摘要
针对电能质量复合扰动分类方法在分类数目和分类性能方面存在的不足,提出一种基于改进经验小波变换(MEWT)和极限梯度提升(XGBoost)的电能质量复合扰动分类方法。首先,对传统经验小波变换进行改进,使之适用于复合扰动特征提取;然后,根据基本扰动MEWT分析结果,从时频域多角度提取能够有效刻画不同扰动特性的特征序列;最后,基于问题转换策略构造以XGBoost为子分类器的多标签复合扰动分类模型,并通过特征选择与超参数优化相结合的模型训练方法进一步提升分类效果。实验结果表明,所提方法可实现48类扰动的有效辨识,较之传统多标签扰动分类方法在分类精度和噪声鲁棒性方面表现更优,且运算速度更快,适用于工程实践。
Aiming at the shortcomings of multiple power quality disturbances classification method in terms of classification number and classification performance,a recognition method of multiple power quality disturbances based on modified empirical wavelet transform(MEWT)and extreme gradient boosting(XGBoost)was proposed.Firstly,the traditional empirical wavelet transform was improved to make it suitable for feature extraction of multiple disturbances.Then,according to the MEWT analysis results of basic disturbances,feature sequences which can effectively depict different disturbance characteristics were extracted from time and frequency domain.Finally,a multi label classification model for multiple disturbances with XGBoost as sub classifier was constructed based on the problem transformation strategy,and the model training method combining feature selection and hyperparameter optimization was used to further improve the effect of classification.The experimental results show that the proposed method can effectively identify 48 types of disturbances.Compared with the traditional multi label disturbance classification method,the proposed method performs better in classification accuracy and noise robustness,and has faster operation speed,which is suitable for engineering practice.
作者
吴建章
梅飞
郑建勇
张宸宇
缪惠宇
Wu Jianzhang;Mei Fei;Zheng Jianyong;Zhang Chenyu;Miao Huiyu(School of Electrical Engineering Southeast University,Nanjing 210096 China;College of Energy and Electrical Engineering Hohai University,Nanjing 211100 China;Suzhou Research Institute of Southeast University,Suzhou 215123 China;State Grid Jiangsu Electric Power Co.Ltd Research Institute,Nanjing 211103 China)
出处
《电工技术学报》
EI
CSCD
北大核心
2022年第1期232-243,253,共13页
Transactions of China Electrotechnical Society
基金
江苏省重点研发计划(BE2020027)
国家电网公司科技(52199918000C)
国家重点研发计划(2018YFB0905000)资助项目。
关键词
电能质量复合扰动
经验小波变换
尺度空间表示
多标签分类
极限梯度提升
Multiple power quality disturbances
empirical wavelet transform
scale-space representation
multi label classification
XGBoost
