13
Chapter 6. Bi-level Incentive Based Demand Response Framework 134

6.4.3 Comparison of the methods

A comparison of the solution methods is performed on both systems in this subsection.

The proposed EPEC model using the diagonalization method is compared with the con-
ventional MPEC approach via the NLP method. The sample results of class-wise in-
centive rate offers and induced demand curtailment of 20th hour for 33-bus systems are
presented in Table 6.10. The results indicate that the proposed method generates the
schedule equivalent to the conventional MPEC approach. This illustrates the potency of
the diagonalization method to achieve a strong equilibrium point under a non-cooperative
environment.
Table 6.10: Incentive rates and induced demand at 20th hour on 33-bus test system

Method NLP Diagonalization

Class ρk,IN
t
C
Ptk,F ρk,IN
t
C
Ptk,F
R 34.96 210.07 35.16 212.59
LI 47.16 259.66 47.39 262.14
MI 53.28 41.60 53.72 42.31
C 91.62 131.74 92.90 135.35
A 23.17 86.61 23.58 89.59

In addition, the economic comparisons of the results using both approaches are summa-
rized in Table 6.11. The results manifest that both methods yield comparable results on
both considered systems.
Table 6.11: Economic comparison for a day using both methods

System Standard 33-bus Indian 108-bus

Index NLP Diagonalization NLP Diagonalization
Total Energy (kWh) 64786.57 64804.44 219219.73 219113.79
Flexible Energy (kWh) 3693.86 3703.56 11585.23 11691.17
DR Contribution (%) 12.59 12.62 11.81 11.92
LSE cost (¢/day) 784834.93 783489.83 2664891.75 2663548.18
DRP payment (¢) 151733.26 153011.69 427564.19 434096.23
Customer Bills (¢) 1902285.87 1901730.19 6357616.98 6347862.61
Total loss (kWh) 1989.17 1975.90 9355.48 9331.78

Most importantly, the utilized method provides an appropriate way to solve the problem
in a non-cooperative environment. It maintains each participant’s privacy through a min-
imum sharing of information. This is not possible using the conventional approach, where
all the participants’ information is known beforehand. Furthermore, the diagonalization
Chapter 6. Bi-level Incentive Based Demand Response Framework 135

method extends a classical optimization-based free approach such as the gradient method.
This renders a conceptual simplicity and ease in solving the problem [10], [12].

6.4.4 Computational Issues

The proposed EPEC model and methods are implemented using MATLAB ® toolbox
YALMIP [141] and IPOPT solver [142] on Window 10 based personal computer Intel(R)
Core(TM) i3-8130U CPU @ 2.20GHz, 4GB RAM. The convergence tolerance value for the
diagonalization method is set to 0.01. It converges within 2-3 iterations for the consid-
ered system. The computational performances for the considered test systems via both
approaches are presented in Table 6.12. The result indicates that the computational costs
of the algorithms have a notable time difference for the system under both methods.
The diagonalization method produces a smaller size per problem, which lowers constraint
enumeration per iteration. Further, it is simple to solve the problem. Though the di-
agonalization method solves the problem faster, its cycling process may take longer to
converge under certain circumstances [90].

Table 6.12: Computational comparison of the methods

System Standard 33-bus Indian 108-bus

Method NLP Diagonalization NLP Diagonalization
# Variables in UL 657 621 2043 1971
# Variables in LL 189 45 333 45
# Equality constraints in UL 612 612 1962 1962
# Inequality constraints in UL 1035 918 3168 2952
# Equality constraints in LL 45 9 81 9
# Inequality constraints in LL 270 54 486 54
CPU Time (sec) 78 14 988 489

6.5 Summary

A bi-level IBDR decision framework considering multiple DRPs is presented in DS. It is

formulated in the competitive environment, where multiple DRPs compete over incentive
rates to improve the system’s economic and technical performance in UL and is constrained
by individual DRP’s objectives at LL. It is modeled in a game-theoretic framework using a
GSG owing to a MLMF game. It is mathematically formulated as an NLP, EPEC problem,
and is solved using the diagonalization method. The simulation results are performed on
Chapter 6. Bi-level Incentive Based Demand Response Framework 136

the standard 33-bus test and Indian 108-bus system. The results illustrate the effectiveness
of the proposed competitive framework for enumerating the potential benefits of each
participating players. It exhibits the competitive behaviour of DRPs while preserving
privacy via minimum cooperation. In addition, the solution algorithm runs adequately
on both systems with a reasonable computation time. In addition, the proposed model
exclusively evinces a retail competition by means of the available participants within DS.
Chapter 7

Conclusions and Future Scope

The distribution system (DS) is a connecting medium between the electrical generation
utility to the customers’ ends in the electrical power hierarchy. It facilitates the distribu-
tion of electrical power to the customers in a secure, efficient, and reliable way. However,
the roles of distribution utility and customers have been transformed significantly with
deregulation, retail competition, and information communication technology (ICT). This
has immersed the contemporary DS with renewable distribution generations (RDGs), en-
ergy storage systems (ESSs), electric vehicles (EVs), prosumers, retailers, demand response
(DR), etc. These technological innovations envisage better-controlled power support, intel-
ligent energy utilization, demand flexibility, and active customers in the system. However,
its derivatives, such as load growth, price volatility, coordination issues, wide-scale imple-
mentation, policies up-gradation, capital investment, etc., are also emerged as challenges
in DS. It has increased DR’s importance as one of the potential assets to address the
sudden peak demand rise, price volatility, efficient energy utilization, and so on.
DR induces load adjustment or peak shaving using dynamic pricing or financial incen-
tives. This subsequently reflects the changes in the system’s techno-economic perfor-
mances. However, DR modeling and assessment in DS have usually been carried out in
line with the electricity market. It measures an aggregated DR response but discard the
realistic features relevant to DS. Moreover, less emphasis is given to the customers’ rea-
sonable behavioral attributes in DR modeling. This indicates inadequacy for realizing
the feasible and rational behavioral models in DR. Besides, increasing DR integration has
brought significant changes in operational problem of DS under the current paradigms.
This makes decision making problems more complex under the ambit of DR.
This thesis presents the DR modeling approaches and the operational strategies in the
contemporary DS. It proposes a new/ improved modeling DR methods based on price
137
Chapter 7. Conclusions and Future Scope 138

elasticity model (PEM) and utility functions (UFs). The PEM-based price-based DR
(PBDR) model proposes an adaptive elasticity model to imitate the customers’ behav-
ior in DR. The idea of adaptive elasticity is further extended in the incentive-based DR
(IBDR) model. Moreover, the shortcoming of the standard IBDR model is addressed by
accounting the exact effect of incentive rate on elasticity. It is taken care by proposing an
analytical approach to extract the incentive elasticity in line with price elasticity. The UFs,
cardinal UFs (CUFs), and ordinal UF (OUF) are employed in PBDR modeling. Both UFs
incorporate the customer’s willingness and risk behaviour in DR models. The adequacy
and efficacy of the proposed DR models are thoroughly investigated on the different class-
wise load patterns synthesized using disaggregated load modeling on the standard 33-bus
test distribution system’s load bus data. Furthermore, economic DR operational strate-
gies have been developed under the emerging DR stakeholders in DS, and their strategic
interactions are deduced through game theory. Additionally, the solution methodologies
have been developed using the improved and customized optimization approaches. The
proposed DR operational strategies and the solution approaches are thoroughly investi-
gated on the standard 33-bus test and Indian-108 bus systems. The thesis is organized
into the seven chapters.
The introduction of DR, its importance, and forthcoming issues in the contemporary DS
are described in chapter 1. Chapter 2 presents a comprehensive literature review on the
DR modeling and its operational framework/strategies in DS. Based on the critical re-
view, a few research gaps are identified and formulated as the research objectives in the
presented work. The critical review covers the DR’s perspectives on the modeling and
strategies. It is observed from the literature that the existing DR modeling methods
are mostly adapted from the consumer theory, microeconomics. However, their applica-
tions in DR are not suitably molded to represent the customers’ behavioral attributes in
multi-period analysis. In addition, the modeling and assessment of DR approaches lack
transition from the wholesale electricity market to the distribution level. Most impor-
tantly, DR involvement in the decision-making problems has been investigated passively
in DS. It needs the systematic development of the operational DR strategies to meet the
evolving paradigms in DS. Further, the sophisticated and suitable modifications in the
optimization methods are required to address the complex DR problem considering the
practical networks constraints.
Chapters 3 deals with the DR modeling using PEM. It has been proposed for PBDR
and IBDR programs. The models present an adaptive PEM for PBDR using a dynamic
elasticity to mimic the customers’ behaviour in DR. The dynamic elasticity is illustrated
using the deterministic and stochastic approaches. The deterministic method is derived
Chapter 7. Conclusions and Future Scope 139

for inter-relating peak hour elasticity to cross-period elasticity. The stochastic approach is
modeled using the stochastic process via geometric Brownian motion. An improved IBDR
model using the PEM is proposed to account for incentive/penalty inclusion in the elastic-
ity variation. It proposes an incentive elasticity equivalent to price elasticity to extract the
exact effect of incentives inclusion. The proposed DR models are applied on the class-wise
load patterns using the disaggregated load modeling on 33-bus test system’s load data.
Further, the developed DR models are thoroughly investigated and assessed on the various
PBDR and IBDR programs. The following conclusions are drawn from this chapter:

