基于WPES-SAE的MPPT控制器多故障诊断方法  

A Multiple Faults Diagnosis Method for MPPT Controller Based on WPES-SAE

作　　者：祝勇俊[1,3] 孙权[2,3] 朱其新[1] ZHU Yong-jun;SUN Quan;ZHU Qi-xin(Suzhou University of Science and Technology,Suzhou 215009,China)

机构地区：[1]苏州科技大学,江苏苏州215009 [2]南京工程学院自动化学院,江苏南京211167 [3]南京航空航天大学自动化学院,江苏南京211106

出　　处：《电力电子技术》2019年第7期98-101,共4页Power Electronics

基　　金：国家自然基金项目(51875380);江苏省住房和城乡建设厅科技项目(2017ZD096)~~

摘　　要：针对光伏发电系统最大功率点跟踪(MPPT)控制器多故障模式情形时难以有效进行故障准确识别,此处提出一种基于堆栈自动编码器(SAE)的功率变换器多故障诊断方法。首先,选取合适的电路测试点并采集电压信号;其次,采用小波包分解方法对各测点电压信号进行故障特征提取,并以小波包能量谱(WPES)作为功率变换器的故障特征向量;最后,采用4层SAE深度学习网络实现多故障模式的准确分类。仿真实验表明,所提方法诊断率可达100%,优于反向传播神经网络(BPNN)及支持向量机(SVM)所得诊断结果,验证了所提方法的准确性和有效性。Aiming at the difficulty in identifying the multiple faults for maximum power point tracking(MPPT)controller,which is an important part in photovoltaic power generation system.Thus,a multiple faults diagnosis method based on stacked autoencoder(SAE)is proposed.First,selecting the appropriate circuit test points and collecting the voltage signals.Secondly,the wavelet packet decomposition method is used to extract the fault features from the voltage signals,and the wavelet package energy spectrum(WPES)is used as the fault feature vector for power converter.Finally,a four-layer SAE deep learning network is applied to accurately classify the multiple fault modes.Simulation results show that the proposed method has a classification rate of 100%,which is better than the results obtained by back propagation neural network(BPNN)and support vector machine(SVM).Also,the accuracy and effectiveness of the proposed method are verified.

关 键 词：故障诊断 最大功率点跟踪 小波包能量谱 堆栈自动编码器 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置]