Electric Power Systems Research 72 (2004) 85–93

A zonal congestion management approach using ac

transmission congestion distribution factors
Ashwani Kumar a , S.C. Srivastava b , S.N. Singh b,∗
a Department of Electrical Engineering, National Institute of Technology, Kurukshetra, India
b Department of Electrical Engineering, Indian Institute of Technology, Kanpur 208016, India

Received 29 December 2003; received in revised form 10 March 2004; accepted 25 March 2004

Available online 15 June 2004

Abstract

In a deregulated electricity market, one of the major concerns of system operator (SO) is to ensure the free and fair electricity trading while
maintaining system security and stability in meeting the pool and contract demands. Achieving a commercially transparent and technically
feasible solution during transmission congestion, therefore, poses a great challenge to SO. Transmission congestion distribution factors (TCDF)
based on sensitivity of ac power flow in the lines due to the unit change in the power injection at the buses have been proposed by which the
congestion zones are identifies to reschedule the generators and loads in that zone for the congestion management. A conceptually reasonable
and computationally feasible approach for the solution of this problem has been developed and is illustrated on two test systems having both
pool and contracts loads.
© 2004 Elsevier B.V. All rights reserved.

Keywords: Transmission congestion management; Congestion zones; Transmission congestion distribution factors; Competitive electricity market

1. Introduction To manage the congestion in real time operations, nor-

mally, following methods are adopted:
Electric power systems, around the world, have been
forced to operate to almost their full capacities due to the 1. Use of available resources for congestion management
environmental and/or economic constraints to build new such as operation of FACTS controllers, rescheduling of
generating plants and transmission lines. The amount of generation based on minimum bids, etc.
electric power that can be transmitted between two loca- 2. Provide the timely information regarding the probability
tions through a transmission network is limited by security of having a particular line congested and economic in-
and stability constraints. Power flow in the lines and trans- centives to system users to adjust their requests and re-
formers should not be allowed to increase to a level where a main within the system constraints.
random event could cause the network collapse because of 3. Physically curtail the transactions.
angular instability, voltage instability or cascaded outages.
System operators (SO) always try to use first option, wher-
When such a limit reaches, the system is said to be con-
ever it is possible. Physical curtailment of loads is consid-
gested. Managing congestion to minimize the restrictions
ered as the last option for congestion management when it is
of the transmission network in the competitive market has,
impossible to wait for the system users to respond according
thus, become the central activity of systems operators. It
to economic criteria. However, the second option should be
has been observed that the unsatisfactory management of
developed for giving the system users sufficient information
transactions could increase the congestion cost which is an
regarding the congestion probability so that they can adjust
unwanted burden on customers [1].
their requests for system services and avoid congestion.
For different power market structures, the approach to
∗ Corresponding author. Tel.: +91 512 2590065; manage congestion may vary. Various congestion manage-
fax: +91 512 2590063. ment schemes for different restructuring paradigms have
E-mail address: [email protected] (S.N. Singh). been reported in literature. Hogan proposed the contract

0378-7796/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.epsr.2004.03.011
86 A. Kumar et al. / Electric Power Systems Research 72 (2004) 85–93

