基于AVMD和CNN的并网型微网线路故障诊断
摘要:为提高微网三相线路故障诊断精度，提出基于自适应变分模态分解(adaptive variational mode decomposition，AVMD)和卷积神经网络(convolutional neural network，CNN)的微网线路故障诊断分类方法。首先建立包含风、光、水系统的微网径向结构模型；采用AVMD将原始故障信号分解得到多个模态分量，其中变分模态分解(variational mode decomposition，VMD)的参数采用天鹰优化(aquila optimizer，AO)算法进行优化；诸多模态中只有少数模态保留了故障信号的信息，利用有效加权峰态相关(effective weighted peak relevance，EWPR)指数对模态分量进行选择，选取最能保留故障信息的3个模态作为敏感模态；剔除噪声和其他无关模态的影响，使用CNN对微网的线路故障进行诊断分类。生成110组故障数据用于训练和验证神经网络，结果表明22组验证数据集中共有21组数据分类正确，此研究方法对故障的诊断精度达到了95.46%。
Abstract：In order to improve the fault diagnosis accuracy of three-phase microgrid lines, a fault diagnosis and classification method based on adaptive variational mode decomposition (AVMD) and convolutional neural network (CNN) is proposed in this paper. Firstly, a microgrid radial structure model including wind, light and water system is established. AVMD is used to decompose the original fault signal into multiple modal components, where the parameters of variational mode decomposition (VMD) are optimized by aquila optimizer (AO). Only a few of the modes retain the fault signal information. Effective weighted peak relevance index (EWPR) is used to select the modal components, and the three modes that can best retain the fault information are selected as sensitive modes. The influence of noise and other irrelevant modes is eliminated, and CNN is used to diagnose and classify the circuit faults of the microgrid. 110 groups of fault data are generated for training and verification of neural network. The results show that 21 groups of data in 22 groups of verification data sets are classified correctly, and the fault diagnosis accuracy of this method reaches 95.46%.
标题：基于AVMD和CNN的并网型微网线路故障诊断
title:Fault Diagnosis of Grid-Connected Microgrid Lines Based on AVMD and CNN
作者：付林瑶,李春华,汪本科,班勇霜
authors:FU Linyao,LI Chunhua,WANG Benke,BAN Yongshuang
关键词：自适应变分模态分解(AVMD),有效加权峰态相关(EWPR)指数,天鹰优化(AO),卷积神经网络(CNN),故障诊断分类,
keywords:adaptive variational mode decomposition (AVMD),effective weighted peak relevance (EWPR) index,aquila optimizer (AO),convolutional neural networks (CNN),fault diagnosis classification,
发表日期：2023-10-12
- 文件大小:
- 4.66 MB
- 下载次数:
- 60
-
高速下载
|
|
