基于RQA与DAGSVM的电能质量扰动识别方法  

Method of Power Quality Disturbance Recognition Based on RQA and DAGSVM

作　　者：陈武 钟建伟 杨永超 梁会军 CHEN Wu;ZHONG Jian-wei;YANG Yong-chao;LIANG Hui-jun(College of Intelligent Systems Science and Engineering,Hubei Minzu University,Enshi Hubei 445000,China)

机构地区：[1]湖北民族大学智能科学与工程学院,湖北恩施445000

出　　处：《计算机仿真》2025年第1期52-56,共5页Computer Simulation

基　　金：国家自然科学基金(62163013);湖北省自然科学基金(2021CFB542,2022CFB264);国网湖北省电力有限公司2022科技项目资助(5215P0220001)。

摘　　要：针对电能质量扰动(power quality disturbance, PQD)随机多变导致的特征交叉及分类性能不足的问题,提出了一种递归定量分析(recurrence quantification analysis, RQA)与有向无环图支持向量机(directed acyclic graph support vector machine, DAGSVM)的PQD分类新方法。首先利用基于复杂网络理论的递归定量分析法定量刻画电能质量扰动的递归图,并构造特征矩阵;其次通过DAGSVM构建PQD分类模型;最后采用基于发现学习的教与学优化算法优化PQD分类器的惩罚系数和核函数参数从而提升PQD分类器性能。结果表明,上述方法对PQD具有较高的识别准确率和良好的抗噪性。Regarding the problem of feature crossover and insufficient classification performance caused by stochastic variation of power quality disturbance(PQD),a new classification method of PQD based on recurrence quantification analysis(RQA)and directed acyclic graph support vector machine(DAGSVM)is proposed.Firstly,the recurrence plot of power quality disturbance is quantitatively described by using the RQA based on complex network theory,and the characteristic matrix is constructed.Secondly,the classification model of PQD is built based on DAGSVM.Finally,the teaching-learning-based optimization algorithm with discovery learning(TLBO/DL)is used to optimize the penalty coefficient and kernel function parameters to improve the performance of the PQD classifier.The results show that the method has high recognition accuracy and good anti-noise performance for PQD.

关 键 词：电能质量扰动信号 分类 教与学优化算法 递归定量分析 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]