Multi-support vector machine power system transient stability assessment based on relief algorithm

It is difficult to choose a best feature subset for the power system transient stability assessment (TSA) problem, and the existing data mining methods for TSA lack sufficient considerations for these situations that wrongly classify unstable samples as stable ones. In response to these deficiencies, this paper proposes a multi-support vector machine (SVM) power system TSA method based on relief algorithm. Firstly, the proposed method selects several feature subsets with different size based on relief algorithm, then uses these selected feature subsets for SVM training, finally, these trained SVMs are integrated to assess the transient stability of power system. The analysis of a simulation system shows that the proposed method can take full advantage of the useful information provided by different feature subsets, and it can get a significant reduction in "misclassification" samples, which provides a useful reference for using data mining theory for TSA in real power systems.