基于PWKNN算法的风电系统故障诊断研究  

Research on Fault Diagnosis of Wind Power System Based on PWKNN Algorithm

作　　者：乐天达 赵强[1] 章志鸿[1] 李志明[1] 童文华 李欣哲[1] LE Tianda;ZHAO Qiang;ZHANG Zhihong;LI Zhiming;TONG Wenhua;LI Xinzhe(Design Center,State Grid Wuxi Power Supply Company,Wuxi 214000,China)

机构地区：[1]国网无锡供电公司设计中心,江苏无锡214000

出　　处：《微型电脑应用》2024年第1期188-192,共5页Microcomputer Applications

摘　　要：为了准确诊断风电系统故障类别,基于改进加权k近邻的粒子群优化算法(PWKNN)提出一种新的诊断方法。PWKNN通过调整权重来反映特征的重要性,并利用距离判断策略计算出多类标分类的相同概率。采用粒子群优化算法(PSO)优化了PWKNN的权值和参数k,利用特征提取训练分类器,结合特征选择的Pearson相关系数来消除无关特征,从而减少分类器的输出时间。对300W风力发电机的四种分类状态进行测试,与传统分类器的比较表明,PWKNN具有更高的分类精度。特征选择可以将平均特征数量从16个减少到2.8个,输出时间可以减少61%。In order to accurately diagnose the fault category of wind power system,a diagnosis method based on improved weighted k-nearest neighbor particle swarm optimization algorithm(PWKNN)is proposed in this paper.PWKNN reflects the importance of features by adjusting the weight,and the distance judgment strategy is used to calculate the same probability of multi class classification.The weight and parameter k of PWKNN are optimized by particle swarm optimization(PSO)algorithm.The classifier is trained by feature extraction,combined with the Pearson correlation coefficient of feature selection irrelevant features are eliminated,the output time of the classifier is reduced.Four classification states of 300W wind turbine are tested.The comparison with the traditional classifier shows that the improved PWKNN has higher classification accuracy.Feature selection can reduce the average number of features from 16 to 2.8,and the output time can be reduced by 61%.

关 键 词：粒子群优化 故障诊断 分类 特征选择 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TM315[自动化与计算机技术—计算机科学与技术]