Ningxia Electrical Power Research Institute of State Grid,Yinchuan,750011
电力变压器故障的准确诊断对于电网的可靠运行至关重要。为此本文提出了一种基于贝叶斯网络和假设检验的溶解气体分析的新型多分类概率诊断模型,该贝叶斯网络模型可嵌入专家知识,从数据中学习数据模式并推断与诊断结果相关的不确定性,并且通过假设检验环节改进数据的选择过程。本文使用IEC TC 10数据集,对比三种传统诊断方法进行了诊断实验,验证了所提出模型的有效性,结果表明本文提出的诊断模型最大诊断准确性为88.9％,相比传统诊断方法有较大提升。
accurate diagnosis of power transformer fault is very important for the reliable operation of power grid. At present, there are a series of transformer fault diagnosis methods and analysis models based on dissolved gas analysis, but the results of these methods are contradictory and cannot generate the uncertainty information related to the diagnosis results. Therefore, this paper proposes a new multi classification probability diagnosis model based on Bayesian network and hypothesis testing for dissolved gas analysis. The Bayesian network model can embed expert knowledge, learn data patterns from data and infer the uncertainty related to diagnosis results, and improve the data selection process through hypothesis testing. In this paper, IEC TC 10 data set is used to test the effectiveness of the proposed model compared with three traditional diagnostic methods. The results show that the maximum diagnostic accuracy of the proposed model is 88.9%, which is greatly improved compared with the traditional diagnosis methods.