network and nodal pricing approach [2] using the spot pric- In this paper, transmission congestion distribution fac-
ing theory [3] for the pool type market. Chao and Peck [4] tors (TCDFs) based on ac power flow Jacobian sensitiv-
proposed an alternative approach which is based on parallel ity have been proposed. These TCDFs has been utilized
markets for link based transmission capacity rights and en- for identification of most sensitive congestion zones/clusters
ergy trading under a set of rules defined and administered for the given system. The effectiveness of proposed method
by the SO. These rules specify the transmission capacity has been demonstrated on 39-bus New England system and
rights required to support any bilateral transactions and 75-bus Indian system.
are adjusted continuously to reflect the changing system
conditions.
Several optimal power flow (OPF) based congestion man- 2. Transmission congestion distribution factors
agement schemes for multiple transaction systems have been
proposed. An approach using the minimum total modifica- 2.1. Mathematical formulation
tion to the desired transactions for relieving congestion was
presented in [5]. A variant of this least modification ap- Transmission congestion distribution factors are defined
proach [6] used a weighting scheme with the weights being as the change in real power flow in a transmission line-k
the surcharges paid by the transactions for transmission us- connected between bus-i and bus-j due to unit change in the
age in the congestion-relieved network. Marginal cost sig- power injection (Pi ) at bus-i. Mathematically, TCDF for
nals were used in [7] for generators to manage congestion. line-k can be written as:
A similar approach is proposed in [8], where the congestion Pij
cost is bundled with the marginal cost at each bus in pool TCDFki = (1)
Pi
model and a congestion cost minimization is adopted in bi-
lateral model. In [9], an OPF based approach that minimizes where Pij is the change in real power flow of line-k.
cost of congestion and service costs is proposed. In [10], a TCDFkn denotes that how much active power flow over a
new mechanism of congestion management in multilateral transmission line connecting bus-i and bus-j would change
transaction networks has been developed based on physical due to active power injection at bus-n. Three different meth-
flows. ods are considered for determination TCDFs are given as
The basic model for congestion management described follows.
in [6] was used in [11] incorporating the FACTS controllers
such as TCSC and SVC for congestion management. In 2.1.1. Method 1
[12], a sensitivity based approach for the optimal location The dc load flow based approach, which has been dis-
of unified power flow controller (UPFC) was proposed cussed in [14], is considered as method 1.
for the congestion management. In [13], an integrated For other methods, based on ac load flow Jacobian sensi-
technical-cum-market based framework has been used for tivity, TCDFs are derived below. The real power flow (Pij )
congestion management, which utilizes interruptible load in a line-k connected between bus-i and bus-j can be written
services as a tool for SO to provide transmission congestion as:
relief using congestion relief indices. Pij = Vi Vj Yij cos(θij + δj − δi ) − Vi2 Yij cos θij (2)
However, none of the methods discussed above provides
the strategies to SO regarding the specific generators to be where Vi and δi are the voltage magnitude and angle at bus-i.
rescheduled or the loads to be curtailed for the congestion Yij and θ ij are magnitude and angle of ijth element of Ybus
management. During the real time dispatch, the knowledge matrix.
of the group of system users, (generators and loads), which Using Taylor’s series approximation and ignoring higher
have similar effects on the transmission constraints is of order terms, Eq. (2) can be written as:
interest. ∂Pij ∂Pij ∂Pij ∂Pij
A congestion clusters based method that identifies the Pij = δi + δj + Vi + Vj . (3)
∂δi ∂δj ∂Vi ∂Vj
groups of system users, which have a similar effect on a
transmission constraint of interest has been proposed in [14]. Eq. (3) can be rewritten as:
These cluster based on congestion distribution factors are
Pij = aij δi + bij δj + cij Vi + dij Vj . (4)
termed as clusters of types 1, 2 and higher, where type 1
cluster represents users with strongest and non-uniform ef- The coefficients appearing in Eq. (4) can be obtained using
fects on transmission constraints of interest. The proposed the partial derivatives of real power flow Eq. (2) with respect
clustering based method has been used to create an efficient to variables δ and V as:
congestion management market, where the transactions in
the most sensitive cluster can help in eliminating conges- aij = Vi Vj Yij sin(θij + δj − δi ) (5)
tion. However, the method is based on dc load flow, which bij = −Vi Vj Yij sin(θij + δj − δi ) (6)
is based on the assumptions of lossless system with unit
voltage at all the buses. cij = Vj Yij cos(θij + δj − δi ) − 2Vi Yij cos θij (7)
 A. Kumar et al. / Electric Power Systems Research 72 (2004) 85–93 87

dij = Vi Yij cos(θij + δj − δi ). (8) From Eqs. (18) and (19), the following equation can be
obtained:
Two cases have been considered for determination of the −1
TCDFs using following Newton–Raphson Jacobian relation- P = J11 δ − J12 J22 J21 δ = Jred δ (20)
ship: where Jred is the reduced Jacobian matrix.





