基于迭代退火算法的高压变频器功率单元频繁故障诊断方法研究*

现有的变压器故障诊断方法较为复杂且计算冗余度较高，在高压变频器的功率单元频繁发生故障时难以高效地检测故障。为此，提出基于迭代退火算法的高压变频器功率单元频繁故障诊断方法。采用小波包分解方法提取高压变频器功率单元的电压信号特征熵，将该特征熵输入到支持向量机模型。使用迭代退火算法优化支持向量机的训练参数，并输出诊断结果。研究结果表明:该方法提取的高压变频器单元故障的平均冗余度最低至3.2%，平均诊断时间为15.1 ms，可实现高压变频器功率单元频繁故障的高效诊断。

Research on diagnosis method of frequent faults in power unit of high voltage inverter based on iterative annealing algorithm

The existing fault diagnosis methods of transformer are complex and have high computational redundancy,which is difficult to detect faults efficiently when the power unit of high voltage converter has frequent faults.Therefore,a diagnosis method of frequent faults in the power unit of high voltage inverter based on iterative annealing algorithm was proposed.The characteristic entropy of voltage signals in the power unit of high voltage inverter was extracted by wavelet packet decomposition method,and the characteristic entropy was input to support vector machine (SVM) model.The training parameters of SVM were optimized by the iterative annealing algorithm,and the diagnostic results were output.The results showed that the average redundancy of the faults of the high voltage inverter unit extracted by this method was as low as 3.2,and the average diagnosis time was 15.1 ms,so the efficient diagnosis on the frequent faults in the power unit of high voltage inverter could be realized.