摘要
针对风电变流器IGBT模块开路故障,在诊断中长时间序列信号的特征时难以提取和识别,文章提出了一种基于栈式稀疏自编码(SSAE)网络和长短期记忆(LSTM)神经网络的开路故障诊断方法。以网侧变流器为主要研究对象,首先,将预处理后的原始电流信号输入SSAE网络,利用无监督学习方式进行逐层贪婪训练,并结合有监督学习方式对SSAE网络进行参数更新和局部微调,进而提取隐含层降维特征,构建特征矩阵;其次,利用LSTM神经网络在处理时间序列中的记忆优势,将特征矩阵作为LSTM网络的输入进行模型的训练;最后,利用Softmax分类器实现故障的识别和分类。诊断结果表明,该方法实现了自动提取网侧变流器的故障电流信号特征;同时所提方法能够风电变流器IGBT模块单一开路和双开路的22种开路故障问题进行准确地识别和分类,平均测试集准确率可达99.64%。
Aiming at the problem of feature extraction and recognition of long-time series signals in the open circuit fault diagnosis of IGBT module of wind power converter,an open-circuit fault diagnosis method based on Stacked Sparse Auto-Encoder(SSAE)network and Long-Short Term Memory(LSTM)neural network is proposed which uses the grid-side converter as the main research object.Firstly,input the preprocessed original current signal into the SSAE network to use the unsupervised learning method to perform layer-by-layer greedy training,and then combine the supervised learning method to update the parameters and local fine-tuning of the SSAE network to extract the dimensionality reduction features of the hidden layer for constructing the feature matrix;Secondly,use the feature matrix as the input of the LSTM network to train the model by taking the advantages of the memory strengths of the LSTM neural network in processing time series;Finally,use the Softmax classifier to identify and classify faults.The diagnosis results show that the method realizes the automatic extraction of the fault current signal characteristics of the grid-side converter;At the same time,the proposed method can accurately identify and classify 22 fault problems of single open circuit and double open circuit of the IGBT module of the wind power converter and the average rate of the test set accuracy can reach 99.64%.
作者
张瑞成
白晓泽
董砚
邸志刚
孙鹤旭
张靖轩
Zhang Ruicheng;Bai Xiaoze;Dong Yan;Di Zhigang;Sun Hexu;Zhang Jingxuan(College of Electrical Engineering,North China University of Science and Technology,Tangshan 063210,China;School of Electrical Engineering,Hebei University of Technology,Tianjin 300310,China;School of Electrical Engineering,Hebei University of Science and Technology,Shijiazhuang 050018,China;Green Intelligent Mining Technology Innovation Center of Hebei Province,Tangshan 063210,China)
出处
《可再生能源》
CAS
CSCD
北大核心
2023年第3期361-369,共9页
Renewable Energy Resources
基金
河北省重点研发计划项目(20314502D)
河北省教育厅科学技术研究项目(ZD2021332,JQN2020020,JQN2022001)
唐山市科技计划项目(21130219C)。