P δ J11 J12 δ From Eq. (20), the value of change in angle can be written
= [J] = . (9) as:
Q V J21 J22 V
δ = [Jred ]−1 P. (21)
2.1.2. Method 2
Eq. (21) can be written in the following form:
Neglecting coupling between P and V and between
Q and δ, Eq. (9) can be simplified as: n

δi = mil Pl i = 1, 2, . . . , n, i = s. (22)
P = [J11 ][δ] (10) l=1

Q = [J22 ][V ]. (11) Substituting Eq. (22) into Eq. (14), the change in power
flow is:
From Eq. (10), we get:  n n

δ = [J11 ]−1 [P] = [M][P]. (12) Pij = aij mil Pl + bij mjl Pl (23)
l=1 l=1
Eq. (12) can be written in the form: Eq. (23) can be rewritten as:
n

δi = mil Pl i = 1, 2, . . . , n, i = s (13) Pij = (aij mi1 + bij mj1 )P1
l=1 + · · · + (aij min + bij mjn )Pn . (24)
where n is the number of buses in the system and s is slack Therefore, the change in the power flow in the line-k con-
bus. necting the bus-i and bus-j is:
It is assumed that the impact of change in the voltage on
real power flow is negligible, and therefore, Eq. (4) can be Pij = TCDFk1 P1 + TCDFk2 P2 + · · · + TCDFkn Pn
written as:
(25)
Pij = aij δi + bij δj . (14)
where TCDFkn = aij min + bij mjn are the another set of ac
Substituting Eq. (13) into Eq. (14), we get: transmission congestion distribution factors corresponding
n n to bus-n for line-k.
 
Pij = aij mil Pl + bij mjl Pl (15)
l=1 l=1
2.2. Congestion zone based transmission management in
deregulated market

Pij = (aij mi1 + bij mj1 )P1 + (aij mi2 + bij mj2 )P2 The TCDFs obtained based on the methodology discussed
+ · · · + (aij min + bij mjn )Pn . (16) in Section 2.1 have been utilized for identifying different
congestion clusters (zones) for a given system. The conges-
Therefore, change in the real power flow can be written tion zone of type 1 is the one having large and non-uniform
as: TCDFs and the congestion zones of type 2 and higher order
have small or similar TCDFs. Therefore, the transactions in
Pij = TCDFk1 P1 + TCDFk2 P2 + · · · + TCDFkn Pn the congestion zone 1 have critical and unequal impact on
(17) the line flow. The congestion zones of types 2, 3 and higher
are farther from the congested line of interest [14]. There-
where TCDFkn = aij min + bij mjn are the transmission con- fore, any transaction outside the most sensitive zone 1 will
gestion distribution factors corresponding to bus-n for line-k contribute very little to the line flow. Thus, the identifica-
connecting bus-i and bus-j. tion of congestion zones will reduce the effort of SO in se-
lecting the participants for congestion management and that
2.1.3. Method 3 will also reduce the computation burden.
Assuming that reactive power injections at all the buses The congestion zone based method is also applicable if
are constant, Eq. (9) can be written as: more than one transmission line get congested simultane-
P = J11 δ + J12 V (18) ously. The congestion clusters/zones for a multi-congestion
case can be obtained by superimposing the clusters’ zones
0 = J21 δ + J22 V. (19) obtained for each of the congested lines separately.
88 A. Kumar et al. / Electric Power Systems Research 72 (2004) 85–93

