A Deep Learning-Based Cyber Intrusion Detection and Mitigation System for Smart Grids

The emergence of power system digitalization initiatives is revolutionizing the way electricity grids are monitored and protected. However, the integration of cyber and physical electrical infrastructures leads to an increase in the risk of cyber intrusions. Attackers can gain access to the smart grid and inject falsified data, leading the protection schemes to activate unnecessary power outage actions. Such outages can be devastating to end users. In this paper, an intrusion detection and mitigation system (IDMS) is proposed using deep learning neural networks (DLNNs) to detect, classify, and locate intrusions in smart grids. Once the disturbance is detected, the IDMS is designed to diagnose the intrusion and classify the attack into a single point or coordinated intrusion. Afterward, the algorithm locates and isolates the contaminated intelligent electronic device (IED) and predicts its current waveform utilizing long short-term memory (LSTM) to maintain power system observability. The proposed IDMS performs the required diagnosis on the modified IEEE 13-bus system. Simulation results demonstrate high accuracy in the proposed detection, classification, location, and prediction approach.