基于WWPA-SVM的光伏故障监测系统  

Photovoltaic Fault Monitoring System Based on WWPA-SVM

作　　者：董煌锋 肖文波[1,2,3] 吴华明 李勇波[1,2] 刘彬[3] DONG Huangfeng;XIAO Wenbo;WU Huaming;LI Yongbo;LIU Bin(Jiangxi Provincial Key Laboratory of Nanchang Hangkong University optoelectronic information sensing technology and instruments;Key Laboratory of Nanchang Hangkong University Nondestructive Testing Technology,Ministry of Education,Jiangxi,Nanchang 330063,China;School of Nanchang Hangkong University Science and Technology,Jiangxi Province,Gongqingcheng 332020,China)

机构地区：[1]南昌航空大学光电信息感知技术与仪器江西省重点实验室 [2]南昌航空大学无损检测技术教育部重点实验室,江西南昌330063 [3]南昌航空大学科技学院,江西共青城332020

出　　处：《惠州学院学报》2024年第6期94-102,128,共10页Journal of Huizhou University

摘　　要：提出一种基于貉藻优化算法(WWPA)优化的支持向量机(SVM)光伏故障监测系统。通过与19种其他优化算法的比较,结果表明WWPA-SVM在准确率、鲁棒性以及计算效率方面均表现出色。在正常数据集上,WWPA-SVM的平均识别准确率达到了99.70%,远超传统SVM的81.45%,同时耗时仅为12.19 s,显著低于其他算法。在不平衡数据集上,WWPA-SVM的平均准确率为99.90%,同样优于SVM的79.23%,耗时为22.55 s。即使在含故障值数据集和含高斯噪声数据集的挑战下,WWPA-SVM的平均准确率分别为95.91%和74.19%,依然保持领先,且耗时分别为9.83和8.40 s。此外,通过在其他光伏电池数据集和wine红酒数据库上的泛化性验证,WWPA-SVM展现了良好的泛化能力,平均准确率分别为96.53%和99.62%,耗时分别为0.17和0.14 s,进一步证实了WWPA-SVM在光伏故障监测中的有效性和优越性。This paper proposes a photovoltaic fault monitoring system based on a wolf weed optimization algorithm(WWPA)optimized support vector machine(SVM).Compared with 19 other optimization algorithms,the results show that WWPA-SVM performs well in terms of accuracy,robustness and computational efficiency.On the normal data set,the average recognition accuracy of WWPA-SVM reached 99.70%,far exceeding the traditional SVM's 81.45%,while taking only 12.19 seconds,significantly lower than other algorithms.On the imbalanced data set,the average accuracy of WWPA-SVM was 99.90%,also superior to SVM's 79.23%,and it took 22.55 seconds.Even when challenged with datasets containing fault values and Gaussian noise,the average accuracy of WWPA-SVM remained leading,with 95.91%and 74.19%respectively,with a time consumption of 9.83 seconds and 8.40 seconds.In addition,through the generalization verification on other photovoltaic cell data sets and wine databases,WWPA-SVM showed good generalization ability with an average accuracy of 96.53%and 99.62%respectively,and a time consumption of 0.17 seconds and 0.14 seconds.These results further confirm the effectiveness and superiority of WWPA-SVM in photovoltaic fault monitoring.

关 键 词：貉藻优化算法 支持向量机 故障监测 泛化能力 

分 类 号：TM615[电气工程—电力系统及自动化]