一种用于变压器故障诊断的贝叶斯网络优化方法  

A Bayesian Network Optimization Method for Transformer Fault Diagnosis

作　　者：仝兆景[1] 荆利菲 兰孟月 TONG Zhaojing;JING Lifei;LAN Mengyue(School of Electrical Engineering and Automation,Henan Polytechnic University,Jiaozuo 454003,China)

机构地区：[1]河南理工大学电气工程与自动化学院,河南焦作454003

出　　处：《电子科技》2024年第8期34-39,共6页Electronic Science and Technology

基　　金：国家自然科学基金(U1504623);河南理工大学教育教学改革基金(2021YJ10)。

摘　　要：针对变压器故障诊断效率低的问题,文中将油中溶解气体分析与人工智能方法相结合,提出了一种改进蝗虫优化算法优化贝叶斯网络的变压器故障诊断方法。利用差分进化算法和与模拟退火算法对蝗虫算法进行改进,提高了算法的优化能力。将改进蝗虫算法应用于贝叶斯网络结构来学习构建变压器故障诊断模型,利用所提方法对变压器进行故障诊断。实验结果表明,该方法诊断正确率达到了92.7%,与其他算法所构建的诊断模型相比具有更高的故障诊断准确率。In view of the low efficiency of transformer fault diagnosis,an improved grasshopper optimization algorithm is proposed by combining dissolved gas analysis in oil with artificial intelligence method to optimize the transformer fault diagnosis method of Bayesian network.The differential evolution algorithm and simulated annealing algorithm are used to improve the locust algorithm,which improve the optimization ability of the algorithm.The improved locust algorithm is applied to the Bayesian network structure learning to construct the transformer fault diagnosis model,and the method proposed in this study is used to diagnose the transformer fault.The experimental results show that the diagnosis accuracy of this method is 92.7%,which is higher than that of other algorithms.

关 键 词：变压器 蝗虫算法 差分进化算法 模拟退火算法 油中溶解气体 贝叶斯网络 故障诊断 结构学习 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]