矿用干式变压器故障诊断和寿命预测技术现状及展望

Current situation and prospect of fault diagnosis technology formine dry-type transformer

干式变压器故障诊断技术通过对监测数据进行分析,能够实现信号特征提取、故障诊断、状态评估和寿命预测,提高干式变压器在井下运行的安全性,降低由于故障带来的损失。文章介绍了信号特征提取方法(频率响应法、小波变换法、堆叠自编码器)、故障诊断方法(贝叶斯网络、支持向量机、BP神经网络)、状态评估和寿命预测方法(交叉熵组合预测法、灰色理论、层次分析法)的研究现状,将几种方法进行对比,并对干式变压器故障诊断技术进行展望。

The feature extraction methods for fault diagnosis signals include frequency response method,wavelet transform method,and stacked auto encoder;Fault diagnosis methods include Bayesian networks,support vector machines,and BP neural networks;State assessment and life prediction methods include cross entropy combination prediction method,grey theory,and analytic hierarchy process.To improve the safety of dry-type transformers during underground operation and reduce losses caused by faults.The research status of fault diagnosis signal feature extraction method,fault diagnosis method,state evaluation and life prediction method,were introduced and compared.Finally,an outlook was given on the fault diagnosis technology of dry-type transformers,pointing out that the fusion of fault diagnosis feature quantities,the application of big data and artificial intelligence can provide a good solution for future research on dry-type transformer fault diagnosis.