基于粒子群与多分类相关向量机的变压器故障诊断

Transformer Fault Diagnosis Based on Particle Swarm Optimization and Multi-classification Correlation Vector Machine

为有效提高变压器故障诊断准确率,文中将粒子群优化算法与多分类相关向量机方法相结合,构建了一种基于粒子群优化多分类相关向量机的方法用以变压器故障诊断。该方法首先将油中溶解气体与4种特征气体比值相结合作为故障特征量,以进一步丰富故障信息。其次,利用粒子群优化算法并结合训练样本数据对多分类相关向量机的核参数进行优化,以获得能够最优的及能有效提高故障分类效率的参数。最后,将9种特征量作为特征输入,并利用已训练完毕的多分类相关向量机进行故障诊断。经实例分析表明,该方法使故障特征量与故障分类模型得到了有效补充、改进及完善,且其故障诊断效率更具优势。

In order to improve the accuracy of transformer fault diagnosis effectively,this paper combines particleswarm optimization（PSO）with multi-classification correlation vector machine to construct a method based on multi-classification correlation vector machine of particle swarm optimization for transformer fault diagnosis. The methodfirstly combines the dissolved gas in oil with the ratio of four characteristic gases as fault characteristic variables tofurther enrich the fault information. Secondly,the kernel parameters of the multi-classification correlation vector ma-chine are optimized by PSO algorithm and training sample data to obtain the parameters which are optimal and canimprove the efficiency of fault classification effectively. Finally,nine characteristic variables are used as feature in-puts,and the trained multi-classification correlation vector machines are used for fault diagnosis. The case analysisshows that this method can effectively supplement and improve the fault characteristic variables and fault classifica-tion model,and its fault diagnosis efficiency is more advantageous.