Cyber-Attack Detection: Modeling and Roof-PV Generation System Defending
Wei Qiu, Kaiqi Sun, Ke‐Jun Li, Yuchuan Li, Junfeng Duan, Kunzhi Zhu
Abstract
The continuous increase of renewable energy installation in the power system such as roof-PV systems decreases the system inertia. It is a non-negligible challenge to the system's operation stability. In addition, with the increasing number of cyber attack events reported globally, the operation of the roof-PV systems may threaten their connected AC system operation during the contingency. Utilizing the grid-connected converters (GCCs), its fast regulating characteristic could rapidly increase or decrease the solar generation output in seconds, which will bring significant influences to the system operation once the cyber attack happens, especially if the system has integrated high proportional renewable energies. This paper proposed a cyber-attack detection model to solve this problem to eliminate its effect on the roof-PV generation system. In this model, the synchrosqueezed wavelet transforms (SWT) are first applied to extract the time-frequency information of frequency measurement. Then a recurrent layer aggregation-based convolutional neural network is introduced to identify the features of cyber attacks using the results from SWT. The comparison experiments indicate that the proposed model has profound performance on the detection accuracy of the roof-PV generation system for cyber attack detection.