Deep Learning-Based Cyber Attack Detection in Power Grids with Increasing Renewable Energy Penetration

As renewable energy sources are integrated into power grids worldwide, the vulnerability to cyber-attacks increases, necessitating robust detection mechanisms to ensure system integrity and stability. In this paper, we propose a novel approach for cyber-attack detection in power grids, specifically based on the conditions of the Sri Lankan power system. Using the concept of wide area network monitoring and utilizing Ceylon Electricity Board generation data, as well as a PSCAD model representing real-time solar farms, synthetic datasets resembling actual demand-supply curves are generated. By analyzing well-known cyber-attack methodologies such as fault data injection, replay attacks, man-in-the-middle, and spoofing, we create attack datasets. Subsequently, we train neural network models including Convolutional Neural Networks (CNNs), Transformer models, and Long Short-Term Memory (LSTM) networks. Through comprehensive comparative analysis of model effectiveness using various parameters, our objective is to swiftly identify cyber-attacks as they occur within the system. The proposed methodology aims to address two primary objectives of cyber-attacks on power grids: energy theft and destabilization. By preemptively detecting and mitigating such attacks, the integrity and stability of the power grid can be safeguarded effectively. Our research contributes to the advancement of cyber-security measures in power systems, particularly in regions experiencing increased penetration of renewable energy sources like Sri Lanka.