PREDICTIVE AND OPTIMIZATION BASED CONTROL OF SMART GRID FOR CYBER SECURITY USING
ARTIFICIAL INTELLIGENCE
NWOGWU KELECHI JUSTICE
ESUT/PG/PHD/2024100000664
DEPARTMENT OF ELECTRICAL AND ELECTRONIC ENGINEERING
FACULTY OF ENGINEERING
ENUGU STATE UNIVERSITY OF SCIENCE AND TECHNOLOGY
(ESUT), ENUGU
PROJECT SUPERVISOR: PROF. ILO F.U
1
JULY 2025
Motivation
The global transition toward a sustainable energy system has catalyzed the evolution of traditional power grids into smart
grids - digitally enhanced infrastructure that integrates advanced communication, automation and control technologies.
While smart grids promise improved efficiency, reliability, environmental performance and efficient energy
management, however, their reliance on interconnected digital systems introduces a complex array of cyber security
vulnerabilities that must be addressed to ensure operational resilience and public safety. (Yordan 2023)
The rapid digital transformation of smart grids has introduced unprecedented levels of connectivity and automation,
increasing their vulnerability to cyber threats such as data breach, ransom ware attacks and grid manipulation. Traditional
cyber security measures struggle to keep pace with these evolving threats, making Ai-driven solutions essential for real
-time detection, mitigation and resilience. Ai-powered security mechanism enables predictive threat analysis, anomaly
detection, and automated incident response, strengthening the grid's ability to withstand sophisticated cyber attacks.
2
1.1 Statement of problems.
The paradigm shift from traditional power grid to smart grid has paved ways for cyber attacks which has several consequences.
The most significant problems are described below which this research work tends to solve.
1. Increasing Cyber Threats in Smart Grids
. With the growing integration of digital communication in power system, smart grids are becoming increasingly vulnerable
to cyber attacks such a data manipulation, denial of service and unauthorized access.
2. Lack of predictive threat detection
3
. Existing control systems in smart grids often operate re-actively. These is a critical gap in predictive models that can foresee
security breaches before they occur and initiate action.
3. Inefficient optimization in real-time control
. Traditional optimization techniques struggle to balance operational efficiency, energy demand and cyber resilience,
especially in real-time multi-agent environment.
4. Insufficient integration of intelligent agent.
. While intelligent agents have potential to autonomously monitor , adapt and respond to threats, their adoption in smart grid
security is still limited and lacks a standardized framework.
5. Interoperability and scalability issues.
. Smart grids consists of heterogeneous components and vast data streams. Implementing predictive optimization using
intelligent agents must overcome difficulties in system scalability, protocol capability and data integration.
4
1.3 Aim /Purpose of the Study
The aim of this research work predictive and optimization based control of smart grid for cybersecurity using artificial
intelligence.
1.4 Research Objectives
The research objectives of this thesis is set out as follows;
i. To review related current literature on predictive and optimization based control of smart grid for cybersecurity using
intelligent agent.
ii. To characterize the existing cyber threats and vulnerabilities in smart grid infrastructure.
iii. To develop predictive models using artificial intelligence for early detection of cyber-attacks
iv. To design and implement optimization-based control strategies for mitigating cyber risks.
v. To integrate predictive analytic with smart grid control systems for proactive decision-making
5
vi. To evaluate the performance of the proposed AI-based control framework under various cyber-attack scenarios.
vii. To enhance grid resilience and reliability through adaptive and intelligent control mechanisms.
viii. To propose a scalable and secure architecture for AI-integrated smart grid systems.
1.6 Significance/Impacts of the Study
The study has a significant and far-reaching technological, operational, economic and societal impacts. Below is the detailed
breakdown of the significance and impacts.
1. Enhanced Cybersecurity
Proactive threat detection: AI can predict cyber threats before they occur by analyzing patterns, anomalies, and historical data.
Real-time response: Predictive and optimization-based AI models allow the system to respond to cyber intrusions in real time,
reducing damage.
Attack mitigation: It can reduce the risk of cascading failures in the smart grid caused by cyberattacks.
2. Smarter Grid Operations
6
Improved decision-making: AI enables data-driven control strategies that adapt to real-time grid conditions and threats.
System resilience: Optimized control can help maintain grid stability and reliability even under cyber threats or uncertain
operating conditions.
Automation: Predictive models reduce human error by automating monitoring and control tasks.
3. Efficient Resource Management
Optimal power flow and load balancing: AI can optimize the distribution of energy across the grid, even during an attack or
disturbance.
Reduced energy waste: Enhances overall efficiency of power systems by minimizing losses and ensuring better demand-side
management.
4. Integration of Renewable Energy Sources
AI improves control and security in a grid with high penetration of renewable energy, which is often decentralized and more
vulnerable to cyber threats.
Predictive control helps to forecast energy generation (like solar or wind), allowing better grid planning and protection.
7
5. Economic and Operational Benefits
Reduced downtime and outages caused by cyberattacks or mismanagement.
Cost savings due to reduced need for manual intervention and more efficient operations.
Long-term sustainability through optimized grid performance.
6. National Security and Critical Infrastructure Protection
The smart grid is a part of critical national infrastructure; securing it with AI-based predictive control is vital for national
security.
Helps in compliance with regulatory standards and strengthens the national cyber defense posture.
7. Societal Impact
Ensures uninterrupted power supply to essential services (hospitals, communication, transportation).
Builds public trust and confidence in smart grid technologies and AI applications.