1. The proposed PBDR models exhibit the rational DR behaviour for imitating the
customers’ attributes. The deterministic method encapsulates the features of both
elasticities in a dynamic elasticity for load adjustment. It provides a lossless DR but
partially obeys intertemporal load shifting.

2. The stochastic method is observed to be very effective for modeling intertemporal

constraint or load recovery through cross-elasticity. It envisages the customers’ di-
verse rationality using the stochastic process. The application results indicate that
the deterministic method gives the optimized load patterns, whereas the stochastic
method gives a realistic load behaviour in DR.

3. The class-wise economic assessment under DR reveals that shifting from flat rate
to dynamic pricing most likely increases the customers’ bills with zero participation
under NDR state. However, these bills are significantly lowered with participation
under ADR. In addition, the increase/decrease in the class-wise bills are varying in
the scale. It shows DR dependency upon the load and price patterns in addition to
their magnitude levels.

4. The proposed augmented IBDR modeling approach gives a precise valuation of incen-
tive inclusion in PEM. It produces 25-60 % higher demand curtailment in comparison
to the standard IBDR model under the different pricing.

5. The comparison and assessment of the different PBDR programs in the augmented
IBDR indicates that the relevance of incentive rate/elasticity tends to decrease as the
dynamism in the pricing increases, when DR contribution is measured at the same
level. It is highest for static pricing (i.e., Flat rate), average for moderate pricing
(i.e., TOU and MFR), and lowest for high dynamic pricing (i.e., RTP, VPP)

6. The economic assessment of aggregated bills under the different DR programs in-
dicates homogeneous variations. It decreases ADR relative to NDR state for all
Chapter 7. Conclusions and Future Scope 140

DR programs. Moreover, the analyses show that the customers with the nominal
demand ratio (peak/valley-off-peak) exhibit DR with much ease in comparison to
high demand ratios. Moreover, the class-wise load patterns exhibit discrete demand
flexibility under the different DR programs.

7. The utility keeps its profit secure under the different DR programs. In addition, the
PBDR and the augmented IBDR model under higher dynamic pricing appropriately
reduce the price volatility due to a notable reduction in the demand requirement
during the peak periods.

The PBDR models using CUF and OUF are developed in chapter 4. The various CUFs
such as quadratic, exponential, logarithmic, and power functions are modeled in DR. In
addition, the impact of different CUFs on DR is evaluated under different load patterns,
parameters, price variations and risk-based indices. Another UF, OUF is also applied in
PBDR modeling using a microeconomic behavior model known as Overlapping generation
(OLG). It models OLG’s parameters, the relative interest rate and degree of willingness
equivalent to the relative price ratios of the different segment periods and degree of risk
behaviour, respectively. Moreover, the customer’s altruistic behaviour is also incorporated
to imitate load recovery in DR. The application results are investigated on 33-bus test
system’s load data considering class-wise load patterns and compared with the existing
approaches. The following conclusions are drawn from this chapter:

1. The CUFs are an effective approach to model the customer’s behaviour in DR. DR
assessment of the different functional forms of CUFs indicates that quadratic and
exponential CUFs are highly sensitive functions, whereas logarithmic and power
CUFs are less sensitive functions.

2. The customers’ degree of risk behaviour estimated using risk-based indices directly
links with DR participation (i.e., load curtailment) for almost all CUFs except loga-
rithmic CUFs. It shows that logarithmic CUF suffices to exhibit DR behaviour but
does not represent the customers’ viewpoint adequately in DR.

3. The DR participation levels of the diversified customers illustrate that the customers
with a high value of absolute risk-aversion (ARA) exhibit the less participation in
DR or vice-versa, irrespective of their demand levels.

4. The proposed OLG based PBDR model effectively exhibits the load adjustment with
the variation in price. It includes the customers’ attributes in DR modeling in the
form of willingness and altruistic behaviour. Moreover, it produces substantial DF
Chapter 7. Conclusions and Future Scope 141

for all types of the customers’ load patterns relative to the existing PBDR models.
On the other hand, DF in the two-state OLG model is affected on the different load
patterns. This indicates that the proposed OLG model exhibits suitable DF as its
state level increases.

5. The OLG model requires few parameters to estimate as opposed to PEM and CUF,
which are simple in modeling. But their parameters’ estimation is a very tedious
and rigorous.

6. The proposed OLG model emphasizes on the customers’ feasible comprehension of

load adjustment under the dynamic prices. Though, PEM and CES-based PBDR
models closely illustrate economic DR but ignore feasible attributes in DR realiza-
tion.

Chapter 5 presents an optimal operation of a hierarchical PBDR framework comprised

of load serving entity (LSE), DRPs and DR customers in DS. It is formulated as a tri-
level, two-stage, nonlinear optimization problem with the nested bi-level form in each
stage. It includes LSE and DRP in the first stage; DRP and the customers in the second
stage. The interaction in each stage is described using a sequential Stackelberg game
(SG). The proposed PBDR framework is solved using a reformulation & decomposition
(R&D) method. The applicability of the formulated problem and the solution approach
are thoroughly investigated and presented on the standard 33-bus test and the Indian-108
bus system. Moreover, the application results are compared with the existing optimization
approach. The following conclusions are drawn from the chapter:

1. The proposed hierarchical PBDR problem demonstrates a comprehensive framework

consisting of multiple DRPs, large number of customers, and the different customer
classes. It provides the LSE’s techno-economic benefits and effectively remunerates
the obtained economic benefits among its shareholders with minimum cooperation.

2. The mix-pricing strategy is seen as adequate for securing the LSE’s economic benefits
and maintaining the customers’ bills with a low flexibility margin/ income level.

3. The demand response flexibility (DRF) index reveals that LSE’s profit and DRP’s
profit tend to decrease as DRF increases. It exemplifies that DR participation level
is appropriate up to an extent. However, it starts to affect the LSE and DRP’s
economic benefits adversely above the certain point.

4. The proposed methodology offers a simple and explicit problem structure by decom-
posing the large problem into the sub-problems. Moreover, it provides information
Chapter 7. Conclusions and Future Scope 142

transmission security with minimal sharing compared to the conventional optimiza-

tion approach.

Chapter 6 presents an optimal operation of the IBDR framework consisting of multiple

DRPs under the competitive retail environment in DS. It is formulated as a bi-level prob-
lem, where the LSE optimizes its operating cost by offering incentives at the upper level
(UL), and individual DRP optimizes its bill saving and discomfort cost at the lower level
(LL). The proposed problem is represented in the form of a multi-leader-multi-follower
game. The game is elucidated using the generalized Stackelberg game (GSG) and mathe-
matically via the equilibrium problems with equilibrium constraints (EPEC). Thereafter,
the formulated equivalent problem is solved using the diagonalization method. The effec-
tiveness of the proposed IBDR framework is investigated and presented on the standard
33-bus test and Indian-108 bus systems. The following conclusions are drawn in this
chapter:

1. The proposed bi-level optimization problem envisages a competitive IBDR frame-

work in the presence of multiple DRPs. It manifests the potential techno-economic
benefits with IBDR participation.

2. The proposed bi-level problem and the solution approach enumerate the potential
benefits of each participant in a non-cooperative environment with the minimal in-
formation sharing.

