Feature Selection For Transient Stability Evaluation In Power System

Feature selecting plays a very important role in power system transient stability assessment. Based on data mining techniques, algorithms for feature selecting are discussed in this dissertation. Aiming at particularities of dynamic data on power system, conforming to existent algorithms, a new algorithm for feature selecting is proposed based on principal component analysis and genetic algorithm.In the dissertation, the main assess methods of power system transient stability are expatiated upon, and and the application and reseach status are described as well. Elementary principles and development status of feature selection methods in domestic and overseas are discussed, and then both advantages and disadvantages of existent methods are analyzed in detail.In studies of evaluating transient stability, most target power system has a small size, such as IEEE39-bus system. In the dissertation the primary feature samples of IEEE16-machine 86-bus system and IEEE50-machine 453-bus system are established. Principal component analysis (PCA) and genetic algorithm (GA) have been used to efficiently reduce the dimension of the primary feature; at last reconstruct the input space by using the idea of fact loading, to accomplish feature selection.The capability and reliability of recognition is provided by the SVM test result.