适应分布式能源大量接入的输配电设备故障诊断方法研究

Research on Fault Diagnosis Method of Qil-immersed Transformer for Power Transmission and Distribution with Access of Large-scale Distributed Energy Resources

摘要:

针对光伏等大量分布式能源的接入对新型电力系统的安全运行带来的输配电设备故障诊断问题，提出了适应分布式能源大量接入下油浸式变压器的故障诊断方法。本文提出基于多源数据融合的DS证据理论（D-S evidence theory），构建深度自编码器故障诊断算法，以多源特征数据作为输入，采用DS证据理论对堆叠自编码器(Stacked Auto Encoder, SAE)模型输出结果进行数据融合，根据融合结果对故障诊断进行分析与研究。本文以某省2019年变压器检修数据为算例，评估多源数据对结果的影响以及其故障分析的正确率。仿真结果表明，多源数据融合可以提高故障诊断正确率，相较于最近邻方法（K-Nearest Neighbor, K-NN）与BP神经网络（Back Propagation Neural Network, BPNN），本文的数据融合模型具有更高的故障精度。

Abstract:

Aiming at fault diagnosis problems of transmission and distribution equipment brought by the access of large amount of distributed energy resources such as photovoltaic for the safe operation of new power systems, a fault diagnosis method of oil-immersed transformer with the access of large amount of distributed energy resources is put forward. DS evidence theory based on multi-source data fusion is proposed to construct a deep autoencoder fault diagnosis algorithm, which takes multi-source feature data as input. DS evidence theory is used to fuse the output results of stacked auto encoder model, and fault diagnosis is analyzed and studied based on the fusion results.Taking transformer maintenance data of a province in 2019 for example, the influence of multi-source data on results and the accuracy of fault analysis are evaluated. Simulation results show that multi-source data fusion can improve the accuracy of fault diagnosis. Compared with K-nearest neighbor and back propagation neural network, the proposed data fusion model has a higher fault accuracy.