基于萤火虫扰动麻雀搜索算法-极限学习机的光伏阵列故障诊断方法研究  

Fault Diagnosis of Photovoltaic Arrays Based on Sparrow Search Algorithm With Firefly Perturbation-extreme Learning Machine

作　　者：赵靖英[1] 吴晶晶[1] 张雪辉[1] 张文煜 姚帅亮 ZHAO Jingying;WU Jingjing;ZHANG Xuehui;ZHANG Wenyu;YAO Shuailiang(State Key Laboratory of Reliability and Intelligence of Electrical Equipment(Hebei University of Technology),Beichen District,Tianjin 300401,China;State Grid Hebei Zhangjiakou Wind Power Storage and Transmission New Energy Co.,Ltd.,Zhangjiakou 075000,Hebei Province,China)

机构地区：[1]省部共建电工装备可靠性与智能化国家重点实验室(河北工业大学),天津市北辰区300401 [2]国网冀北张家口风光储输新能源有限公司,河北省张家口市075000

出　　处：《电网技术》2023年第4期1612-1622,共11页Power System Technology

基　　金：国家自然科学基金项目(51377044);河北省自然科学基金项目(E2019202481)。

摘　　要：光伏阵列具有随机性、间歇性输出特点,发生故障严重影响电力系统安全运行。针对有效表征不同程度局部阴影与雨天接地故障的故障特征量缺乏的问题,分析不同故障状态下光伏阵列运行特征,提出一种新的6维故障特征向量:开路电压Uoc、最大功率点电压Um与短路电流Isc、最大功率点电流Im分别表征短路与断路故障;U-I特性曲线二阶导数零点数表征局部阴影故障,并利用遗传模拟退火算法优化的模糊C均值聚类算法(the fuzzy C-means clustering algorithm optimized by the genetic simulated annealing algorithm,GSA-FCM)验证Um、Im表征不同程度局部阴影故障的有效性;并网电流总谐波畸变率表征雨天接地故障。引入萤火虫扰动的麻雀搜索算法(sparrow search algorithm with firefly perturbation,FSSA)优化传统极限学习机(extreme learning machine,ELM),建立FSSA-ELM模型,解决传统故障诊断方法实现复杂、收敛速度慢的问题。基于现场数据驱动,建立考虑对地寄生电容的光伏系统仿真模型和实验平台,设计2种不同辐照度范围的仿真方案和实验方案进行方法验证,结果表明,FSSA-ELM模型结合ELM实现简单且FSSA收敛速度快的特点,利用6维故障特征向量,可准确识别光伏阵列故障类型。With the characteristics of random and intermittent outputs,the faults of the serious photovoltaic arrays affects the safe operation of the system.In view of the lack of fault characteristic parameters that effectively characterize the different degrees of a partial shading or a ground fault in the rainy days,the operating characteristics of the photovoltaic arrays under different faults is analyzed,and a new 6-dimensional fault feature vector is proposed:The open circuit voltage Uoc,the maximum power point voltage Um and short circuit current Isc,the maximum power point current Im are introduced to characterize the short circuit and the open circuit faults respectively.The number of zero points of the second derivative of the U-I curve is introduced to characterize the partial shading.And the fuzzy c-means clustering algorithm optimized by the genetic simulated annealing algorithm(GSA-FCM)is used to verify the validity of the different degrees of the partial shading characterized by the Um and Im.The total harmonic distortion rate of the grid-connected current is introduced to characterize the ground fault in rainy days.The sparrow search algorithm with firefly perturbation(FSSA)is introduced to optimize the traditional extreme learning machine(ELM).An FSSA-ELM model is established to solve the problems such as complex implementation and slow convergence in the traditional fault diagnosis methods.Driven by the field data,the simulation model of a photovoltaic system with the parasitic capacitance to the ground and the experimental platform are developed.Two schemes with different irradiance ranges and experimental schemes are designed to verify the method.The results show that the FSSA-ELM model,combined with the simple implementation of the ELM and the fast convergence speed of the FSSA,accurately identifies the fault types of the photovoltaic arrays by using the 6-dimensional fault feature vector.

关 键 词：光伏阵列 故障诊断 并网电流总谐波畸变率 故障特征量 萤火虫扰动麻雀搜索算法–极限学习机 寄生电容 

分 类 号：TM721[电气工程—电力系统及自动化]