3. The IBDR participation in the DS significantly lowers the LSE’s operating & pur-
chasing costs, customers’ bills, power losses, and demand requirements from the
utility grid.

4. The discrete incentive rates and corresponding induced DR demands of the different
customer classes reveal that each customer’s class actively participates in DR.

7.1 Salient Contributions

The salient contributions of the thesis are outlined as follows:

1. An economic price-based DR model using an adaptive elasticity in PEM is proposed

to imitate the customers’ behaviour in DR.
Chapter 7. Conclusions and Future Scope 143

2. A new and improved economic incentive-based DR model is developed to consider

the exact effect of incentive rate inclusion in PEM on DR performance.

3. The comprehensive PBDR models using the various CUFs are developed, and their
DR response are assessed under the different load patterns, parameters, price, and
risk-based indices.

4. A micro-economic behavioral model, OLG, is adopted and developed for PBDR

representation.

5. A hierarchical tri-level PBDR strategy is developed considering the pragmatic char-

acteristics of DS under the DR environment.

6. Developed a bi-level IBDR strategy under the ambit of multiple DRPs in view of
the competitive environment in DS.

7.2 Future Scope

The future research can be carried out in the following directions as summarized below:

1. The proposed adaptive PBDR and new IBDR model are based on PEM to illustrate
the customer’s pragmatic behaviour in DS. Since DS consists of the different cus-
tomer classes’ load patterns. Hence, the proposed PEM-based DR approach can be
extended to investigate the customer’s DR response when price patterns are designed
according to customers’ load patterns.

2. The CUFs based PBDR models are an effective modeling approach in DR. However,
less emphasis is given to their parameter estimation and calibration under DR vari-
ation. Thus, the future research can be focused on the better estimation approaches
using predictive tools such as machine learning, artificial neural network, etc.

3. The OLG based DR program can be further extended into a more generalization of
the OLG model for multi-states. Moreover, a more explicit model can be the area
of future work to include the customers’ attributes in DR modeling.

4. The proposed DR models are based on the assumption of the customers’ rational
behaviour. The future research may be conducted in the direction of customers’
irrational behaviour in DR modeling.
Chapter 7. Conclusions and Future Scope 144

5. The proposed hierarchical PBDR framework can be investigated considering multi-

objective problems under distributed generation, congested networks, renewable pen-
etration, multiple LSEs, etc., in the future work. In addition, the quantification of
uncertainty in price and demand can be part of future research work.

6. The developed IBDR framework can be extended with the other emerging stakehold-
ers such as EV, ESS, RES using multiple LSEs under the competitive environment
in the contemporary DS.
Appendix A

Test Distribution Systems

A.1 Study system 1

The standard 33-bus distribution is rated at 12.66 kV with the nominal active and reactive
power demand of 3715 kW and 2300 kVAr, respectively. The system’s single line diagram
is represented in Fig. A.1.

Figure A.1: 33-bus distribution system

The system’s base parameters are Sbase = 100MVA and Vbase = 12.66kV. The load flow
data of the system are tabulated below.

145
Appendix A. Test Distribution Systems 146

Table A.1: Line and load data of the standard 33-bus test distribution system

Branch R X Rec. end load

Sen. end Rec. end
PL QL
no. (Ω) (Ω)
(kW) (kVAr)
1 1 2 0.0922 0.047 100 60
2 2 3 0.493 0.2512 90 40
3 3 4 0.3661 0.1864 120 80
4 4 5 0.3811 0.1941 60 30
5 5 6 0.819 0.707 60 20
6 6 7 0.1872 0.6188 200 100
7 7 8 0.7115 0.2351 200 100
8 8 9 1.0299 0.74 60 20
9 9 10 1.044 0.74 60 20
10 10 11 0.1967 0.0651 45 30
11 11 12 0.3744 0.1298 60 35
12 12 13 1.468 1.1549 60 35
13 13 14 0.5416 0.7129 120 80
14 14 15 0.5909 0.526 60 10
15 15 16 0.7462 0.5449 60 20
16 16 17 1.2889 1.721 60 20
17 17 18 0.732 0.5739 90 40
18 2 19 0.164 0.1565 90 40
19 19 20 1.5042 1.3555 90 40
20 20 21 0.4095 0.4784 90 40
21 21 22 0.7089 0.9373 90 40
22 3 23 0.4512 0.3084 90 50
23 23 24 0.898 0.7091 420 200
24 24 25 0.8959 0.7071 420 200
25 6 26 0.2031 0.1034 60 25
26 26 27 0.2842 0.1447 60 25
27 27 28 1.0589 0.9338 60 20
28 28 29 0.8043 0.7006 120 70
29 29 30 0.5074 0.2585 200 600
30 30 31 0.9745 0.9629 150 70
31 31 32 0.3105 0.3619 210 100
32 32 33 0.3411 0.5302 60 40
Appendix A. Test Distribution Systems 147

A.2 Study System 2

This study system is 108-bus practical radial Indian distribution system of Jhawlawar city,
Rajasthan, rated at 11 kV. It has nominal active and reactive power demand of 12132 kW
and 9099 kVAr, respectively. A single-line diagram of the system is shown in Fig. A.2.

Figure A.2: Indian 108-bus distribution system

The system’s base parameters are Sbase = 100MVA and Vbase = 11kV. The load flow data
of the system are tabulated below.

Table A.2: Line and load data of Indian 108-bus distribution system

R X Rec. end load

Branch no. Sen. end Rec. end
(Ω) (Ω) P (kW) Q (kVAr)
1 1 2 0.0814 0.0921 0 0
2 2 3 0.0555 0.0628 0 0
3 3 4 0.1443 0.1632 0 0
4 4 5 0.297 0.047 0 0
Continued on next page
Appendix A. Test Distribution Systems 148

Table A.2 – Continued from previous page

R X Rec. end load
Branch no. Sen. end Rec. end
(Ω) (Ω) P (kW) Q (kVAr)
5 5 6 1.155 0.1827 96 72
6 6 7 0.297 0.047 38.4 28.8
7 7 8 0.594 0.094 38.4 28.8
8 4 9 0.0666 0.0753 60 45
9 9 10 0.594 0.094 0 0
10 10 11 0.627 0.0992 60 45
11 11 12 0.297 0.047 60 45
12 12 13 0.495 0.0783 60 45
13 13 14 0.297 0.047 98.4 73.8
14 14 15 0.297 0.047 60 45
15 9 16 0.0814 0.0921 96 72
16 16 17 0.0925 0.1046 96 72
17 17 18 0.0396 0.0385 96 72
18 18 19 0.0352 0.0342 96 72
19 19 20 0.0396 0.0385 96 72
20 20 21 0.0352 0.0342 96 72
21 21 22 0.0999 0.0424 96 72
22 22 23 0.0555 0.0236 96 72
23 23 24 0.0888 0.0377 96 72
24 24 25 0.0888 0.0377 0 0
25 25 26 0.1665 0.0707 96 72
26 26 27 0.165 0.0261 96 72
27 27 28 0.297 0.047 96 72
28 28 29 0.231 0.0365 96 72
29 29 30 0.297 0.047 0 0
30 30 31 0.396 0.0626 96 72
31 31 32 0.363 0.0574 96 72
32 32 33 0.231 0.0365 96 72
33 33 34 0.264 0.0418 96 72
34 34 35 0.363 0.0574 96 72
35 18 36 0.627 0.0992 96 72
36 36 37 0.627 0.0992 96 72
37 37 38 0.627 0.0992 96 72
38 19 39 0.495 0.0783 96 72
39 39 40 0.198 0.0313 96 72
Continued on next page
Appendix A. Test Distribution Systems 149

