Transient stability preventive control of power systems using chaotic particle swarm optimization combined with two-stage support vector machine

•An improved Relief-Wrapper method for feature selection is proposed.•A two-stage support vector machine (SVM) method is proposed for security boundary approximation.•The chaotic search method is used for particle swarm optimization (PSO) to avoid premature convergence.•Economic and effective preventive control strategy is obtained using the chaotic PSO combined with two-stage SVM. This paper presents a chaotic particle swarm optimization (CPSO) algorithm combined with data mining method for transient stability preventive control. The data mining method is utilized to approximate the security region considering transient stability. Therefore, the application effects of different input features and data-mining classifiers are compared first. Then, a two-stage support vector machine (SVM) approach is proposed to generate two models, including a linear SVM model with controllable features provides preventive adjustment rules, and a more accurate SVM model to approximate the actual security region. Finally, the CPSO in combination with the two-stage SVM is proposed to calculate the optimal preventive control strategies. Comprehensive studies are conducted on a 16-machine 68-bus system and 48-machine 140-bus system to verify the effectiveness.