Volume 2: Innovative Solutions for Energy Transitions: Part I

Detection of False Data Injection Attack in Automatic Generation Control System Based on Local Predictor and Support Vector Machine Z. Y. Chen, K. S. Xiahou, M. S. Li, T. Y. Ji, Q. H. Wu* ,

Abstract

False data injection attack (FDIA) invades the automatic generation control (AGC) system and degrades the control performance, which may cause unstable operation of power system. Fast and accurate detection can help to reduce the impact of attacks. This paper presents a novel detection method, combining local predictor (LP) and support vector machine (SVM), for the FDIA of AGC system. The effects of different types of cyber attacks on AGC system are analyzed. The LP is applied to identify the local pattern of each point of historical data, and it extracts the information in a high dimension space with accurate predictions. The similar data obtained from LP are adopted to train the SVM, and the LP-SVM algorithm is presented to detect the attacks of AGC system. Simulation studies undertaken on a single-area AGC system reveal that the LP-SVM method outperforms traditional SVM and naive Bayes (NB).

Keywords Automatic Generation Control System, False Data Injection Attacks, Local Predictor, Support Vector Machine