Real-Time Transient Stability Prediction Based on Relevance Vector Learning Mechanism for Large-Scale Power System

Abstract

One of the most challenging problems in real-time operation of power system is the prediction of transient stability. Fast and accurate techniques are imperative to achieve on-line transient stability assessment (TSA). This problem has been approached by various machine learning algorithms, however they find a class decision estimate rather than a probabilistic confidence of the class distribution. To counter the shortcoming of common machine learning methods, a novel machine learning technique, i.e. 'relevance vector machine' (RVM), for TSA is presented in this paper. RVM is based on a probabilistic Bayesian learning framework, and as a feature it can yield a decision function that depends on only a very fewer number of so-called relevance vectors. The proposed method is tested on a practical power system, and compared with a state-of-the-art 'support vector machine' (SVM) classifier. The classification performance is evaluated using false discriminate rate (FDR). It is demonstrated that the RVM classifier can yield a decision function that is much sparser than the SVM classifier while providing more higher classification accuracy. Consequently, the RVM classifier greatly reduces the computational complexity, making it more suitable for real-time implementation.