Improving Diagnostic Performance of High-Voltage Circuit Breakers on Imbalanced Data Using an Oversampling Method

Abstract

In recent years, applications of artificial intelligence (AI) techniques in fault diagnosis of high-voltage circuit breakers (HVCBs) have gained wide attention. In real applications, how-ever, HVCBs work in the normal state most of the time. Therefore, the problem of imbalanced monitoring data is prevalent, which threatens the generalization capability of AI-based diagnosis methods, resulting in poor fault diagnostic performance. To address this problem, an oversampling method called Density-weighted Minority Oversampling (DWMO) was proposed to balance monitoring data in this study. Experimental results on HVCB monitoring data with various imbalance ratios show that DWMO can improve the diagnostic performance of AI techniques and outperformed other commonly used oversampling methods including Synthetic Minority Oversampling Technique (SMOTE), Borderline-SMOTE, Adaptive Synthetic Sampling Approach, and Majority Weighted Minority Oversampling Technique.