8. Academic and Technological Advancement
8
Advances the body of knowledge in:
Cyber-physical systems
Smart grid optimization
AI for control and security
Encourages interdisciplinary research between electrical engineering, computer science, and data science.
Scope of the study
The study focuses on modern electrical grids that integrates Advanced communication networks, Distributed energy resources
(DERs), Smart sensors/meters and Automated control systems.
Research Questions
1. How can artificial intelligence be used to predict cyber threats in smart grid systems?
9
2. What optimization techniques can be integrated with AI to improve the resilience of smart grid control systems against
cyber attacks?
3. How effective are AI-based predictive control methods compared to traditional control strategies in securing smart
grids?
4. What are the most common cyber-physical vulnerabilities in smart grids, and how can AI mitigate their impact?
5. Can real-time optimization and prediction using AI enhance the operational reliability of smart grids under cyber
threat conditions?
LITERATURE REVIEW
10
Tabular Summary of References
Technique Methods &
# Author(s) Title Work Done Research Gap
s Used Materials
Autoencoder
Federated
Disturbance FL with Deploy on
Husnoo et learning +
1 FedDiSC vs attack DP-SIGNSG edge devices;
al. (2023) autoencode
detection D, ICS latency in FL
r
datasets
Optimal
Extending to
Liu et al. DRL for DQN + attack Polish grid
2 other RES and
(2020) Wind-EPS CVSS policy test case
uncertainty
modeling
RL-POM Model-free Online Simulation
Kurt et al. Robustness to
3 DP attack RL anomaly on SG
(2018) unseen attacks
detection (POMDP) detection testbeds
DL Taxonomy
Autoencod Real-world DL
Abdi et al. proactive & Literature
4 ers, CNN, deployment
(2024) defense benchmarki review
MTD gaps
survey ng
11
Technique Methods &
# Author(s) Title Work Done Research Gap
s Used Materials
ML CNN- Anomaly
ScienceDi Real-time
techniques LSTM, detection Datasets
5 rect deployment,
for SVM, performanc analysis
(2023) transparency
SG-CPS hyperopt e
PSO-
Bitirgen optimized PSO + False data
IEEE test Scalability and
6 & Filik CNN- CNN- injection
systems generalization
(2023) LSTM for LSTM detection
FDIA
Cheetah-P Hierarchical
Mohamed Hardware/soft
SO Hybrid microgrid Microgrid
7 et al. ware
microgrid PSO security simulations
(2024) integration
control control
PMU
PMU Dueling Secure Evaluate
Cali et al. streams in
8 anomaly AI, PMU PMU-based latency and
(2024) cyber-physic
detection data detection false alarms
al SOC
Edge Edge-AI, Real-time
Mdpi Architecture IoT-SG
9 Intelligenc offloading offloading
(2021) & frameworks
e survey models under attack
12
Technique Methods &
# Author(s) Title Work Done Research Gap
s Used Materials
cybersecurit
y review
Applied EV,
Review of
Mdpi control & Control + security, Cross-sector
10 energy sector
(2023) AI AI maintenanc applicability
use
overview e
Converter
Converter Attack
Search0 -driven SAE + 123-bus
11 instability resilience in
(2023) stability Deep RL IEEE system
detection variable grids
AI
Multi-
Distribute Vulnerabilit Autonomous
Hug agent Microgrid,
d energy, y control-
12 (2018) & control + PMU, attack
cyber- quantificati integrated
others attack points
attacks on cybersecurity
modeling
13
7. Research Gaps and Challenges
Despite progress, several gaps persist:
Data Scarcity: Limited public datasets for cyberattack modeling in smart grids
Scalability: Many models struggle with high-dimensional, real-time data processing
Generalization: AI models may not adapt well to novel or sophisticated attack strategies
Interpretability: Black-box models lack transparency, hindering trust in high-risk environments
Future research should explore federated learning, transfer learning, and robust validation frameworks to enhance
model generalizability and security guarantees.
14
REFERENCES
Y. Zhang, L. Wang, and J. Liu, ―A Comprehensive Review of Smart Grid Technologies and Their Applications,‖ IEEE
Access, vol. 12, pp. 11234-11250, 2024. doi: 10.1109/ACCESS.2024.3149234.
S. Kumar, A. Ghosh, and R. Kumar, ―Advanced Metering Infrastructure: A Survey on Recent Developments and Future
Directions,‖ IEEE Transactions on Smart Grid, vol. 15, no. 2, pp. 874-890, 2023. doi: 10.1109/TSG.2023.3145821.
H. Lee and M. Choi, ―Integration of IoT and AI in Smart Grids: A Survey,‖ IEEE Internet of Things Journal, vol. 10, no.
5, pp. 4672-4684, 2023. doi: 10.1109/JIOT.2023.3223096.
M. Patel, R. Singh, and A. Sharma, ―Real-Time Monitoring and Control in Smart Grids Using IoT Technologies,‖ IEEE
Journal on Selected Areas in Communications, vol. 41, no. 4, pp. 1018-1030, 2023. doi: 10.1109/JSAC.2023.3234521.
J. Chen, L. Zhang, and C. Liu, ―Predictive Maintenance for Smart Grid Systems: Techniques and Applications,‖ IEEE
Transactions on Industrial Informatics, vol. 20, no. 3, pp. 1250-1262, 2023. doi: 10.1109/TII.2023.3269875.
15
K. Lee and D. Park, ―Optimization of Energy Distribution in Smart Grids Using Machine Learning,‖ IEEE Transactions on
Power Systems, vol. 38, no. 2, pp. 1267-1280, 2023. doi: 10.1109/TPWRS.2023.3248650.
16
