基于小波变换和深度迁移学习的变压器故障自动化诊断  

Automatic Transformer Fault Diagnosis Based on Wavelet Transform and Deep Transfer Learning

作　　者：武锋利 WU Fengli(State Grid Shaanxi Electric Power Co.,Ltd.Weinan Power Supply Company,Weinan 714000,China)

机构地区：[1]国网陕西省电力有限公司渭南供电公司,渭南714000

出　　处：《自动化与仪表》2025年第2期115-118,123,共5页Automation & Instrumentation

摘　　要：为了提升变压器故障诊断性能,维持变压器的稳定运行,提出基于小波变换和深度迁移学习的变压器故障自动化诊断研究。通过常见故障深层探究确定故障信号种类,基于改进小波变换算法去除故障信号噪声,基于深度迁移学习技术构建变压器故障诊断模型,以源域数据交叉熵损失最小化、跨域分布差异损失最小化为目标,对构建模型进行2次优化改进,将无噪故障信号输入至训练好的诊断模型中,即可获得变压器故障诊断结果。测试结果显示,应用提出方法处理后的变压器故障信号质量更好,变压器故障诊断结果获取时间、迭代次数均小于限值,变压器故障诊断结果与测试故障类型保持一致。In order to improve the performance of transformer fault diagnosis and maintain the stable operation of transformer,an automatic transformer fault diagnosis based on wavelet transform and deep transfer learning is proposed.The fault signal types were determined by deep exploration of common faults,the fault signal noise was removed based on improved wavelet transform algorithm,and the transformer fault diagnosis model was constructed based on deep transfer learning technology.The constructed model was optimized and improved twice with the goal of minimizing cross-entropy loss of source domain data and minimizing cross-domain differential loss.The transformer fault diagnosis result can be obtained by inputting the noiseless fault signal into the trained diagnosis model.The test results show that the transformer fault signal quality is better after the proposed method is applied.The acquisition time and iterations of the transformer fault diagnosis result are less than the limit value.The transformer fault diagnosis result is consistent with the test fault type.

关 键 词：变压器故障 深度迁移学习 局部放电信号 故障种类判定 改进小波变换算法 跨域分布差异损失 

分 类 号：TM41[电气工程—电器] TP274[自动化与计算机技术—检测技术与自动化装置]