Transmission Line Faults Classification Method Based On Hilbert-Huang Transform And Fuzzy Support Vector Machine

With the continuous development of society’s economy and the increasing power system load, the occurrence frequency of transmission line fault increases. Accurate, rapid fault classification is a prerequisite to ensure realization of the transmission fault location and accident analysis.The method of transmission line fault classification based on fuzzy support vector machines is studied and its aim is when a transmission line short-circuit fault occurred it can achieve fault classification quickly and accurately.First, the method of Hilbert-Huang Transform is introduced. This method is applied successfully in signal feature extraction and it can well characterize the fault information.By using the method of EMD(Empirical Mode Decomposition), it can obtain IMF(Intrinsic Mode Function) of the different current signal. And then by applying HHT to the signal the hilbert marginal spectrum of A,B,C and zero sequence current are got.The transmission line fault classification based on FSVM is studied.First, The method of SVM(Support Vector Machine) is used to establish a preliminary classification model. The RBF kernel function is selected and Grid-search, genetic algorithms and PSO algorithms are applied to seclect the best C(Penalty Parameter) and σ (Kernel Width).The piecewise membership function is defined in the high-dimensional feature space based on SVM.The method of support vector regression on the basis of fuzzy support vector machine classification and solves the fuzzy membership of test set data belonging to each type of fault are introduced in this paper. It can get the final classification by comparing the pre-configured threshold, obtaining the classification results and correcting FSVM classification label. Simulation results show that this method not only has high recognition rate of fault, but also has high fault tolerance.The data dimension reduction through PCA(Principal Component Analysis) method and three-dimensional graphics display method is studied.Through PCA dimension reduction the three main element component are selected and it can characterize the original information fully and display in the space of three-dimensional graphics. At last,the different types of data with different symbols are displayed. Finally, comparative analysis between the SVM classification results and FSVM in three-dimensional space maps are to be done to verify the superiority of FSVM in transmission line fault classification with its high fault tolerance.