Design And Implementation Of Hydropower Fault Classifer Based On Support Vector Machine

As hydropower faults are so complicated, traditional reasoning mechanism in hydropower simulation system has been unable to meet actual needs, so using data mining to replace the original reasoning mechanism in fault categorization is a new attempt in this field. Through the development of hydropower simulation system, people realize the limitation of functions of the traditional fault simulation in training the trainees, and man-made design and implementation of hydropower faults by computer programs are difficult to ensure their accuracy, therefore, these limitations the application of new technologies.First, hydropower faults are expressed in natural language in this thesis, which is to understand, and, according to TF-IDF weights algorithm, they are described in the form of the weighted word vector by using vector space model as well as Chinese words segmentation technology. Then, because hydropower faults are described in form of vector, so Support Vector Machine technology is used for fault categorization. Support Vector Machine and related technologies are introduced, such as the choice of kernel functions, and multi-faults categorization SVM algorithms are analyzed detailedly.In order to improve the accuracy of hydropower fault categorization, semi-supervised learning method is introduced in this thesis. Semi-supervised learning can make better use of a large number of unlabeled samples and it can improve the accuracy and ensure the efficiency at the same time. CPTSVM, an improved progressive transductive inference semi-supervised learning algorithm, is proposed. The experiments show that the algorithm can improve the accuracy of semi-supervised learning and can reduce the complexity of the algorithm.Finally, a hydropower fault classifier based on CPTSVM is designed and implemented, and the Precision, Recall and other evaluation indexes of the classifier can reach application requirements. Also the purpose for hydropower fault categorization and diagnosis can be achieved by this classifier, and it is foreseeable that this classifier can have good prospects in the general-purpose utilities hydropower simulation system.