基于AdaBoost.M2-ISSA-ELM算法的电力变压器故障诊断方法  

Power transformer fault diagnosis method based on AdaBoost.M2-ISSA-ELM algorithm

作　　者：王艳[1] 王寅初 赵洪山[1] 李伟[1] 连洪钵 康磊 WANG Yan;WANG Yinchu;ZHAO Hongshan;LI Wei;LIAN Hongbo;KANG Lei(School of Electrical and Electronic Engineering,North China Electric Power University,Baoding 071003,China)

机构地区：[1]华北电力大学电气与电子工程学院,河北保定071003

出　　处：《电力自动化设备》2024年第9期205-211,218,共8页Electric Power Automation Equipment

基　　金：国家自然科学基金资助项目(51807063)。

摘　　要：为提高电力变压器故障诊断精度,将集成学习和群体智能优化算法相结合,提出一种电力变压器故障诊断方法。使用极限学习机(ELM)作为基学习算法,构建集成学习框架下的基分类器,并针对ELM模型性能受参数初始化影响较大、易陷入局部最优问题,引入基于正弦优化的改进麻雀搜索算法(ISSA)优化相关参数,提高基分类器的分类性能。使用改进的自适应增强(AdaBoost.M2)算法构建集成学习模型,扩展基分类器的输出,并引入伪损失函数替代传统AdaBoost算法中的加权误差,以增强集成分类器综合表达能力,得到基于AdaBoost.M2-ISSA-ELM算法的电力变压器故障诊断模型,进一步提高模型识别精度。通过909组油中溶解气体分析(DGA)样本对所提方法进行实例分析,结果表明该方法具有较好的诊断精度和分类性能,能够实现电力变压器故障类型的准确识别。In order to improve the accuracy of power transformer fault diagnosis,a power transformer fault diagnosis method is proposed by combining ensemble learning and swarm intelligence optimization algorithm.Using the extreme learning machine(ELM)as the basic learning algorithm to construct the base classifier under the integrated learning framework,and aiming at the problem that the performance of the ELM model is greatly affected by parameter initialization and is prone to fall into local optimization,an improved sparrow search algorithm(ISSA)based on sinusoidal optimization is introduced to optimize relevant parameters,and improve the classification performance of the basic classifier.Then,an improved adaptive boosting(AdaBoost.M2)algorithm is used to build an ensemble learning model,expand the output of the base classifier,and the pseudo loss function is introduced to replace the weighted error in the traditional AdaBoost algorithm to enhance the comprehensive expression ability of the integrated classifier.A power transformer fault diagnosis model based on AdaBoost.M2-ISSA-ELM algorithm is obtained,which further improves the recognition accuracy of the model.The proposed method is analyzed through 909 sets of dissolved gases analysis(DGA)samples,and the results show that the method has good diagnostic accuracy and classification performance,and can achieve accurate identification of power transformer fault types.

关 键 词：电力变压器 故障诊断 集成学习 智能优化算法 极限学习机 

分 类 号：TM41[电气工程—电器]