A Missing Data Tolerance Data-driven Method for Open-Circuit Fault Diagnosis of Three-phase Inverters Based on Random Forest and Resampling Scheme

Data-driven methods have shown promising performance for fault diagnosis of three-phase power inverters. In practice, missing data problems may occur during the real-time sampling phase, which can lead to a low-quality dataset and poor performance of data-driven methods. In this paper, a new missing-data tolerance method is proposed for open-circuit fault diagnosis in three-phase inverters. First, a data-driven diagnostic model is trained by Random Forest and then a resampling scheme is proposed to solve the missing data problem to improve the online performance. Moreover, the relationship between the loss amount of data and diagnostic accuracy is analyzed. In the end, several test results are given to verify the effectiveness of the proposed method.