2.2.1. Spot market for transmission congestion ing more than one supplier and/or one consumer can be ex-
management pressed as:
Pool dispatch is practiced in several countries and stands  
k k
at the core of all deregulated markets so far. Essentially, PGm − PDn = 0 k = 1, 2, . . . , tk (27)
it exists with a short-term spot market usually involving a m n

day-ahead dispatch and scheduling arrangements. The bi- where PGm and PDn stand for the injections into the bus-m
lateral market may also exist as well, but this does not ex- and taken out at bus-n. tk is the total number of contracts.
plicitly affect the operations of the SO or the “Pool Rules”
which stipulate how the SO should function. 2.2.3. Re-dispatch for the congestion management
Thus, the use of the proposed congestion zones would be After the bids are received from the market participants,
to have separate spot market for the transmission conges- the market coordinator will select certain bids necessary to
tion management [14]. It is assumed to have separate bids facilitate pool market along with the bilateral or multilateral
by the system users, one into the electricity market, and transactions. A transmission market provider finds the solu-
other into the transmission management market. As done in tion to this problem by solving the following optimization
the most of the restructured electric power utilities, all bal- problem:
anced bilateral or multilateral transactions requests are sub-
Ntr

mitted for approval to the transmission market coordinator.
A separate biding process for participating in the transmis- Minimize Cl (Pl )Pl (28)
l
sion congestion management market takes place in parallel.
It is also assumed that the system provider will post on-line Subject to
likely congestion clusters/zones based on requested transac-
Ntr

tion and manage congestion from the groups of participants,
Ntr based on the bid cost function Cl (Pl ) for changing the ((TCDFkl )Pl ) + Fk0 ≤ Fkmax k = 1, 2, . . . , nl (29)
l=1
injection into the bus-l at a price Cl in pool and mix of bi-
lateral and multilateral contracts. Plmin ≤ Pl ≤ Plmax l = 1, 2, . . . , Ntr (30)
2.2.2. Modeling of bilateral or multilateral contracts Ntr
 Ntr
 ∂PL
The conceptual model of bilateral dispatch is that sell- Pl − Pl = 0 (31)
ers and buyers enter into transactions where the quantities ∂Pl
l l
traded and the trade prices are at the discretion of these par-
PGm − PDn = 0 (32)
ties and not the SO. These transactions are then brought to
the SO with a request that transmission facilities for the rel-  
k k
PGm − PDn =0 k = 1, 2, . . . , tk (33)
evant amount of power transferred to be provided. If there m n
is no static and dynamic security violation, the SO simply
dispatches all requested transactions and charges for the ser- where Pl is the real power adjustment at bus-l and Cl
vice. are the incremental or decremented price bids submitted by
In practical system, not all the generators have bilateral generators or by interruptible loads. These are the prices
contract with load bus-i and similarly not all loads will have at which generators are willing to adjust their real power
bilateral contract with generator-j. Mathematically, each bi- outputs or loads are willing to adjust the power demand.
lateral transaction between power seller at bus-m and power Fk0 is the power flow caused by all contracts requesting the
purchaser at bus-n is of the following form: transmission service. Fkmax is the line flow limit of the line
connecting bus-i and bus-j. PL is the total loss obtained from
PGm − PDn = 0. (26) an exact loss formula [16] in order to incorporate the change
in the losses in the system occurring due to re-dispatch of
Furthermore, the amount of generation from generator j generators, or curtailment/increment in the interrupted loads.
scheduled for bilateral contracts is within certain range from The number of participants in the transmission congestion
its maximum generating capacity. The bilateral concept can market, Ntr will be generally much smaller than the number
be generalized to be multi-node case where the seller, for of active participants in the electricity market. In the pool
example, a generator company, may inject power at several based market structure, the optimization model will not con-
nodes and the buyer also draw load at several nodes. Unlike tain Eqs. (32) and (33).
pool dispatch, there will be a transaction power balance in
that the aggregate injection equals the aggregate draw off
for each transaction. 3. Numerical examples
The contracted demands of load buses to be provided by
generator-m have to be shared in the proportion decided op- The effectiveness of the proposed concept of conges-
timally by SO. Mathematically, multilateral contract involv- tion zone/cluster based congestion management system
 A. Kumar et al. / Electric Power Systems Research 72 (2004) 85–93 89

