摘要
针对变压器故障诊断中模型训练时间长,容易过拟合,噪声敏感等问题,提出一种深度降噪极限学习机变压器故障诊断方法。将极限学习机与降噪自编码器结合构建降噪自编码极限学习机,并将其堆叠构建深度降噪极限学习机模型进行特征提取,将提取的特征输入常规极限学习机进行分类,整体构成深度降噪极限学习机分类算法。该算法能有效应对电压器油中溶解气体分析数据中的噪声且具有非常快的学习速度。仿真实验结果表明,相比于传统BP神经网络,文中方法有更高的故障诊断正确率和更短的训练时间,是一种有效的变压器故障诊断方法。
In view of the problems of long training time, easy over-fitting and noise sensitivity in transformer fault diagnosis, a transformer fault diagnosis method for deep de-noising extreme learning machine is proposed in this paper. The extreme learning machine is combined with the de-noising auto-encoder to construct the de-noising auto-encoding extreme learning machine. Several de-noising auto-encoding extreme learning machines are stacked to build a deep de-noising extreme learning machine model for feature extraction, and the extracted features are classified into a conventional extreme learning machine, and the whole structure is the deep de-noising extreme learning machine classification algorithm. The algorithm can effectively deal with the noise in the dissolved gas analysis data of transformer oil and has a very fast learning speed. The simulation results show that, compared with the traditional BP neural network, this method has higher fault diagnosis accuracy and shorter training time, which is an effective method of transformer fault diagnosis.
作者
王春明
朱永利
Wang Chunming;Zhu Yongli(School of Control and Computer Engineering, North China Electric Power University, Baoding 071003, Hebei, China)
出处
《电测与仪表》
北大核心
2019年第15期143-147,共5页
Electrical Measurement & Instrumentation
基金
国家自然科学基金资助项目(51677072)
关键词
变压器
故障诊断
深度极限学习机
降噪
油中溶解气体分析
transformer
fault diagnosis
deep extreme learning machine
de-noising
dissolved gas-in-oil analysis
