基于时频域多特征和优化KELM的电能质量扰动检测  

Power Quality Disturbance Detection Based on Time-frequency Domain Multi-features and Optimized KELM

作　　者：徐琳 范松海 赵淳 隗震[2,3] 刘畅 XU Lin;FAN Songhai;ZHAO Chun;Wei ZHEN;LIU Chang(Electric Power Research Institute of State Grid Sichuan Electric Power Company,Chengdu 610041,China;NARI Group Corporation(State Grid Electric Power Research Institute),Nanjing 211000,China;Wuhan NARI Limited Liability Company,State Grid Electric Power Research Institute,Wuhan 430079,China)

机构地区：[1]国网四川省电力公司电力科学研究院,四川成都610041 [2]南瑞集团有限公司(国网电力科学研究院有限公司),江苏南京211000 [3]国网电科院武汉南瑞有限责任公司,湖北武汉430079

出　　处：《山东电力技术》2025年第3期59-67,共9页Shandong Electric Power

基　　金：国家电网有限公司科技项目“农网台区末端融合感知、智能诊断及服务提升技术研究及应用”(52199922000M)。

摘　　要：电能质量扰动的准确分类是改善和治理电能质量的前提。为提高电能质量快速检测的准确性,提出一种基于时频多特征和改进核极端学习机(kernel extreme learning machine,KELM)的电能质量扰动(power quality disturbance,PQD)分类方法。该方法首先利用小波变换和S变换提取各电能质量扰动信号的特征量,然后根据提取的特征量构造具有分类规则的KELM模型,并使用混沌粒子群优化(chaotic particle swarm optimization,CPSO)对KELM的参数进行自适应优化。实例仿真结果和对比分析表明,该方法能有效识别7种常见的单一扰动信号和10种复合扰动信号,并且抗噪声能力更强,分类精度高于KELM和PSO-KELM模型。该方法为电能质量的改善和治理提供了新的思路和方法。Accurate classification of power quality disturbances is the premise for improving and controlling power quality.In order to improve the accuracy of rapid detection of power quality,this paper proposes a power quality disturbance(PQD)classification method based on time-frequency multi-features and improved kernel extreme learning machine(KELM).This method first uses wavelet transform and S transform to extract the feature quantities of each power quality disturbance signal.Then a KELM model with classification rules is constructed based on the extracted feature quantities,and chaotic particle swarm optimization(CPSO)is used to adaptively optimize the parameters of KELM.Example simulation results and comparative analysis show that this method can effectively identify seven kinds of common single disturbance signals and ten kinds of compound disturbance signals.The proposed method has stronger anti-noise ability,and its classification accuracy is higher than the KELM and PSO-KELM models,providing new ideas for the improvement and management of power quality.

关 键 词：电能质量扰动分类 时频多特征 混沌粒子群优化 核极限学习机 

分 类 号：TM93[电气工程—电力电子与电力传动]