基于极限学习机与规则推理的NPC三电平逆变器二级故障诊断方法

A method of two layers fault diagnosis based on ELM and RI theories for NPC inverter

摘要:

针对NPC三电平逆变器故障诊断问题,提出一种基于极限学习机与规则推理的二级故障诊断方法。分析了依据输出电流诊断故障的可区分性,以及故障模式的分类。然后对输出电流提取故障特征,并采用极限学习机完成故障初级分类。对于初级分类结果为电流不可区分故障情况,再根据桥臂电压信息运用规则推理法实现故障二级精确诊断。诊断实验表明,该方法能够实现NPC三电平的多模式故障诊断,且故障诊断方法简单、定位精确、快速、鲁棒性强。

Abstract:

Aiming to diagnosis faults of power devices open circuit, a new fault diagnosis method based on extreme learning machine(ELM) and rules inference (RI) theories for NPC inverter is proposed. The paper analyzes the distinguishability of faulty mode with current output and make a classification according to theSdistinguishability. Then fault features is extracted from output current, and extreme learning machine is established to make primary fault diagnosis. If the primary result is the current undistinguishable fault, the rule inference based on information of bridge voltage is used to accurately locatethe fault components, which is the second level diagnosis. At last, the results of diagnostics show that the proposed method can diagnosis and detect the many fault patterns and that the method has a good performance of accurate positioning, simple structure, fast diagnosis and robustness.