Application of support vector regression model based on phase space reconstruction to power system wide-area stability prediction

With the development of wide area measurement technology, it will open up new possibilities for dynamic power system protection and control. In this paper, a novel time series prediction algorithm via support vector regression (SVR) technique is presented, which utilizes synchronized phasor data to provide fast transient stability swings prediction for the use of emergency control. Basic theory analysis of support vector regression algorithm based on phase space reconstruction of time series is minutely introduced and a multi-step prediction formula of generator rotor angles is presented. And, Final Prediction Error (FPE) principle is suggested to select the embedding dimension of the prediction model. Compared with traditional recurrent neural networks (RNN) prediction method, SVR adopts the new type of structural risk minimization principle, so it owns excellent generalization ability. The proposed approach has been tested on a practical power system, and the result indicates the effectiveness of such prediction model.