基于格拉姆角场与并行CNN的并网逆变器开关管健康诊断  

Health diagnosis of switch tube in grid-connected inverter based on Gramian angular field and parallel CNN

作　　者：李宗源 陈谦[1] 钱倍奇 牛应灏 张政伟 LI Zongyuan;CHEN Qian;QIAN Beiqi;NIU Yinghao;ZHANG Zhengwei(College of Energy and Electrical Engineering,Hohai University,Nanjing 211100,China)

机构地区：[1]河海大学能源与电气学院,江苏南京211100

出　　处：《电力自动化设备》2024年第8期153-159,共7页Electric Power Automation Equipment

基　　金：国家自然科学基金资助项目(51837004)。

摘　　要：针对并网逆变器开关管实际运行中易出现缺陷状态而导致电压/电流波形异常的问题,提出了一种基于格拉姆角场与并行卷积神经网络(CNN)相结合的逆变器开关管健康诊断方法,以实现对逆变器进行监测及预测性诊断。采集逆变器输出端电压与电流信号,设定并计算虚拟电阻参数再将其转化为一维时序序列;利用格拉姆角场对其进行变换,提取出与逆变器开关管缺陷相关的格拉姆角和场与格拉姆角差场2组图像数据;将生成的2组图像同时送入CNN进行并行学习训练。实验结果表明所提方法及训练模型能及时有效地对逆变器异常状态进行诊断,且诊断准确率高,鲁棒性好。Aiming at the issue of abnormal voltage/current waveforms caused by defects in the actual operation of grid-connected inverter switch tubes,a health diagnosis method for inverter switch tubes based on the combination of Gramian angular field(GAF)and parallel convolutional neural network(CNN)is proposed to achieve monitoring and predictive diagnosis of inverter.The voltage and current signals at the output end of the inverter are collected,and the virtual resistance parameters are set,calculated and then converted into one-dimensional time series.The method of GAF transform is used to extract two groups of image data relating to the discrete of the inverter switch,i.e.,Gramian angular summation field image and Gramian angular difference field image.The two groups of images are sent to CNN simultaneously for parallel learning training.The experimental results show that the abnormal state diagnosis of inverters can be realized by the proposed method and training model,which is of high accuracy and good robustness.

关 键 词：并网逆变器 健康诊断 开关管缺陷 格拉姆角场 并行CNN 

分 类 号：TM464[电气工程—电器]