基于多级决策融合模型的电力变压器故障深度诊断方法  

Deep fault diagnosis of power transformer based on multilevel decision fusion model

作　　者：李刚[1,2] 于长海 范辉 刘云鹏[2] 宋雨[1] 

机构地区：[1]华北电力大学控制与计算机工程学院,河北保定071003 [2]华北电力大学新能源电力系统国家重点实验室,北京102206 [3]国网河北省电力公司电力科学研究院,河北石家庄050011

出　　处：《电力自动化设备》2017年第11期138-144,共7页Electric Power Automation Equipment

基　　金：国家自然科学基金资助项目(51407076);国家电网公司科技项目(5204DY170010);河北省自然科学基金资助项目(F2014502050);中央高校基本科研业务费专项资金资助项目(2015ZD28)~~

摘　　要：针对电力变压器故障的深度诊断问题,提出一种深度置信网络与D-S证据理论相结合的方法。采用深度置信网络对电力变压器故障的多维数据进行特征提取及分类,并结合D-S证据理论解决故障诊断中的不确定性问题,构造了电力变压器故障诊断的多级决策融合模型。以变压器油中溶解气体、局放量以及历史故障数据和家族质量史等数据为样本进行仿真实验,结果表明所提方法对于具备大量多源信息的电力变压器故障诊断问题是有效的。A method combining deep confidence nets with D-S evidence theory is proposed to solve the deep fault diagnosis problem of power transformer.A multi-level decision fusion model is established for fault diagnosis of power transformer,by using deep confidence nets to extract and classify the characteristics of multi-source information in faults of power transformer and using D-S evidence theory to solve the uncertainties of fault diagnosis.Simulation is carried out with dissolved gases in transformer oil,partial discharge value,historical fault data and family history of quality as samples,whose results show that the proposed method is effective for fault diagnosis of power transformer,which has a large amount of multisource information.

关 键 词：电力变压器 故障诊断 多源信息融合 深度置信网络 D-S证据理论 不确定性分析 

分 类 号：TM41[电气工程—电器]