基于蜻蜓算法优化ELM的电能质量扰动诊断与识别研究  

Study on the Optimization of Power Quality Disturbance Diagnosis and Identification of ELM Based on Dragonfly Algorithm

作　　者：王宏刚[1] 田洪迅[1] 谈军[2] 周辉[3] WANG Honggang;TIAN Hongxun;TAN Jun;ZHOU Hui(State Grid Corporation of China,Beijing 100000,China;NARI Group Corporation (State Grid Electric Power Research Institute) ,Nanjing 210000,China;State Grid Zhejiang Electric Power Company,Hangzhou 310000,China)

机构地区：[1]国家电网公司,北京100000 [2]南瑞集团(国网电力科学研究院),南京210000 [3]国网浙江省电力公司,杭州310000

出　　处：《电力电容器与无功补偿》2019年第1期142-147,共6页Power Capacitor & Reactive Power Compensation

摘　　要：针对极限学习机预测结果易受其初始化输入权值和偏置值的影响,提出一种DA算法优化ELM的电能质量扰动诊断和识别模型。选择5种电能质量扰动信号为研究对象,研究结果表明,与GA_ELM、PSO_ELM和ELM相比,本文提出的算法DA_ELM可以有效提高电能质量扰动识别的准确率,为电能质量扰动诊断和识别提供新的方法和途径。In view of the predicted result of the extreme learning machine being likely susceptible to the initial input weights and the offset value, a kind of DA algorithm is proposed to optimize power quality disturbance diagnosis and model identification of ELM.Five kinds of power quality disturbance signals are chosen as the study object. It is shown by the study result that compared with GA_ELM, PSO-ELM and ELM, the algorithm DA_SVM proposed in this paper can effectively improve the accuracy of the power quality disturbance identification, have fast convergence rate and provide new methods and means for the power quality disturbance diagnosis and identification.

关 键 词：蜻蜓算法 极限学习机 电能质量 样本熵 小波分解 

分 类 号：TM744[电气工程—电力系统及自动化]