Fig. 1. Congestion zones for 39-bus New England system for line 34-14.

utilizing the TCDFs is illustrated on 39-bus New Eng- 3.1. The 39-bus New England system
land system [17] and 75-bus Indian system [18]. The
congestion zones are decided on the basis of TCDFs and 3.1.1. Single congestion case
accordingly the most sensitive congestion zone has been The 39-bus New England system is a simplified represen-
selected by SO for managing congestion. The most sen- tation of the 345 kV transmission system in the New England
sitive zone 1 is a cluster of buses having TCDFs, which region which consists of 10 generators and 29 load buses.
are highly non-uniform and large in magnitude. The con- For this system, the congestion zones based on TCDFs ob-
gestion zones of order 2 and higher comprise of buses tained using method 3 for a congested line connected be-
with TCDFs of smaller magnitude with almost uniform tween bus-34 and bus-14 are shown in Fig. 1. The TCDFs
variation. The proposed congestion management approach corresponding to each bus are given in Table 1 for differ-
(Eqs. (28)–(33)) has been solved using GAMS optimization ent zones. The zones obtained using method 2 consisted of
tool [15]. the same set of buses as with method 3 although the TCDF
The results of congestion management in terms of con- values were different.
gestion costs and rescheduling of generator outputs for It is observed from Table 1 that the zone 1 is the most sen-
pool market, pool plus bilateral/multilateral market struc- sitive zone with larger magnitude and strongest non-uniform
tures have been determined using the following three distribution of TCDFs. The magnitudes of TCDFs in zone 4
methods: are higher than zones 2 and 3 but the distribution of TCDFs
Method 1: Utilizing TCDFs obtained with dc load are uniform.
flow model [14] The congestion cost determined for pool based mar-
Method 2: Utilizing TCDFs obtained with ket model with both bilateral and multilateral con-
decoupled ac load flow model as defined tracts using TCDFs obtained from different methods
in Eqs. (16) and (17) are shown in Fig. 2. Generators 3, 8, and 10 were se-
Method 3: Utilizing TCDFs obtained with coupled lected by SO for congestion management based on
ac load flow model as defined in their qualifying bids to reschedule their generation.
Eqs. (24) and (25) The different combinations of bilateral and multilat-
The test results on the two systems are given below. eral contracts considered along with the pool model are

Table 1
Zones (1–4; bus no., TCDF) for IEEE 39-bus system for congested line 34-14
1 2 3 4

1, −0.005 31, 0.131 2, −0.035 17, −0.023 9, 0.029 4, 0.153 23, 0.154
3, 0.202 32, 0.195 8, −0.020 18, −0.031 26, 0.027 5, 0.153 24, 0.152
13, −0.069 33, 0.260 10, −0.041 19, −0.033 27, 0.049 6, 0.154 30, 0.154
14, −0.256 34, 0.421 11, −0.037 25, −0.022 28, 0.028 7, 0.154 36, 0.152
15, −0.045 35, 0.232 12, −0.041 29, 0.028 21, 0.153 39, 0.153
16, −0.005 37, 0.076 22, 0.154
20, 0.196 38, 0.020
90 A. Kumar et al. / Electric Power Systems Research 72 (2004) 85–93

Method 1 Method 2 Method 3 G3 G8 G10

5000 0.6
Congestion Cost $/h

0.4
Rescheduled Gen. in

4000
0.2
3000
0
p.u.

2000 P C1 C2 C3 C4 C5
-0.2
1000 -0.4
Pool model with contracts
0 -0.6
P C1 C2 C3 C4 C5
Pool model with contracts -0.8

Fig. 2. Congestion cost for 39-bus system. Fig. 4. Rescheduled generation for 39-bus (method 2).

0.8 G3 G8 G10 G3 G8 G10

0.6
0.6
Rescheduled Gen. in

0.4 0.4
Rescheduled Gen. in p.u.

0.2
0 0.2
p.u.

