摘要
为了快速准确地诊断变压器故障,提出一种基于量子粒子群优化的快速相关向量机(quantum particle swarm optimized fast relevance vector machine,QPSO-FRVM)变压器故障诊断模型。首先建立了快速相关向量机多层次分类模型,在此基础上提出劣化度故障特征提取方法;其次分析了影响相关向量机分类性能的2个因素,借助量子粒子群算法确定每一层的核函数参数以及故障特征提取方法。最后利用训练好的QPSO-FRVM模型进行变压器的故障诊断,并与IEC三比值法、SVM模型进行对此。仿真结果表明,FRVM缩短了训练时间,具有比RVM更高的诊断效率;同时在小样本情况下,对核参数和特征提取方法均进行优化选择的QPSO-FRVM模型,具有比IEC三比值法和SVM模型更高的诊断准确率,为实现变压器快速准确的故障诊断提供一种新的参考。
Quantum particle swarm optimized （QPSO） feature selection methods-fast relevance vector machine （FRVM） based transformer fault diagnosis model is proposed in this paper. First, multi-layer classifier model based on fast relevance vector machine was built, then deteriotion degree feature selection method was proposed; parameters of kernel functions and fault feature extraction method were two significant factors influencing the performance of the classifier and were optimized by QPSO method. The numerical examples testify that FRVM decreased training time a lot than RVM and the proposed model optimizing both the feature selection methods and kernel function parameters has higher classification accuracy than IEC three ratio and SVM model.
出处
《电网技术》
EI
CSCD
北大核心
2013年第11期3262-3267,共6页
Power System Technology
基金
山东省科技发展计划项目(2012GGE27004
2012GQX20115)~~
关键词
相关向量机
量子粒子群优化算法
故障诊断
特征选择
relevance vector machine quantum particle swarm optimization fault diagnosis feature selection
