低压故障数据泛化分析及基于ISSA-WELM的预测模型研究  

Generalization Analysis of Low-voltage Fault Data and Research on Prediction Model Based on ISSA-WELM

作　　者：王伟 顾洪源 吴正盛 黎达理 何青 刘骅霆 薛炎龙 WANG Wei;GU Hongyuan;WU Zhengsheng;LI Dali;HE Qing;LIU Huating;XUE Yanlong(Nanning Power Supply Bureau of Guangxi Power Grid Co.,Ltd.,Guangxi Nanning 530000,China;Luoyang Sanlong Installation&Overhaul Co.,Ltd.,Henan Luoyang 471000,China)

机构地区：[1]广西电网有限责任公司南宁供电局,广西南宁530000 [2]洛阳三隆安装检修有限公司,河南洛阳471000

出　　处：《广西电力》2024年第5期32-39,共8页Guangxi Electric Power

摘　　要：本文采用逻辑树、聚类以及相关性等分析方法,对2020年2023年度低压配电网故障进行数据泛化分析,从故障主次原因、责任归属占比、低压问题改进策略等方面进行探讨研究,提出了一种改进的麻雀算法(Improved SparrowSearch Algorithm,ISSA)和加权极限学习机(Weighted Extreme Leaning Machine,WELM)优化算法预测模型。为验证改进模型的预测性能,本文选择极限学习机(Extreme Leaning Machine,ELM)、麻雀算法优化极限学习机(ISSA-ELM)、粒子群算法优化极限学习机(Particle Swarm Optimization-Optimized Extreme Leaning Machine,PSO-ELM)、灰狼算法优化极限学习机(Grey Wolf Optimizer optimized Extreme Leaning Machine,GWO-ELM)、误差反向传播算法(Back Propagation,BP)、支持向量回归(Support Vector Regression,SVR)、随机森林(Random Forest,RF)和径向基函数网络(Radial BasisFunction Network,RBF)共8个算法进行对比。仿真结果显示,ISSA-WELM模型具备优秀的泛化能力和预测精度,能实现对低压故障工单的实时检测,提升数据预测的准确性,对后续低压运检工作具有指导意义。In this paper,the analysis methods of logic tree,clustering and correlation are used to generalize the data of lowvoltage distribution network faults from 2020 to 2023.The main and secondary causes of faults,the proportion of responsibility attribution,and the improvement strategy of low-voltage problems are discussed and studied.Meanwhile,a prediction model of improved Sparrow Search Algorithm(ISSA)and Weighted Extreme Learning Machine(WELM)optimization algorithm is proposed.In order to verify the prediction performance of the improved model,this paper compares eight algorithms,including Extreme Learning Machine(ELM),Sparrow Algorithm Optimisation Extreme Learning Machine(ISSA-ELM),Particle Swarm Optimisation Extreme Learning Machine(PSO-ELM),Grey Wolf Optimisation Extreme Learning Machine(GWO-ELM),Error Backpropagation Algorithm(BP),Support Vector Regression(SVR),Random Forest(RF)and Radial Basis Function Network(RBF).The simulation results show that the ISSA-WELM model has excellent generalization ability and prediction accuracy,which realizes the real-time detection of low-voltage fault work orders,improves the prediction accuracy of the data,and provides guiding significance for the subsequent low-voltage operation and repair work.

关 键 词：低压配电网 故障数据泛化 逻辑树 聚类 相关性 ISSA-WELM模型 

分 类 号：TM73[电气工程—电力系统及自动化] TP181[自动化与计算机技术—控制理论与控制工程]