基于小波神经网络与KNN机器学习算法的六相永磁同步电机故障态势感知方法  

Faul tstate perception method for six-phase PMSM based on wavelet neural network and KNN machine learning algorithm

作　　者：张昊宇 姚钢[1] 殷志柱[2] 周荔丹[1] Zhang Haoyu;Yao Gang;Yin Zhizhu;Zhou Lidan(Key Laboratory of Power Transmission and Conversion Control of Ministry of Education,Shanghai Jiao Tong University,Shanghai 200240,China;Central Academe eo Shanghai Electric Group Co.,Ltd.,Shanghai 200070,China)

机构地区：[1]上海交通大学电力传输与功率变换控制教育部重点实验室,上海200240 [2]上海电气集团中央研究院,上海200070

出　　处：《电测与仪表》2019年第2期1-9,共9页Electrical Measurement & Instrumentation

基　　金：国家自然基金科学基金资助项目(61374155);高校博士点专项科研基金(20130073110030)

摘　　要：为了避免六相永磁同步电机在运行过程中因缺相引发更严重的电机故障和系统崩坏,需对电机在故障发生前进行提前预测判断和在故障发生后进行故障类型识别。根据故障下定子磁动势不变原理,推导Y移30°中性点隔离型六相永磁同步电机在各缺相故障下的数学模型。通过小波包分析方法提取故障工况下的特征值,构建小波神经网络模型对故障发生进行预测判断,避免系统误触发;构建KNN机器学习系统,对故障类型进行快速识别,以实现对故障态势的感知。利用MATLAB软件和Python的Scikit-Learn机器学习库进行仿真实验,对比验证该方法在六相永磁同步电机故障态势感知中可靠有效。In order to avoid the more serious motor fault and system breakdown which were caused by open phase during the running of six-phase permanent magnet synchronous motor(PMSM),it is necessary to do the fault prediction and diagnosis.This paper derived mathematical model of neutral point isolation six-phase PMSM shifted by 30°with the principle of stator magnetomotive force invariance.Wavelet packet analysis was adopted to collect the feature values and the wavelet neural network was built to do the fault prediction and avoid system spurious triggering.The K-Nearest Neighbor(KNN)machine learning system had also been built to diagnose the fault types quickly which could realize fault state perception.MATLAB and Scikit-Learn library of Python were used to do simulations which could verify the reliability and effectiveness of strategy.

关 键 词：永磁同步电机 神经网络 机器学习 小波包分解 故障态势感知 

分 类 号：TM933[电气工程—电力电子与电力传动]