摘要
为了进一步提高变压器故障诊断精度,提出基于云模型与改进D-S证据理论的变压器故障诊断方法。首先,利用油中溶解气故障样本建立各类型故障的标准云模型,并计算待测样本与故障标准云间的隶属度值;其次,根据隶属度确定不同故障下的基本概率分配;然后,利用Pignistic概率距离构建相似度并对基本概率分配进行修正,再引入平均支持度加权优化证据融合规则;最后,使用基于云模型和改进证据理论建立的诊断模型对实际案例进行测试与诊断分析。实验结果表明:基于云模型和改进D-S证据理论建立的模型故障诊断精度达88.4±2.8%,较支持向量机、K最近邻分类算法和灰色关联分析法的识别率分别提高了7.8%、3.8%、15.7%,验证了所建模型具有更优越的故障诊断性能。
For improving fault diagnosis accuracy of power transformer further,the fault diagnosis method based on cloud model and improved Dempster-Shafer evidence theory is proposed in this paper. Firstly,the standard cloud model for each type of fault is set up by the use of dissolved gas in oil samples and the membership degrees between the samples to be measured and fault standard cloud and testing samples are calculated. Secondly,the basic probability assignments for different fault types are defined in accordance with the obtained membership degrees. Then,similarity is calculated by Pignistic probability distance to correct the basic probability assignments and the average support weighting is introduced to modify the fusion rules of D-S theory. Finally,the diagnosis model based on cloud model and improved D-S theory is used to perform test and diagnosis analysis of the real case. The experimental results show that the fault diagnosis accuracy of the proposed method,which is set up on the basis of cloud model and improved Dempster-Shafer evidence theory,is up to 88.4±2.8%,which is 7.8%,3.8% and 15.7% higher than that of support vector machine,K-nearest neighbor algorithm and grey correlation analysis method,respectively,which verify that the setup model has more superior fault diagnosis performance.
作者
张宽
吐松江·卡日
高文胜
伊力哈木·亚尔买买提
孙国良
何志洋
ZHANG Kuan;TUSONGJIANG·Kari;GAO Wensheng;YILIHAMU·Yaermaimaiti;SUN Guoliang;HE Zhiyang(School of Electrical Engineering,Xinjiang University,Urumqi 830049,China;Department of Electrical Engineering,Tsinghua University,Beijing 100084,China)
出处
《高压电器》
CAS
CSCD
北大核心
2022年第4期196-204,共9页
High Voltage Apparatus
基金
国家自然科学基金(52067021)
新疆优秀青年科技人才培养项目(2019Q012)
清华大学电力系统及大型发电设备安全控制和仿真国家重点实验室开发课题(SKLD20KM19)
新疆维吾尔自治区高校科研计划(XJEDU2019Y013)
新疆维吾尔自治区自然科学基金(2022D01C35)。
关键词
变压器
故障诊断
油中溶解气体
云模型
证据理论
power transformer
fault diagnosis
dissolved gas analysis
cloud model
evidence theory
