基于Kmeans++聚类的光伏系统直流电弧故障检测研究  

RESEARCH ON DC ARC FAULT DETECTION OF PV SYSTEM BASED ON Kmeans++CLUSTERING

作　　者：邬洲 张军 李隆 李宝龙 陈辉 Wu Zhou;Zhang Jun;Li Long;Li Baolong;Chen Hui(College of Automation Engineering,Shanghai University of Electric Power,Shanghai 200090,China;Shanghai JA Solar Technology Co.,Ltd.,Shanghai 200436,China;Zhejiang Power Transmission and Transformation Engineering Co.,Ltd.,Hangzhou 310016,China)

机构地区：[1]上海电力大学自动化工程学院,上海200090 [2]上海晶澳太阳能科技有限公司,上海200436 [3]浙江省送变电工程有限公司,杭州310016

出　　处：《太阳能学报》2024年第11期320-329,共10页Acta Energiae Solaris Sinica

基　　金：上海市科委科技创新双碳项目(21DZ1207300)。

摘　　要：针对光伏系统中直流侧串联电弧故障由于信号微弱且具有强烈的随机性,从而导致故障不易识别的问题,提出基于互补集合经验模态分解(CEEMD)与K均值聚类(Kmeans++)相结合的故障检测方法。首先,通过使用CEEMD将光伏系统直流侧电流信号分解为若干个本征模态分量(IMF),然后使用皮尔逊相关系数来筛选有效的模态分量以进行信号的重构。其次,对重构后的信号进行时频域特征提取,并应用Kmeans++进行故障识别。实验结果表明,采用所提方法能有效地检测故障。鉴于实际光伏系统运行的复杂性,研究不同外部干扰对电弧检测算法的影响,并通过实验数据验证该方法在抗干扰性方面的优越性。最后,与基于PNN和SVM的故障检测方法进行比较,验证了所提电弧故障检测方法的有效性。A fault detection method based on the combination of complementary set empirical mode decomposition(CEEMD)and Kmeans clustering(Kmeans++)is proposed to address the issue of difficult identification of DC side series arc faults in photovoltaic systems due to weak signals and strong randomness.Firstly,the DC side current signal of the photovoltaic system is decomposed into several intrinsic mode functions(IMF)using CEEMD,and then Pearson correlation coefficients are used to filter out effective modal components for signal reconstruction.Secondly,extract time-frequency domain features of the reconstructed signal and apply Kmeans++for fault identification.The experimental results show that the proposed method can effectively detect faults.Given the complexity of actual photovoltaic system operation,the influence of different external disturbances on arc detection algorithms was studied,and the superiority of this method in anti-interference was verified through experimental data.Finally,the proposed method was compared with fault detection methods based on PNN and SVM,and experimental data was used to verify its good fault recognition accuracy.

关 键 词：光伏效应 电弧 故障检测 模态分解 特征提取 聚类分析 

分 类 号：TM501.2[电气工程—电器]