Fault Diagnosis for Smart Grid by a Hybrid Method of Rough Sets and Neural Network

Traditional fault diagnosis methods based on relay protection are no longer suitable with multiple distributed generators in Smart Grid. In order to improve the accuracy and rapidity of fault diagnosis with DG interconnected, a novel hybrid method of intuitionistic uncertainty rough sets and BP neural network was introduced. Firstly, based on data pretreatment, the original fault diagnosis samples were discretized by the hybrid clustering method. Then, the decision attribute was reduced to delete redundant information for obtaining the minimum fault feature subset. In the course of identifying fault diagnosis through BP neural network, some output results were modified by using the inference capability of expert system. The worked example for Xigaze power system in China's Tibet shows the effectiveness of the method and the fault identification rate is improved by 30%.