Table A.2 – Continued from previous page

R X Rec. end load
Branch no. Sen. end Rec. end
(Ω) (Ω) P (kW) Q (kVAr)
40 39 41 0.231 0.0365 96 72
41 21 42 0.495 0.0783 96 72
42 42 43 0.165 0.0261 96 72
43 43 44 0.165 0.0261 96 72
44 23 45 0.297 0.047 96 72
45 45 46 0.297 0.047 96 72
46 46 47 0.297 0.047 96 72
47 46 48 0.231 0.0365 96 72
48 26 49 0.297 0.047 192 144
49 28 50 0.198 0.0313 96 72
50 50 51 0.363 0.0574 0 0
51 30 52 0.495 0.0783 96 72
52 10 53 0.297 0.047 60 45
53 10 54 0.594 0.094 60 45
54 54 55 0.297 0.047 0 0
55 1 56 0.666 0.2826 0 0
56 56 57 1.554 0.6594 0 0
57 57 58 1.11 0.471 0 0
58 58 59 0.333 0.1413 0 0
59 59 60 0.3441 0.146 1200 900
60 57 61 0.825 0.1305 240 180
61 58 62 1.155 0.1827 240 180
62 1 63 1.056 0.7128 2400 1800
63 1 64 0.396 0.2673 0 0
64 64 65 0.099 0.0668 0 0
65 65 66 0.858 0.1357 240 180
66 66 67 0.693 0.1096 0 0
67 67 68 1.023 0.1618 240 180
68 68 69 0.891 0.1409 0 0
69 69 70 1.155 0.1827 0 0
70 70 71 0.693 0.1096 0 0
71 71 72 0.759 0.1201 60 45
72 72 73 0.66 0.1044 60 45
73 67 74 0.726 0.1148 0 0
74 74 75 1.023 0.1618 240 180
Continued on next page
Appendix A. Test Distribution Systems 150

Table A.2 – Continued from previous page

R X Rec. end load
Branch no. Sen. end Rec. end
(Ω) (Ω) P (kW) Q (kVAr)
75 75 76 0.891 0.1409 0 0
76 76 77 0.495 0.0783 96 72
77 65 78 0.3663 0.1554 0 0
78 78 79 0.222 0.0942 0 0
79 79 80 0.198 0.0313 0 0
80 80 81 0.462 0.0731 0 0
81 81 82 0.264 0.0418 0 0
82 82 83 0.363 0.0574 0 0
83 83 84 0.495 0.0783 60 45
84 84 85 1.089 0.1723 240 180
85 85 86 0.66 0.1044 240 180
86 78 87 2.97 0.4698 240 180
87 79 88 0.363 0.0574 240 180
88 80 89 0.924 0.1462 0 0
89 89 90 0.462 0.0731 60 45
90 81 91 0.627 0.0992 96 72
91 82 92 0.165 0.0261 38.4 28.8
92 83 93 0.231 0.0365 60 45
93 1 94 0.6549 0.2779 0 0
94 94 95 0.666 0.2826 0 0
95 95 96 1.887 0.8007 0 0
96 96 97 0.0999 0.0424 768 576
97 97 98 0.858 0.1357 0 0
98 98 99 2.046 0.3236 0 0
99 99 100 0.297 0.047 0 0
100 100 101 0.693 0.1096 0 0
101 101 102 0.726 0.1148 151.2 113.4
102 102 103 0.825 0.1305 0 0
103 103 104 0.825 0.1305 0 0
104 104 105 0.858 0.1357 240 180
105 95 106 0.297 0.047 240 180
106 99 107 0.726 0.1148 151.2 113.4
107 102 108 0.099 0.0157 240 180
Appendix B

Customer Classes and Technical

Information

Table B.1: Customer classes information used in 33-bus distribution system

Demand Max flexible S/CF demand range

Classes Buses No. of Cust.
(kW) demand (kW) (κk ) [min, max]
R 2-18 950 121 290 -0.2 [2, 15]
LI 19-25 1290 17 320 0 [10, 25]
MI 26-28 180 6 60 0.2 [50, 100]
C 11-18 555 28 160 1 [20, 45]
A 29-33 740 63 165 -0.5 [8, 18]
R: Residential, LI: Large Industrial, MI: Medium Industrial, C: Commercial, A: Agriculture

Table B.2: Customer classes information used in 108-bus distribution system

Demand Max. Flexible

Classes Buses No. of Cust.
(kW) demand (kW)
R 2-43 3067 386 887
LI 55-71 4560 60 1166
MI 72-89 1476 49 420
C 44-54 984 51 266
A 90-108 2045 171 491

151
Appendix B. Customer Classes and Technical Information 152

Table B.3: Nominal load factor for the various customer classes

Time Customer class Time Customer class

(Hr.) R LI MI C A (Hr.) R LI MI C A
1 0.40 0.80 0.10 0.10 0.50 13 0.70 0.95 0.95 0.75 0.10
2 0.36 0.80 0.10 0.10 0.56 14 0.80 0.80 1.00 0.80 0.10
3 0.28 0.80 0.10 0.10 0.42 15 0.80 1.00 1.00 0.82 0.15
4 0.22 0.80 0.10 0.10 0.46 16 0.80 1.00 1.00 0.85 0.20
5 0.34 0.85 0.40 0.15 0.68 17 0.85 0.95 0.95 0.90 0.30
6 0.42 0.75 0.58 0.15 0.70 18 0.90 0.95 0.95 1.00 0.45
7 0.68 0.85 0.85 0.20 0.78 19 0.95 0.85 0.85 1.00 0.50
8 0.75 0.90 0.90 0.26 0.80 20 1.00 0.85 0.85 1.00 0.68
9 0.88 0.90 0.90 0.50 0.30 21 0.76 0.80 0.80 1.00 0.74
10 0.76 0.90 0.90 0.55 0.25 22 0.60 0.75 0.30 0.80 0.80
11 0.70 0.95 0.95 0.60 0.20 23 0.50 0.80 0.20 0.20 0.90
12 0.63 0.95 0.95 0.65 0.15 24 0.40 0.80 0.15 0.10 0.70
R: Residential, LI: Large Industrial, MI: Medium Industrial, C: Commercial, A: Agriculture

Table B.4: Techno-economic parameter used in the study

Value
Parameters
33-bus 108-bus
P G,n,t /P G,n,t (MW) 0/0.05 0/0.15
QG,n,t /QG,n,t (MVAr) 0/0.03 0/0.12
V n /V n (p.u) 0.90/1.05 0.85/1.05
Bibliography

[1] F. C. Schweppe, M. C. Caramanis, R. D. Tabors, and R. E. Bohn, Spot

Pricing of Electricity. Boston, MA: Springer US, 1988. [Online]. Available: http:
//link.springer.com/10.1007/978-1-4613-1683-1

[2] A. Faruqui and J. Palmer, “Dynamic pricing and its discontents,” Regulation, vol. 34, p. 16,
2011.

[3] R. Deng, Z. Yang, M.-Y. Chow, and J. Chen, “A survey on demand response in smart
grids: Mathematical models and approaches,” IEEE Transactions on Industrial Informatics,
vol. 11, no. 3, pp. 570–582, 2015.

[4] F. Rahimi and A. Ipakchi, “Demand response as a market resource under the smart grid
paradigm,” IEEE Transactions on smart grid, vol. 1, no. 1, pp. 82–88, 2010.

[5] P. Palensky and D. Dietrich, “Demand side management: Demand response, intelligent
energy systems, and smart loads,” IEEE transactions on industrial informatics, vol. 7, no. 3,
pp. 381–388, 2011.

[6] P. Siano, “Demand response and smart grids—a survey,” Renewable and sustainable energy
reviews, vol. 30, pp. 461–478, 2014.

[7] U. D. of Energy, “Benefits of demand response in electricity markets and recommendations

for achieving them - a report to the united states congress pursant to section 1252 of the
energy policy act of 2005,” U.S. Department of Energy, Tech. Rep., 02/2006 2006.

[8] R. Bharvirkar, “Retail demand response in southwest power pool,” 2009.

[9] P. T. Baboli, M. Eghbal, M. P. Moghaddam, and H. Aalami, “Customer behavior based

demand response model,” in 2012 IEEE Power and Energy Society General Meeting. IEEE,
2012, pp. 1–7.

[10] H. Aalami, M. P. Moghaddam, and G. Yousefi, “Modeling and prioritizing demand response
programs in power markets,” Electric Power Systems Research, vol. 80, no. 4, pp. 426–435,
2010.

[11] M. Rahmani-andebili, “Modeling nonlinear incentive-based and price-based demand response

programs and implementing on real power markets,” Electric Power Systems Research, vol.
132, pp. 115–124, 2016.

153
References 154

[12] M. P. Moghaddam, A. Abdollahi, and M. Rashidinejad, “Flexible demand response programs

modeling in competitive electricity markets,” Applied Energy, vol. 88, no. 9, pp. 3257–3269,
2011.