-0.2 P C1 C2 C3 C4 C5
0
-0.4
-0.6 P C1 C2 C3 C4 C5
-0.2
-0.8 Pool model with contracts
-1 -0.4
Pool model with contracts
Fig. 3. Rescheduled generation for 39-bus (method 1). -0.6

-0.8
as follows: Fig. 5. Rescheduled generation for IEEE 39-bus (method 3).
P: Pool model only
C1: Pool model with one bilateral contract congestion costs shown in this figure are the optimal value of
between buses 3–35 the objective function given in Eq. (28). Hence, the ac load
C2: Pool model with two bilateral contracts flow based methods for congestion management provide
between buses 3–35 and 8–21 more economical solution as compared to the dc load flow
C3: Pool model with one multilateral contract based method. It is due to the fact that in ac method the gen-
between buses 3–35, 36 eration rescheduling is found to be smaller than dc method.
C4: Pool model with one bilateral contract Further, dc load flow based method (method 1) involves as-
between bus 8-21 and multilateral contract sumptions such as it neglects the losses in the system and the
between 3–35, 36 voltages are assumed to be 1.0 pu at all the buses, whereas ac
C5: Pool model with two bilateral contracts load flow based method considers the losses and the voltage
between buses 8–21, 10–35 and multilateral variations and gives more accurate values of TCDFs used in
contract between buses 3–35, 36 the optimization model for congestion management. In [19],
From Fig. 2, it can be observed that the congestion costs a similar set of sensitivity factors termed as power transfer
are found to be quite less in methods 2 and 3 as compared to distribution factors (PTDFs) were derived based on dc load
method 1 for all types of market models taken for study. The flow as well as ac load flow models. It was demonstrated

Fig. 6. Congestion zones for a multi-congestion case for 39-bus New England system.
 A. Kumar et al. / Electric Power Systems Research 72 (2004) 85–93 91

Table 2
Rescheduled generation (in pu) and congestion cost for different cases of transactions
P C1 C2 C3 C4 C5

G3 −0.8000 −0.0209 −0.2222 −0.0228 −0.2564 −0.2976

G4 0.5000 0.2000 0.2000 0.2000 0.2000 0.2000
G6 −0.3358 −1.0000 −1.0000 −1.0000 −1.0000 −1.0000
G8 0.2502 0.4442 0.5278 0.4482 0.5624 0.3009
G10 0.4000 0.4000 0.2000 0.4000 0.2000 0.4000
Cost ($/h) 5390.2000 5084.6900 5438.8100 5101.1800 5628.1500 5346.8400

Fig. 7. Congestion zones for 75-bus Indian system for line 26-41.

through numerical example that the ac-PTDFs are accurate multi-congestion case has been determined selecting two
as compared to the dc-PTDFs. Out of the two ac load flow new generators 4 and 6 from the sensitive zone 2 to resched-
based methods, the results for the method 3 are more eco- ule their outputs for the congestion management. These are
nomical as the coupling between P–V and Q–δ has given in Table 2. It is evident from Table 2 that congestion
been considered, which is neglected in method 2. cost increases if more than one line gets congested in the
Figs. 3–5 show the values of rescheduled generation for network.
pool model with both bilateral and multilateral contracts
analyzed with different methods. 3.2. The 75-bus Indian system
From Figs. 3–5, it is observed that the value of the resched-
uled generation for ac based method is less than the dc based This system represents reduced network of one of the
method for the pool market as well as pool market with bi- State Electricity Boards and contains 15 generators, 24
lateral and multilateral transactions. transformers, and 97 lines [14]. For this system, congestion
zones based on magnitude and non-uniformity of TCDFs
3.1.2. Multi-congestion case
Method1 Method 2 Method 3
The proposed zones/clusters based method for conges- 6000
tion management is also applicable if more than one trans-
Congestion Cost $/h

