Detecting Intelligent Load Redistribution Attack Based on Power Load Pattern Learning in Cyber-Physical Power Systems

The massive integration of advanced cyber technologies with power grids has expanded the attack surface area of cyber-physical power systems (CPPSs), where timely detection is of paramount importance for their safe and reliable operation. However, most studies on securing CPPS rarely considered resource constraints, resulting in the unsatisfactory performance of security solutions in practical applications. To gain insight into attack behavior in realistic scenarios, this article fully develops the concept of load redistribution (LR) attacks and designs an intelligent version that considers both concealment property and resource limitation. Then, aiming to develop an effective countermeasure, this article provides a novel attack detection method based on power load pattern learning, which consists of a power load predictor and subsequent attack detector, to determine the existence of intelligent LR attacks. Specifically, a multichannel power load predictor based on the SARIMAX model is proposed to capture both temporal and spatial correlations of power load data for accurate prediction. Using augmented features, a dictionary-learning-based attack detector that can handle the class imbalance problem by unsupervised learning is capable to detect the intelligent LR attacks. Finally, experiments on numerical simulation and the CPPS-HITL platform are conducted to verify the effectiveness and practical availability of the proposed method.