基于DGA的粒子群相关向量机变压器故障诊断

Fault Diagnosis of Transformer Using Relevance Vector Machine with Particle Swarm Optimization Based on DGA

传统智能算法中因算法自身的固有缺陷,从而导致变压器故障诊断结果不理想。为此,针对相关向量机中核函数参数的选取对分类效果产生影响的问题,笔者在对运用粒子群算法优化相关向量机的可行性进行充分分析的基础上,构建了粒子群优化的相关向量机方法,以DGA作为特征输入,利用粒子群优化算法对核函数参数σ进行优化,以获得最优的相关向量机故障诊断方法,从而提高变压器的故障诊断精度。实例对比分析表明,与SVM、RVM方法相比,粒子群相关向量机方法具有更高的诊断精度。

The inherent defects of the traditional intelligent algorithm results in unsatisfactory result of transformer fault diagnosis. In this paper, to eliminate the effect of parameter selection in relevance vector machine on classification results, the feasibility of using particle swarm optimization algorithm to optimize the relevance vector machine is analyzed, and a relevance vector machine method with particle swarm optimization is constructed. While DGA was considered as feature input,The particle swarm optimization algorithm is employed to optimize the kernel function parameter σ, hence to obtain an optimal fault diagnosis method with relevance vector machine and improve the precision of transformer fault diagnosis. Examples contrast and analysis show that, compared with support vector machine and relevance vector machine methods, the relevance vector machine method with particle swarm optimization has higher diagnostic accuracy.