5000
mission line gets congested simultaneously in the system.
4000
The zones for a multi-congestion case has been obtained by
superimposing the clusters/zones obtained for each of the 3000
congested lines individually. For the IEEE 39-bus system, a 2000
multi-congestion case has been studied considering conges- 1000
tion of lines 34–14 and 36–21, simultaneously. The conges- 0
P C1 C2 C3 C4 C5 C6
tion clusters for this case are shown in Fig. 6.
Pool model with contracts
In this case, the zones 1 and 2 are the most sensitive
zones. The rescheduled generation and congestion cost for Fig. 8. Congestion cost for75-bus Indian network.
92 A. Kumar et al. / Electric Power Systems Research 72 (2004) 85–93

G12 G3 G13

1
Resceduled Gen. in

0.5

0
p.u.

-0.5
P C1 C2 C3 C4 C5 C6

-1
Pool Model with Contracts
-1.5

Fig. 9. Rescheduled generation for 75-bus Indian network (method 1).

for congestion in lines 26–41 were obtained which are G12 G3 G13
shown in Fig. 7. Generators 3, 12, and 13 from zone-1 were 1
selected to participate in the congestion management based 0.5
Rescheduled Gen. in

on their qualifying bids in the market.

Fig. 8 shows the congestion costs for the following types 0
of market models considered for study using the three dif- P C1 C2 C3 C4 C5 C6
p.u.

-0.5
ferent methods of analysis.
-1
P: Pool model only Pool model with contracts
C1: Pool model with one bilateral contract between -1.5
buses 3–27
Fig. 10. Rescheduled generation for 75-bus Indian network (method 2).
C2: Pool model with two bilateral contracts
between buses 3–27 and 12–24
C3: Pool model with three bilateral contracts G12 G3 G13
between buses 3–27, 12–24, and 13–55 1

C4: Pool model with one multilateral contract

Rescheduled Gen. in

0.5
between buses 12–27, 73
0
C5: Pool model with one bilateral contract between
p.u.

P C1 C2 C3 C4 C5 C6
buses 13–55 and one multilateral contract -0.5
between buses 12–27, 73
-1
C6: Pool model with two bilateral contracts Pool model with contracts
between buses 3–24 and 13–55 and one -1.5
multilateral contract between buses 12–27, 73
Fig. 11. Rescheduled generation for 75-bus Indian network (method 3).
From Fig. 8, it is observed that the congestion costs are
quite less with methods 2 and 3 as compared to method 1 for
times were recorded on Pentium IV, 1.9 GHz, 512 kB RAM
all types of market models taken for study. Out of the two
computer. The dc based method (method 1) took 0.19 and
ac load flow based methods, the congestion costs are lower
10.19 s as total solution time including 0.05 and 0.18 s for
with method 3 as compared to the method 2 for all types of
the calculation of the TCDFs for 39-bus and 75-bus systems,
market models. Hence, the ac load flow based method for
respectively. The ac load flow based methods (methods 2
congestion management is more economical as compared to
and 3) took total time of about 0.24 and 11.41 s including
dc load flow based method for congestion management.
the 0.08 and 0.23 s for calculation of TCDFs for 39-bus and
Figs. 9–11 show the values of rescheduled generation for
75-bus systems, respectively. Thus, the CPU time required
different types of models analyzed with different methods.
by the proposed ac load flow sensitivity based method is
It is observed that the value of the rescheduled generation
quite close to that with the dc load flow based method.
with the ac based methods is less than the dc based method
for the pool market as well as pool market with bilateral and
multilateral contracts.
4. Conclusion

