基于粒子群算法优化支持向量机的变压器绕组变形分类方法

变压器是电网最为核心的设备,绕组变形是变压器主要的故障类型之一,频率响应分析法(frequency response analysis, FRA)是目前广泛应用的绕组变形检测方法。为提高绕组变形分类诊断的性能,文中提出基于粒子群算法优化支持向量机(particle swarm optimization-support vector machine, PSO-SVM)的变压器绕组变形分类方法,采用数学统计方法提取频率响应曲线的特征参量,并输入到支持向量机模型进行训练,利用粒子群算法优化支持向量机模型参数,使其能够有效区分不同的绕组故障类型。为证明文中方法在变压器绕组故障诊断方面的有效性,在一台特制模型变压器上进行了一系列故障模拟实验。数据处理结果表明,训练后的支持向量模型表现出了极高的性能,并且,相比传统的网格搜索参数优化算法,粒子群算法优化的支持向量机可以显著提高变压器绕组变形故障的分类性能。

Transformer Winding Deformation Classification Method Based on Particle Swarm Algorithm Optimizing Support Vector Machine

Transformer is the core equipment of power grid,winding deformation is one of transformer common faults,and frequency response analysis is a widely used winding deformation detection method.In order to improve the performance of classifying and diagnosing winding deformation,this paper proposes a transformer winding deformation classification method based on particle swarm algorithm optimizing support vector machine,the mathematical statistical method is used to extract the characteristic parameters of frequency response curves,which are imported into SVM model to perform the training.The particle swarm algorithm is used to optimize the parameters of SVM model,to effectively distinguish variable winding fault types.In order to prove the effectiveness of the presented method in diagnosing transformer winding faults,a series of fault emulating experiments are carried out on a specially manufactured model transformer.The data process result indicates that the SVM model after training presents high performance;moreover,compared to the traditional grid searching parameter optimization method,the SVM which is optimized by particle swarm algorithm can remarkably improve the performance of classifying transformer winding deformation faults.