An Ensemble Learning-Based Cyber-Attacks Detection Method of Cyber-Physical Power Systems

The cyber-physical power system (CPPS) is evolved from the traditional power system by the deployment of information networks with control systems, computational units, and communication networks. However, their integration into CPPS introduces new challenges in maintaining security. Various malicious cyber-attacks are drawn much attention recently after the discovery of the impact on CPPS state estimation. Thus, it is of great significance for operators to detect cyber-attacks and identify the types of attacks in CPPS to make decisions appropriately. This problem can be viewed as a multi-class classification problem from the perspective of machine learning. Accordingly, this paper proposes an ensemble learning-based cyber-attacks detection model for CPPS state estimation. In the proposed model, we use the subspace features and then send the data to the classifier for ensemble, where decision trees and random vector functional link networks are introduced as the basic classifier. The proposed model is validated by employing simulated data on IEEE 14-bus and 30-bus systems. Simulation results demonstrate the performance of the proposed cyber-attacks detection model.