[13] H. Aalami, M. P. Moghaddam, and G. Yousefi, “Evaluation of nonlinear models for time-
based rates demand response programs,” International Journal of Electrical Power & Energy
Systems, vol. 65, pp. 282–290, 2015.

[14] D. S. Kirschen, G. Strbac, P. Cumperayot, and D. de Paiva Mendes, “Factoring the elasticity
of demand in electricity prices,” IEEE Transactions on Power Systems, vol. 15, no. 2, pp.
612–617, 2000.

[15] D. S. Kirschen, “Demand-side view of electricity markets,” IEEE Transactions on power

systems, vol. 18, no. 2, pp. 520–527, 2003.

[16] B. Neenan and J. Eom, “Price elasticity of demand for electricity: a primer and synthesis,”
Electric Power Research Institute, 2008.

[17] G. Barbose, C. Goldman, and B. Neenan, “Real time pricing tariffs: A survey of utility
program experience,” LBNL-54238, March, 2004.

[18] L. K. Panwar, S. Reddy, A. Verma, and B. Panigrahi, “Dynamic incentive framework for
demand response in distribution system using moving time horizon control,” IET Generation,
Transmission & Distribution, vol. 11, no. 17, pp. 4338–4347, 2017.

[19] A. Asadinejad, A. Rahimpour, K. Tomsovic, H. Qi, and C.-f. Chen, “Evaluation of residential
customer elasticity for incentive based demand response programs,” Electric Power Systems
Research, vol. 158, pp. 26–36, 2018.

[20] C.-L. Su and D. Kirschen, “Quantifying the effect of demand response on electricity markets,”
IEEE Transactions on Power Systems, vol. 24, no. 3, pp. 1199–1207, 2009.

[21] A. Asadinejad, M. G. Varzaneh, K. Tomsovic, C.-f. Chen, and R. Sawhney, “Residential

customers elasticity estimation and clustering based on their contribution at incentive based
demand response,” in 2016 IEEE Power and Energy Society General Meeting (PESGM).
IEEE, 2016, pp. 1–5.

[22] A. Asadinejad and K. Tomsovic, “Optimal use of incentive and price based demand response
to reduce costs and price volatility,” Electric Power Systems Research, vol. 144, pp. 215–223,
2017.

[23] Y. Otani and M. El-Hodiri, Microeconomic Theory. Springer Berlin Heidelberg, 1987, vol. 47,
no. 186.

[24] D. Salvatore et al., Microeconomics: theory and applications. Oxford University Press, 2008.

[25] B. Chai, J. Chen, Z. Yang, and Y. Zhang, “Demand response management with multiple
utility companies: A two-level game approach,” IEEE Transactions on Smart Grid, vol. 5,
no. 2, pp. 722–731, 2014.
References 155

[26] P. Shinde and K. S. Swarup, “Stackelberg game-based demand response in multiple utility
environments for electric vehicle charging,” IET Electrical Systems in Transportation, vol. 8,
no. 3, pp. 167–174, 2018.

[27] S. Mohajeryami, I. N. Moghaddam, M. Doostan, B. Vatani, and P. Schwarz, “A novel eco-

nomic model for price-based demand response,” Electric Power Systems Research, vol. 135,
pp. 1–9, 2016.

[28] R. Sharifi, S. H. Fathi, A. Anvari-Moghaddam, J. M. Guerrero, and V. Vahidinasab, “An

economic customer-oriented demand response model in electricity markets,” in 2018 IEEE
International Conference on Industrial Technology (ICIT). IEEE, 2018, pp. 1149–1153.

[29] R. Sharifi, A. Anvari-Moghaddam, S. H. Fathi, J. M. Guerrero, and V. Vahidinasab, “Eco-

nomic demand response model in liberalised electricity markets with respect to flexibility of
consumers,” IET Generation, Transmission & Distribution, vol. 11, no. 17, pp. 4291–4298,
2017.

[30] S. Maharjan, Q. Zhu, Y. Zhang, S. Gjessing, and T. Basar, “Dependable demand response
management in the smart grid: A Stackelberg game approach,” IEEE Transactions on Smart
Grid, vol. 4, no. 1, pp. 120–132, 2013.

[31] M. Yu and S. H. Hong, “A real-time demand-response algorithm for smart grids: A Stackel-
berg game approach,” IEEE Transactions on smart grid, vol. 7, no. 2, pp. 879–888, 2015.

[32] A. Safdarian, M. Fotuhi-Firuzabad, and M. Lehtonen, “Integration of price-based demand

response in DisCos’ short-term decision model,” IEEE Transactions on Smart Grid, vol. 5,
no. 5, pp. 2235–2245, 2014.

[33] ——, “A medium-term decision model for disCos: Forward contracting and TOU pricing,”
IEEE Transactions on Power Systems, vol. 30, no. 3, pp. 1143–1154, 2015.

[34] T. Lu, Z. Wang, J. Wang, Q. Ai, and C. Wang, “A data-driven Stackelberg market strat-
egy for demand response-enabled distribution systems,” IEEE Transactions on Smart Grid,
vol. 10, no. 3, pp. 2345–2357, 2018.

[35] D. T. Nguyen, H. T. Nguyen, and L. B. Le, “Dynamic pricing design for demand response
integration in power distribution networks,” IEEE Transactions on power systems, vol. 31,
no. 5, pp. 3457–3472, 2016.

[36] C. Feng, Z. Li, M. Shahidehpour, F. Wen, and Q. Li, “Stackelberg game based transactive
pricing for optimal demand response in power distribution systems,” International Journal
of Electrical Power & Energy Systems, vol. 118, p. 105764, 2020.

[37] M. Zugno, J. M. Morales, P. Pinson, and H. Madsen, “A bilevel model for electricity retailers’
participation in a demand response market environment,” Energy Economics, vol. 36, pp.
182–197, 2013.

[38] M. R. Sarker, M. A. Ortega-Vazquez, and D. S. Kirschen, “Optimal coordination and schedul-

ing of demand response via monetary incentives,” IEEE Transactions on Smart Grid, vol. 6,
no. 3, pp. 1341–1352, 2014.
References 156

[39] M. Yu and S. H. Hong, “Incentive-based demand response considering hierarchical electricity

market: A Stackelberg game approach,” Applied Energy, vol. 203, pp. 267–279, 2017.

[40] W. Wei, F. Liu, and S. Mei, “Energy pricing and dispatch for smart grid retailers under
demand response and market price uncertainty,” IEEE transactions on smart grid, vol. 6,
no. 3, pp. 1364–1374, 2014.

[41] C. Feng, Z. Li, M. Shahidehpour, F. Wen, and Q. Li, “Stackelberg game based transactive
pricing for optimal demand response in power distribution systems,” International Journal
of Electrical Power and Energy Systems, vol. 118, 2020.

[42] J. Wang, Q. Huang, W. Hu, J. Li, Z. Zhang, D. Cai, X. Zhang, and N. Liu, “Ensuring
profitability of retailers via Shapley Value based demand response,” International Journal of
Electrical Power & Energy Systems, vol. 108, pp. 72–85, 2019.

[43] Z. Li, S. Wang, X. Zheng, F. De Leon, and T. Hong, “Dynamic demand response using
customer coupons considering multiple load aggregators to simultaneously achieve efficiency
and fairness,” IEEE Transactions on Smart Grid, vol. 9, no. 4, pp. 3112–3121, 2018.

[44] M. A. F. Ghazvini, P. Faria, S. Ramos, H. Morais, and Z. Vale, “Incentive-based demand

response programs designed by asset-light retail electricity providers for the day-ahead mar-
ket,” Energy, vol. 82, pp. 786–799, 2015.

[45] M. A. F. Ghazvini, J. Soares, N. Horta, R. Neves, R. Castro, and Z. Vale, “A multi-objective

model for scheduling of short-term incentive-based demand response programs offered by
electricity retailers,” Applied energy, vol. 151, pp. 102–118, 2015.

[46] S. Yousefi, M. P. Moghaddam, and V. J. Majd, “Optimal real time pricing in an agent-based
retail market using a comprehensive demand response model,” Energy, vol. 36, no. 9, pp.
5716–5727, 2011.

