基于改进粒子群优化XGBoost的变压器故障诊断方法  

Fault Diagnosis Method of Transformer Based on Improved Particle Swarm Optimization XGBoost

作　　者：龚泽威一 饶桐 王钢 李钊 骆钊[2] 朱家祥 彭晶 于虹 曹占国 GONG Zeweiyi;RAO Tong;WANG Gang;LI Zhao;LUO Zhao;ZHU Jiaxiang;PENG Jing;YU Hong;CAO Zhanguo(Electric Power Research Institute of Yunnan Power Grid Co.,Ltd.,Kunming 650217,China;Faculty of Electric Power Engineering,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]云南电网有限责任公司电力科学研究院,昆明650217 [2]昆明理工大学电力工程学院,昆明650500

出　　处：《高压电器》2023年第8期61-69,共9页High Voltage Apparatus

基　　金：国家自然科学基金资助项目(51907084);云南电网有限责任公司科技项目(YNKJXM20220011);云南省应用基础研究计划(202101AT070080)。

摘　　要：变压器作为电网传输和变换电能的主要设备,对DGA数据进行异常分析,可为变压器故障诊断提供理论依据。鉴于此,文中提出了基于DGA和IPSO-XGBoost的变压器故障诊断方法。首先,将特征气体划分为无编码比值作为特征参量输入极端梯度提升(XGBoost)模型,提出了基于XGBoost的变压器故障诊断模型;其次,通过动态调整惯性权重和加速因子对粒子群算法(PSO)进行改进,并利用改进的粒子群算法(IPSO)对XGBoost的关键参数进行迭代优化;最后,随机抽取1614例故障类型已知的DGA数据进行算例分析。结果表明:相比于其它传统机器学习分类模型,XGBoost的变压器故障诊断正确率更高,且与传统PSO算法相比,所提方法可以更好克服粒子群寻优速度慢和易陷入局部最优等问题,可为变压器安全稳定运行提供有力保障。Transformer is the main equipment for power transmission and conversion in power grid.Abnormal analy-sis of DGA data can provide a theoretical basis for fault diagnosis of transformer.In view of this,the fault diagnosis method of transformer based on dissolved gas analysis(DGA)and improve particle swarm optimization(ipso)-ex-treme gradient boosting(XGBoost)is proposed in this paper.Firstly,the characteristic gas is divided into non-coding ratios as characteristic parameters to input XGBoost model,and the fault diagnosis model of transformer based on XG-Boost is proposed.Secondly,the PSO algorithm is improved by dynamically adjusting the inertia weight and accelera-tion factor,and the IPSO algorithm is used to optimize the key parameters of XGBoost.Finally,the known DGA data of 1614 cases of fault types is sampled randomly for calculation analysis.The results show that compared with other traditional machine learning classification models,the fault diagnosis accuracy of transformer based on XGBoost is higher.Compared with the traditional PSO algorithm,the proposed method can better overcome such problems as slow particle swarm optimization speed and easy to fall into local optimum,which can provide a strong guarantee for safe and stable operation of transformer.

关 键 词：变压器 故障诊断 极端梯度提升 粒子群算法 无编码比值 

分 类 号：TM407[电气工程—电器] TP18[自动化与计算机技术—控制理论与控制工程]