A single-class attack detection algorithm for smart grid AGC system based on improved support vector machines algorithm

Chao Hong, Jiaye Tao, Yun Fu, Yiwei Yang, Lipeng Wei, Zhihong Liang, Junrong Liu

Abstract: Automatic generation control (AGC) system is an important part of modern power systems which faces the threat of cyber attacks. For the detection of cyber attacks on the AGC system, a single-class attack detection method based on an improved support vector machines (SVM) algorithm is proposed. This algorithm first uses a SVM-recursive feature elimination (SVM-RFE) method to reduce feature dimensionality of sample data points, obtaining the optimal feature subset. Then, time difference is used to expand the data dimensionality and model the dynamic information of the system based on SVM algorithm. Simulation verification is conducted based on the IEEE 39-bus model, and the results show that the proposed algorithm has good detection performance.