[47] G. Gutiérrez-Alcaraz, J. H. Tovar-Hernández, and C. N. Lu, “Effects of demand response

programs on distribution system operation,” International Journal of Electrical Power and
Energy Systems, vol. 74, pp. 230–237, 2016.

[48] B. Zeng, J. Zhang, X. Yang, J. Wang, J. Dong, and Y. Zhang, “Integrated planning for
transition to low-carbon distribution system with renewable energy generation and demand
response,” IEEE Transactions on Power Systems, vol. 29, no. 3, pp. 1153–1165, 2014.

[49] R. Deng, Z. Yang, F. Hou, M.-Y. Chow, and J. Chen, “Distributed real-time demand response
in multiseller–multibuyer smart distribution grid,” IEEE Transactions on Power Systems,
vol. 30, no. 5, pp. 2364–2374, 2014.

[50] J. Yusta, H. Khodr, and A. Urdaneta, “Optimal pricing of default customers in electrical
distribution systems: Effect behavior performance of demand response models,” Electric
Power Systems Research, vol. 77, no. 5-6, pp. 548–558, 2007.

[51] J. Kirtley, S. Kennedy, and J. Wang, “A demand responsive bidding mechanism with
price elasticity matrix in wholesale electricity pools,” Electrical Engineering, 2008. [Online].
Available: http://dspace.mit.edu/handle/1721.1/54447
References 157

[52] A. J. Roscoe and G. Ault, “Supporting high penetrations of renewable generation via im-
plementation of real-time electricity pricing and demand response,” IET Renewable Power
Generation, vol. 4, no. 4, pp. 369–382, 2010.

[53] C. Zhao, J. Wang, J.-P. Watson, and Y. Guan, “Multi-stage robust unit commitment con-
sidering wind and demand response uncertainties,” IEEE Transactions on Power Systems,
vol. 28, no. 3, pp. 2708–2717, 2013.

[54] V. Gómez, M. Chertkov, S. Backhaus, and H. J. Kappen, “Learning price-elasticity of smart

consumers in power distribution systems,” in 2012 IEEE Third International Conference on
Smart Grid Communications (SmartGridComm). IEEE, 2012, pp. 647–652.

[55] A. Dadkhah and B. Vahidi, “On the network economic, technical and reliability charac-
teristics improvement through demand-response implementation considering consumers’ be-
haviour,” IET Generation, Transmission & Distribution, vol. 12, no. 2, pp. 431–440, 2017.

[56] E. Shahryari, H. Shayeghi, B. Mohammadi-ivatloo, and M. Moradzadeh, “An improved

incentive-based demand response program in day-ahead and intra-day electricity markets,”
Energy, vol. 155, pp. 205–214, 2018.

[57] N. Venkatesan, J. Solanki, and S. K. Solanki, “Demand response model and its effects on
voltage profile of a distribution system,” IEEE Power and Energy Society General Meeting,
2011.

[58] M. P. Moghaddam, A. Abdollahi, and M. Rashidinejad, “Flexible demand response programs

modeling in competitive electricity markets,” Applied Energy, vol. 88, no. 9, pp. 3257–3269,
2011.

[59] N. Li, L. Chen, and S. H. Low, “Optimal demand response based on utility maximization in
power networks,” in 2011 IEEE power and energy society general meeting. IEEE, 2011, pp.
1–8.

[60] Z. Fan, “A distributed demand response algorithm and its application to PHEV charging in
smart grids,” IEEE Transactions on Smart Grid, vol. 3, no. 3, pp. 1280–1290, 2012.

[61] J. Ma, H. H. Chen, L. Song, and Y. Li, “Residential load scheduling in smart grid: A cost
efficiency perspective,” IEEE transactions on smart grid, vol. 7, no. 2, pp. 771–784, 2015.

[62] M. Fahrioglu and F. L. Alvarado, “Using utility information to calibrate customer demand
management behavior models,” IEEE transactions on power systems, vol. 16, no. 2, pp.
317–322, 2001.

[63] V. Pradhan, V. M. Balijepalli, and S. A. Khaparde, “An effective model for demand response
management systems of residential electricity consumers,” IEEE Systems Journal, vol. 10,
no. 2, pp. 434–445, 2014.

[64] D. Romer, Advanced Macroeconomics, USA: Mc Graw Hill. USA: Mc Graw Hill, 2012.

[65] D. Gowland, “Lectures on macroeconomics,” 1990.

References 158

[66] L. Göransson, J. Goop, T. Unger, M. Odenberger, and F. Johnsson, “Linkages between

demand-side management and congestion in the European electricity transmission system,”
Energy, vol. 69, pp. 860–872, 2014.

[67] A. J. Conejo, J. M. Morales, and L. Baringo, “Real-time demand response model,” IEEE
Transactions on Smart Grid, vol. 1, no. 3, pp. 236–242, 2010.

[68] A. Mosaddegh, C. A. Cañizares, and K. Bhattacharya, “Optimal demand response for distri-
bution feeders with existing smart loads,” IEEE Transactions on Smart Grid, vol. 9, no. 5,
pp. 5291–5300, 2017.

[69] I. Sharma, K. Bhattacharya, and C. Cañizares, “Smart distribution system operations with
price-responsive and controllable loads,” IEEE Transactions on Smart Grid, vol. 6, no. 2,
pp. 795–807, 2014.

[70] N. Mohammad and Y. Mishra, “Coordination of wind generation and demand response to
minimise operation cost in day-ahead electricity markets using bi-level optimisation frame-
work,” IET Generation, Transmission & Distribution, vol. 12, no. 16, pp. 3793–3802, 2018.

[71] P. Yang, G. Tang, and A. Nehorai, “A game-theoretic approach for optimal time-of-use
electricity pricing,” IEEE Transactions on Power Systems, vol. 28, no. 2, pp. 884–892, 2012.

[72] L. Gkatzikis, I. Koutsopoulos, and T. Salonidis, “The role of aggregators in smart grid
demand response markets,” IEEE Journal on selected areas in communications, vol. 31,
no. 7, pp. 1247–1257, 2013.

[73] D. Aussel, L. Brotcorne, S. Lepaul, and L. von Niederhäusern, “A trilevel model for best
response in energy demand-side management,” European Journal of Operational Research,
vol. 281, no. 2, pp. 299–315, 2020.

[74] Y. Jia, Z. Mi, Y. Yu, Z. Song, and H. Fan, “Tri-level decision-making framework for strategic
trading of demand response aggregator,” IET Renewable Power Generation, vol. 13, no. 12,
pp. 2195–2206, 2019.

[75] D. Qiu, D. Papadaskalopoulos, Y. Ye, and G. Strbac, “Investigating the effects of demand
flexibility on electricity retailers’ business through a tri-level optimisation model,” IET Gen-
eration, Transmission & Distribution, vol. 14, no. 9, pp. 1739–1750, 2020.

[76] A.-Y. Yoon, Y.-J. Kim, and S.-I. Moon, “Optimal retail pricing for demand response of hvac
systems in commercial buildings considering distribution network voltages,” IEEE Transac-
tions on Smart Grid, vol. 10, no. 5, pp. 5492–5505, 2018.

[77] L. Li, “Coordination between smart distribution networks and multi-microgrids considering
demand side management: A trilevel framework,” Omega, vol. 102, p. 102326, 2021.

[78] D. P. Bertsekas, Constrained optimization and Lagrange multiplier methods. Academic press,
2014.

[79] B. Zeng and Y. An, “Solving bilevel mixed integer program by reformulations and decompo-
sition,” Optimization online, pp. 1–34, 2014.
References 159

[80] H. Haghighat and B. Zeng, “Bilevel mixed integer transmission expansion planning,” IEEE
Transactions on Power Systems, vol. 33, no. 6, pp. 7309–7312, 2018.

[81] P. McNamara and S. McLoone, “Hierarchical demand response for peak minimization using
Dantzig–Wolfe decomposition,” IEEE Transactions on Smart Grid, vol. 6, no. 6, pp. 2807–
2815, 2015.

[82] M. Shahidehopour and Y. Fu, “Benders decomposition: applying benders decomposition to

power systems,” IEEE Power and Energy Magazine, vol. 3, no. 2, pp. 20–21, 2005.

