Power System Transient Stability Assessment Based On Pattern Recognition And Ensemble Learning

Pattern recognition may get rid of the restrictions of model described system operating characteristics, which can be used for online analysis. Its computation load is little influenced by system scale and whole assessment speed is very quick.Naive Bayes classifier is commonly used classifier in pattern recognition, it is now recognized as the probability of a simple and effective classification method, with simple, stuggy and highly effective characteristic. But because it is the establishment under the attribute variable relative kind of variable independent supposition premise, moreover this supposition often cannot satisfy in the actual problem, thus affecting its classified precision. This paper uses EPRI PSASP software simulation and carries on the fault data collection; The amount of sampling procedures for the preparation of features, forming a for transient stability assessment of the original sample set. On this basis, the proposed clustering based on gray correlation feature selection method, carried out on the amount of transient stability features of dimensionality reduction, relaxed the independence of the naive Bayesian classifier restrictions to some extent; And to naive Bayesian classifier as the base classifier, Bagging and AdaBoost using two kinds of ensemble learning algorithm to further improve the classification performance. Through to the New England 10 machine 39 node system's simulation computation, the results showed that this paper methods are valid and correct.