3.3. Computational time This paper proposes a simple and efficient congestion
management method based on congestion zones/clusters ob-
In order to compare the computational time required by tained with new set of transmission congestion distribution
the proposed ac load flow based methods (methods 2 and 3) factors derived from the ac load flow Jacobian sensitivity.
and the exiting dc load flow based method (method 1), CPU The ac load flow based methods provide more economical
 A. Kumar et al. / Electric Power Systems Research 72 (2004) 85–93 93

solution for the congestion management as compared to the [8] H. Singh, S. Hao, A. Papalexopoulos, Transmission congestion man-
dc load flow based method proposed in reference [14]. The agement in competitive electricity markets, IEEE Trans. Power Syst.
13 (2) (1998) 672–680.
proposed ac load flow sensitivity based methods are more [9] Fu. Jian, John W. Lamont, A combined framework for service identi-
accurate as compared to the dc load flow based method, as fication and congestion management, IEEE Trans. Power Syst. 16 (1)
established in [19]. The computational time required by the (2001) 56–61.
proposed ac load flow method is quite close to the dc load [10] S. Tao, G. Gross, A congestion-management allocation mechanism
flow method. The method is efficient for on-line implemen- for multiple transaction network, IEEE Trans. Power Syst. 17 (3)
(2002) 826–833.
tation of congestion management as the proposed TCDFs [11] S.C. Srivastava, Perveen Kumar, Optimal power dispatch in dereg-
can be updated quite fast utilizing only base case load flow ulated market considering congestion management, in: Proceedings
Jacobian sensitivity. The market administrator can post the of the Conference on Electric Utility Deregulation and Restructuring
information of congested zones/clusters so that the market and Power Technologies, DRPT, 2000a, City University, London,
participants can bid and adjust their outputs for congestion 4–7 April 2000, pp. 53–59.
[12] S.N. Singh, A.K. David, Optimal location of FACTS devices for
management, accordingly. congestion management, Int. J. Electric Power Syst. Res. 58 (2001)
71–79.
[13] L.A. Tuan, K. Bhattacharya, Interruptible load services for transmis-
References sion congestion relief, in: Proceedings of the 14th Power System
Computational Conference (PSCC,02), Sevilla, Spain, 24–28 June
[1] Federal Energy Regulatory Commission (FERC), “Regional Trans- 2002.
mission Organizations,” Washington, DC, Docket RM99-2-000, Or- [14] C.-N. Yu, M. Ilic, Congestion clusters-based markets for transmission
der 2000, 20 December 1999. management, in: Proceedings of the IEEE Power Engineering Society
[2] W.W. Hogan, Contract networks for electric power transmission, J. Winter Meeting, New York, NY, January 1999, pp. 1–11.
Regul. Econ. 4 (1992) 211–242. [15] GAMS Release 2.50, A User’s Guide, GAMS Development Corpo-
[3] F.C. Schweppe, M.C. Caramanis, R.D. Tabors, R.E. Bohn, Spot ration, 1999.
pricing of Electricity, Kluwer, Norwell, MA, 1988. [16] O.I. Elgerd, Electric Energy Systems Theory: An Introduction, Tata
[4] H. Chao, S. Peck, A market mechanism for electric power transmis- McGraw Hill Publishing Company Ltd., New Delhi, 1989.
sion, J. Regul. Econ. 10 (1996) 25–59. [17] K.R. Padiyar, Power System Dynamics: Stability and Control, Wiley,
[5] F.D. Galiana, M. Ilic, A mathematical framework for the analysis Singapore, 1996, p. 601.
and management of power transactions under open access, IEEE [18] S.N. Singh, S.C. Srivastava, Corrective action planning to achieve
Trans. Power Syst. 13 (No. 2) (1998) 681–687. optimal power flow solution, in: Proceedings of the IEEE, vol. 142,
[6] R.S. Fang, A.K. David, Transmission congestion management in an pt. C, November 1995, pp. 576–582.
electricity market, IEEE Trans. Power Syst. 4 (3) (1999) 877–883. [19] A. Kumar, S.C. Srivastava, AC power transfer distribution factors
[7] H. Glavisch, F. Alvarado, Management of multiple congested condi- for allocating power transactions in a deregulated environment, in:
tions in unbundled operation of power systems, IEEE Trans. Power Proceedings of the IEEE Power Engineering Review, July 2002,
Syst. 13 (2) (1998) 1013–1019. pp. 42–43.