[83] G. Bautista, M. F. Anjos, and A. Vannelli, “Formulation of oligopolistic competition in AC

power networks: An NLP approach,” IEEE Transactions on Power Systems, vol. 22, no. 1,
pp. 105–115, 2007.

[84] B. F. Hobbs, C. B. Metzler, and J.-S. Pang, “Strategic gaming analysis for electric power
systems: An MPEC approach,” IEEE transactions on power systems, vol. 15, no. 2, pp.
638–645, 2000.

[85] J. B. Cardell, C. C. Hitt, and W. W. Hogan, “Market power and strategic interaction in
electricity networks,” Resource and energy economics, vol. 19, no. 1-2, pp. 109–137, 1997.

[86] C.-L. Su, “A sequential NCP algorithm for solving equilibrium problems with equilibrium
constraints,” Manuscript, Department of Management Science and Engineering, Stanford
University, Stanford, CA, 2004.

[87] S. Leyffer and T. Munson, “Solving multi-leader–common-follower games,” Optimisation

Methods & Software, vol. 25, no. 4, pp. 601–623, 2010.

[88] M. de Luján Latorre and S. Granville, “The Stackelberg equilibrium applied to AC power
systems-a noninterior point algorithm,” IEEE Transactions on Power Systems, vol. 18, no. 2,
pp. 611–618, 2003.

[89] A. A. Algarni and K. Bhattacharya, “A generic operations framework for discos in retail
electricity markets,” IEEE Transactions on Power Systems, vol. 24, no. 1, pp. 356–367,
2009.

[90] H. Haghighat and S. W. Kennedy, “A bilevel approach to operational decision making of a

distribution company in competitive environments,” IEEE Transactions on Power Systems,
vol. 27, no. 4, pp. 1797–1807, 2012.

[91] G. A. Jiménez-Estévez, R. Palma-Behnke, R. Torres-Avila, and L. S. Vargas, “A competitive

market integration model for distributed generation,” IEEE Transactions on power systems,
vol. 22, no. 4, pp. 2161–2169, 2007.

[92] R. H. Boroumand, “Electricity markets and oligopolistic behaviors: The impact of a mul-
timarket structure,” Research in International Business and Finance, vol. 33, pp. 319–333,
2015.
References 160

[93] J. M. López-Lezama, A. Padilha-Feltrin, J. Contreras, and J. I. Muñoz, “Optimal contract

pricing of distributed generation in distribution networks,” IEEE transactions on power sys-
tems, vol. 26, no. 1, pp. 128–136, 2010.

[94] Q. Hu, F. Li, X. Fang, and L. Bai, “A framework of residential demand aggregation with
financial incentives,” IEEE Transactions on Smart Grid, vol. 9, no. 1, pp. 497–505, 2016.

[95] D. Jang, J. Eom, M. J. Park, and J. J. Rho, “Variability of electricity load patterns and its
effect on demand response: A critical peak pricing experiment on korean commercial and
industrial customers,” Energy Policy, vol. 88, pp. 11–26, 2016.

[96] D. P. Bertsekas, “The method of multipliers for equality constrained problems,” Constrained
optimization and Lagrange multiplier methods, pp. 96–157, 1982.

[97] B. Group et al., “Quantifying demand response benefits in PJM,” Prepared for PJM Inter-
connection, LLC and the Mid-Atlantic Distributed Resources Initiative (MADRI), 2007.

[98] J. Jiang, Y. Kou, Z. Bie, and G. Li, “Optimal Real-Time Pricing of Electricity Based on
Demand Response,” Energy Procedia, vol. 159, pp. 304–308, 2019.

[99] A. Zerrahn and W. P. Schill, “On the representation of demand-side management in power
system models,” Energy, vol. 84, pp. 840–845, 2015.

[100] W. A. Thompson, H. M. Taylor, and S. Karlin, “An Introduction to Stochastic Modeling.”

Journal of the American Statistical Association, vol. 80, no. 390, p. 491, 1985.

[101] A. Etheridge, A Course in Financial Calculus. Cambridge University Press, aug 2002,
vol. 40, no. 09. [Online]. Available: https://www.cambridge.org/core/product/identifier/
9780511810107/type/book

[102] Y.-d. Lyuu, “Ito Process,” pp. 489–564, 2012.

[103] L. A. Greening, D. L. Greene, and C. Difiglio, “Energy efficiency and consumption — the
rebound effect — a survey,” Energy Policy, vol. 28, no. 6-7, pp. 389–401, jun 2000. [Online].
Available: https://linkinghub.elsevier.com/retrieve/pii/S0301421500000215

[104] A. Faruqui, “The ethics of dynamic pricing,” in Smart grid. Elsevier, 2012, pp. 61–83.

[105] J. Burkardt, “The truncated normal distribution,” Department of Scientific Computing Web-
site, Florida State University, pp. 1–35, 2014.

[106] M. E. Baran and F. F. Wu, “Network reconfiguration in distribution systems for loss reduction
and load balancing,” IEEE Power Engineering Review, vol. 9, no. 4, pp. 101–102, 1989.

[107] “Distribution of electricity consumption across India in financial year 2020,

by sector,” 2020. [Online]. Available: https://www.statista.com/statistics/1130112/
india-electricity-consumption-share-by-sector/

[108] N. Kanwar, N. Gupta, K. R. Niazi, and A. Swarnkar, “Optimal distributed generation al-
location in radial distribution systems considering customer-wise dedicated feeders and load
patterns,” Journal of Modern Power Systems and Clean Energy, vol. 3, no. 4, pp. 475–484,
2015.
References 161

[109] O. E. Board, “Ontario Energy Board,” https://www.oeb.ca/rates-and-your-bill/

electricity-rates/, 2008, [Online; accessed 2020].

[110] S. D. Das and R. Srikanth, “Viability of power distribution in india–challenges and way
forward,” Energy Policy, vol. 147, p. 111882, 2020.

[111] J. Stewart, H. Haeri, and L. Garth, “Demand Response Elasticities Analysis,” 2018.

[112] S. M. B. Sadati, J. Moshtagh, M. Shafie-khah, and J. P. Catalão, “Smart distribution sys-

tem operational scheduling considering electric vehicle parking lot and demand response
programs,” Electric Power Systems Research, vol. 160, pp. 404–418, 2018.

[113] B. Shen, G. Ghatikar, C. C. Ni, P. Martin, and G. Wikler, “Addressing Energy Demand
through Demand Response : International Experiences and Practices,” no. June, 2012.

[114] H. Ming, A. A. Thatte, and L. Xie, “Revenue Inadequacy with Demand Response Providers:
A Critical Appraisal,” IEEE Transactions on Smart Grid, vol. 10, no. 3, pp. 3282–3291,
2019.

[115] F. P. Ramsey, “A contribution to the theory of taxation,” The economic journal, vol. 37, no.
145, pp. 47–61, 1927.

[116] M. Badtke-Berkow, M. Centore, K. Mohlin, and B. Spiller, “A Primer on Time-Variant

Electricity Pricing,” Environmental Defense Fund, pp. 1–24, 2015. [Online]. Available:
https://www.edf.org/sites/default/files/a primer on time-variant pricing.pdf

[117] M. Mulder and B. Willems, “The dutch retail electricity market,” Energy policy, vol. 127,
pp. 228–239, 2019.

[118] P. Samadi, A.-H. Mohsenian-Rad, R. Schober, V. W. Wong, and J. Jatskevich, “Optimal

real-time pricing algorithm based on utility maximization for smart grid,” in 2010 First
IEEE International Conference on Smart Grid Communications. IEEE, 2010, pp. 415–420.

[119] A. Niromandfam, A. S. Yazdankhah, and R. Kazemzadeh, “Modeling demand response based

on utility function considering wind profit maximization in the day-ahead market,” Journal
of Cleaner Production, vol. 251, p. 119317, 2020.

[120] H. Zhong, L. Xie, and Q. Xia, “Coupon incentive-based demand response: Theory and case
study,” IEEE Transactions on Power Systems, vol. 28, no. 2, pp. 1266–1276, 2012.

[121] A. Sumaiti, S. R. Konda, L. Panwar, V. Gupta, R. Kumar, and B. K. Panigrahi, “Aggregated

demand response scheduling in competitive market considering load behavior through fuzzy
intelligence,” IEEE Transactions on Industry Applications, vol. 56, no. 4, pp. 4236–4247,
2020.

