Feature Selection and Fault Diagnosis Method for Switchgear Cabinet

In this paper, the scheme to monitor the fault features in switchgear is designed. From the features reflecting different types of faults such as insulation, mechanical, temperature rise, arc and so on, the original feature set for switchgear fault can be achieved which is including some diagnostic indicators. Two kinds of score with local preserving and global separation is weighted by introducing a weight coefficient, then the improved Laplacian score is formed to sort the importance level of fault features, which refines the local preserving for adjacent samples and global separation for non-adjacent samples of the features subset. By using fuzzy support vector machine (SVM) classifier to check feature subset, and then the optimal fault feature subset of switchgear is obtained. Finally, the fault diagnosis of switchgear is implemented by using Mahalanobis distance (MD) to quantify the similarity of fault features and standard samples. According to the instance analysis of monitoring data from a switching station, the correction of the proposed method is verified.