关键词: 逆变器, 三电平, 故障诊断, 决策树相关向量机, 经验模态分解

Abstract: Aiming at the problem of open-circuit fault arising in diode-clamped three-level inverter, a fault diagnosis strategy, based on the empirical mode decomposition and the decision tree relevance vector machine (EMD-DTRVM), is proposed in this paper. First, the operation conditions of main circuit in inverter are analyzed to classify faults, and the inverter bridge voltages are selected as the measured signals. Then, the fault features are extracted in the form of energy and energy entropy by the empirical mode decomposition. Moreover, particle swarm clustering algorithm is used to build the decision tree structure. By training the RVM classified models, the fault diagnosis of power component in three-level inverter is finally accomplished. Compared with traditional approaches, the proposed EMD-DTRVM strategy not only achieves shorter diagnosis time and higher accuracy, but also works simple and efficient with less parameters.

Key words: Inverter; three-level, fault diagnosis, decision tree relevance vector machine, empirical mode decomposition