[122] A. Saha, “Expo-power utility: A ‘flexible’form for absolute and relative risk aversion,” Amer-
ican Journal of Agricultural Economics, vol. 75, no. 4, pp. 905–913, 1993.

[123] H. U. Gerber and G. Pafum, “Utility functions: from risk theory to finance,” North American
Actuarial Journal, vol. 2, no. 3, pp. 74–91, 1998.
References 162

[124] D. A. Wood and R. Khosravanian, “Exponential utility functions aid upstream decision
making,” Journal of Natural Gas Science and Engineering, vol. 27, pp. 1482–1494, 2015.

[125] F. P. Kelly, A. K. Maulloo, and D. K. H. Tan, “Rate control for communication networks:
shadow prices, proportional fairness and stability,” Journal of the Operational Research so-
ciety, vol. 49, no. 3, pp. 237–252, 1998.

[126] M. Miyake, “Logarithmically homogeneous preferences,” Journal of Mathematical Eco-

nomics, vol. 67, pp. 1–9, 2016.

[127] R. E. Kihlstrom, D. Romer, and S. Williams, “Risk aversion with random initial wealth,”
Econometrica: Journal of the Econometric Society, pp. 911–920, 1981.

[128] P. A. Diamond, “National debt in a neoclassical growth model,” The American Economic
Review, vol. 55, no. 5, pp. 1126–1150, 1965.

[129] D. T. Nguyen, M. Negnevitsky, and M. de Groot, “Modeling load recovery impact for demand
response applications,” IEEE Transactions on Power Systems, vol. 28, no. 2, pp. 1216–1225,
2012.

[130] P. Michel, E. Thibault, and J.-P. Vidal, “Intergenerational altruism and neoclassical growth
models,” 2004.

[131] M. S. Kimball, “Making sense of two-sided altruism,” Journal of Monetary Economics,

vol. 20, no. 2, pp. 301–326, 1987.

[132] L. A. Zadeh, “Zadeh, fuzzy sets,” Inform Control, vol. 8, pp. 338–353, 1965.

[133] A. Faruqui, D. Harris, and R. Hledik, “Unlocking the¿ 53 billion savings from smart meters
in the eu: How increasing the adoption of dynamic tariffs could make or break the eu’s smart
grid investment,” Energy Policy, vol. 38, no. 10, pp. 6222–6231, 2010.

[134] M. Zugno, J. M. Morales, P. Pinson, and H. Madsen, “A bilevel model for electricity retailers’
participation in a demand response market environment,” Energy Economics, vol. 36, pp.
182–197, 2013.

[135] M. Yu, S. H. Hong, Y. Ding, and X. Ye, “An incentive-based demand response (DR) model
considering composited DR resources,” IEEE Transactions on Industrial Electronics, vol. 66,
no. 2, pp. 1488–1498, 2018.

[136] M. J. Osborne et al., An introduction to game theory. Oxford university press New York,
2004, vol. 3, no. 3.

[137] G. L. Torres and V. H. Quintana, “An interior-point method for nonlinear optimal power
flow using voltage rectangular coordinates,” IEEE transactions on Power Systems, vol. 13,
no. 4, pp. 1211–1218, 1998.

[138] E. Hooshmand and A. Rabiee, “Energy management in distribution systems, considering

the impact of reconfiguration, ress, esss and dr: A trade-off between cost and reliability,”
Renewable energy, vol. 139, pp. 346–358, 2019.
References 163

[139] N. K. Meena, A. Swarnkar, N. Gupta, and K. R. Niazi, “Optimal integration of ders in

coordination with existing vrs in distribution networks,” IET Generation, Transmission &
Distribution, vol. 12, no. 11, pp. 2520–2529, 2018.

[140] S. Borenstein, “The long-run efficiency of real-time electricity pricing,” The Energy Journal,
vol. 26, no. 3, 2005.

[141] “YALMIP: A toolbox for modeling and optimization in MATLAB, author=Lofberg, Jo-
han,” in 2004 IEEE international conference on robotics and automation (IEEE Cat. No.
04CH37508). IEEE, 2004, pp. 284–289.

[142] A. Wächter and L. T. Biegler, “On the implementation of an interior-point filter line-
search algorithm for large-scale nonlinear programming,” Mathematical programming, vol.
106, no. 1, pp. 25–57, 2006.

[143] F. Facchinei, A. Fischer, and V. Piccialli, “On generalized Nash games and variational in-
equalities,” Operations Research Letters, vol. 35, no. 2, pp. 159–164, 2007.

[144] N. Mahmoudi, E. Heydarian-Forushani, M. Shafie-khah, T. K. Saha, M. H. Golshan, and

P. Siano, “A bottom-up approach for demand response aggregators’ participation in electric-
ity markets,” Electric Power Systems Research, vol. 143, pp. 121–129, 2017.

[145] B. Allen and M. Hellwig, “Bertrand-Edgeworth oligopoly in large markets,” The Review of
Economic Studies, vol. 53, no. 2, pp. 175–204, 1986.

[146] E. Maskin, “The existence of equilibrium with price-setting firms,” The American Economic
Review, vol. 76, no. 2, pp. 382–386, 1986.

[147] J.-S. Pang and M. Fukushima, “Quasi-variational inequalities, generalized Nash equilibria,
and multi-leader-follower games,” Computational Management Science, vol. 2, no. 1, pp.
21–56, 2005.

[148] A. A. Kulkarni and U. V. Shanbhag, “A shared-constraint approach to multi-leader multi-

follower games,” Set-valued and variational analysis, vol. 22, no. 4, pp. 691–720, 2014.

[149] R. Fletcher and S. Leyffer, “Solving mathematical programs with complementarity con-
straints as nonlinear programs,” Optimization Methods and Software, vol. 19, no. 1, pp.
15–40, 2004.

[150] N. K. Meena, A. Swarnkar, N. Gupta, and K. R. Niazi, “Multi-objective Taguchi approach

for optimal DG integration in distribution systems,” IET Generation, Transmission & Dis-
tribution, vol. 11, no. 9, pp. 2418–2428, 2017.
Publications

Following papers have been published/accepted/communicated out of this thesis work.

International Journals

1. V. C. Pandey, N. Gupta, K. R. Niazi, and A. Swarnkar, “An Economic Price based Demand
Response using Overlapping Generation Model in Distribution Systems,” Electric Power
Systems Research, vol. 213, pp. 108794, 2022.

2. V. C. Pandey, N. Gupta, K. R. Niazi, A. Swarnkar, and R. A. Thokar, “Modeling and As-

sessment of Incentive Based Demand Response Using Price Elasticity Model in Distribution
Systems,” Electric Power Systems Research, vol. 206, pp. 107836, 2022.

3. V. C. Pandey, N. Gupta, K. R. Niazi, A. Swarnkar and R. A. Thokar,“A Hierarchical

Price-Based Demand Response Framework in Distribution Network,” IEEE Transactions
on Smart Grid, vol. 13, no. 2, pp. 1151-1164, March 2022, doi: 10.1109/TSG.2021.3135561.

4. V. C. Pandey, N. Gupta, K. R. Niazi, A. Swarnkar, and R. A. Thokar, “An Adaptive Demand

Response Framework using Price Elasticity Model in Distribution Networks,” Electric Power
Systems Research, vol. 202, pp.107597, 2022.

5. V. C. Pandey, N. Gupta, K. R. Niazi, and A. Swarnkar, “A Bi-level Incentive Based Demand

Response Framework under Retail Competition in Distribution Systems (Under Review).

6. V. C. Pandey, N. Gupta, K. R. Niazi, and A. Swarnkar, “Modeling and Assessment of

Nonlinear Cardinal Utility Functions in Price Based Demand Response, (Under Review).

165
Brief bio-data

Vipin Chandra Pandey received the B. Tech in electrical engineer-

ing from Uttarakhand Technical University, Uttarakhand, India, in
2013, and the M. Tech in power systems from Malaviya National
Institute of Technology, Jaipur India, in 2016. He is currently
a Research Scholar in the Department of Electrical Engineering,
Malaviya National Institute of Technology, Jaipur, India. His re-
search interests include distribution system operation, artificial in-
telligence applications to power system optimization, demand re-
sponse and uncertainty modeling of distributed energy resources.

167