A Neural Network Based Resilient Control Design for Distributed Power Systems Under Faults and Attacks

A novel active resilient control system is developed for distributed power systems (DPSs) under false data injection (FDI) attacks, and faults. The proposed system works based on a new anomaly detection (AD) design which consists of a Luenberger observer and an artificial neural network (ANN). The ANN structure is developed by Extended Kalman filter (EKF) to improve the ANN ability for the online AD in the power system. Based on the feedback data received from the AD system, the resilient controller will be designed, which eliminates the need for control reconfiguration. The resiliency of the proposed design against FDI attacks and faults in the sensors is tested on a Load Frequency Control (LFC) system through numerical simulations. Based on simulation results, the proposed active resilient control system can successfully detect anomalies in the actuators and compensate for their effects on DPSs.