基于RepLKNet卷积网络的风电变流器开路故障诊断方法
Open circuit fault diagnosis method for wind power converters based on RepLKNet convolutional network
摘要
针对传统风电变流器故障诊断方法难以提取故障特征的问题,文章基于 RepLKNet卷积网络提出一种风电变流器开路故障诊断方法.首先,采集风电变流器三相输出相电压作为原始信号,经变分模态分解后,通过 Pearson 相关性分析选取包含主要故障特征的本征模态函数,并对其进行重构和归一化;其次,将归一化后的信号进行一维组合,采用格拉姆角差场转化为二维特征图像后输入 RepLKNet卷积网络,进行自动提取故障特征;最后,通过 Softmax 层进行决策和分类.仿真结果表明,文章所提方法可自动提取信号中的故障特征,对变流器22种单、双开路故障的准确率达99.79%.
To address the difficulty of fault feature extraction in traditional fault diagnosis methods for wind power converters,a open circuit fault diagnosis method is proposed for wind power converters based on RepLKNet convolutional network.Firstly,the three-phase output phase voltages of the wind power converter are collected as original signals.After Variational Mode Decomposition,Intrinsic Mode Functions containing dominant fault features are selected via Pearson correlation analysis,followed by signal reconstruction and normalization.Secondly,the normalized signals are assembled into a one-dimensional sequence,which is then converted into two-dimensional feature images using the Gramian Angular Difference Field.The obtained images are fed into the RepLKNet convolutional network for automatic fault features extraction.Finally,classification decisions are implemented through the Softmax layer.Simulation results demonstrate that the proposed method can automatically extract fault features embedded in signals,achieving a diagnostic accuracy of 99.79%for 22 types of single and double open circuit faults of the converter.
年,卷(期):
2026, 44(9)
分类号:
TK51
机标分类号:
TP391
TH133.33
TN911.7
在线出版日期:
2026年10月14日
基金项目:
2024ZD0801200:国家科技重大专项CXY2024013:河北省教育厅科技项目21130219C:唐山市科技计划项目,