基于CNN-LSTM网络的交直流电网故障线路识别方法  

Fault Line Identification Method of AC/DC Power Grid Based on CNN-LSTM Network

作　　者：宋匡玮 吴浩 陈伟哲 SONG Kuangwei;WU Hao;CHEN Weizhe(School of Automation and Information Engineering,Sichuan University of Science&Engineering,Yibin 644000,China;Artificial Intelligence Key Laboratory of Sichuan Province,Yibin 644000,China;Yibin Wuliangye Co.,Ltd.,Yibin 644000,China)

机构地区：[1]四川轻化工大学自动化与信息工程学院,四川宜宾644000 [2]人工智能四川省重点实验室,四川宜宾644000 [3]宜宾五粮液股份有限公司,四川宜宾644000

出　　处：《四川轻化工大学学报（自然科学版）》2024年第5期50-58,共9页Journal of Sichuan University of Science & Engineering(Natural Science Edition)

基　　金：四川省科技厅项目(2022YFS0518,2022ZHCG0035);自贡市科技局项目(2019YYJC13,2019YYJC02,2020YGJC16);人工智能四川省重点实验室项目(2023RYY06);企业信息化与物联网测控技术四川省高校重点实验室项目(2022WYY04);四川轻化工大学人才引进项目(2021RC12);四川轻化工大学研究生创新基金项目(Y2023284)。

摘　　要：针对交直流混联电网故障特征复杂、故障线路识别率低的局限性,提出一种基于模糊熵(FuzzyEn)结合改进的卷积神经-长短期记忆网络(CNN-LSTM)的交直流电网故障线路识别方法。首先使用改进的小波阈值滤波算法对数据进行降噪处理,再利用模糊熵提取信号的故障特征。接下来在卷积神经网络(CNN)的全连接层上,添加一层长短期记忆网络(LSTM),对模糊熵提取的故障特征进行分类,最终实现故障线路的准确识别。实验结果表明,该算法识别准确率达到了99.5%,能够有效地诊断出交直流混联电网故障线路,同时在10 dB噪声干扰下仍达到96.0%的识别准确率,具有较好的抗噪声能力。To address the limitations of complex fault characteristics and low fault line identification rate in AC/DC hybrid power grids,a novel approach based on Fuzzy Entropy(FuzzyEn)that is used to accurately identify AC/DC power grid fault lines has been presented,which combines with the advanced Convolutional Neural Networks-Long Short Term Memory(CNN-LSTM).The process begins with the application of an improved wavelet threshold filtering algorithm to remove noise from the data.Subsequently,FuzzyEn is employed to extract crucial fault features from the signals.Long Short Term Memory(LSTM)is then incorporated into the fully connected layer of Convolutional Neural Networks(CNN)to classify the extracted fault features,which ultimately enables the precise identification of fault lines.Experimental results demonstrate that the recognition accuracy of this algorithm reaches 99.5%,which can effectively diagnose fault lines in AC/DC hybrid power grids.Moreover,the algorithm maintains a recognition accuracy of over 96.0%even in the presence of 10 dB noise,which has a good anti-noise ability.

关 键 词：交直流混联电网 小波阈值滤波 卷积神经网络 长短时记忆网络 模糊熵 故障线路识别 

分 类 号：TM773[电气工程—电力系统及自动化]