Failure Mode Effect Classification for Power Electronics Converters Operating in a Grid-Connected System

Power electronic interfaces are the key aspects for achieving efficient grid integration for various distributed generation applications. As these interfaces continue to increase, their failure will result in major power losses and unstable operation of the electrical network. This article proposes a failure mode effect classification (FMEC) approach for localizing the faults in power electronic converters. The approach is developed with model-driven fault detection for identifying the fault signatures and data-driven fault identification for classifying the fault. This aims at identifying the fault effect on inputs, components, and sensors without compromising with the power stage of the converter. Furthermore, numerical simulations are carried out with a three-phase converter to acquire the fault signature library, and k-nearest neighbor approach is used to train the datasets. The fault signature library handles the information related to filter residuals obtained from the fault magnitude of each fault scenario. The proposed approach is validated through the experimental analysis of a real-time operation of a three-phase converter. The classifier training showed 96.5% accuracy, testing accuracy is 95.75%, and the fault detection time is 0.04 s. The testing results of the FMEC accurately identified various faults under varying load conditions without compromising the dynamic performance of the algorithm.