摘要
负荷识别是非侵入式负荷监测的关键环节.针对原始电压电流轨迹特征选择有限、识别准确度低的问题,提出一种基于改进YOLOv5s(YOLOv5s是YOLOv5(you only look once的第5个版本)系列中预训练结构最小的模型)的非侵入式负荷识别算法.将坐标注意力(coordinate attention,简称CA)模块添加至YOLOv5s的主干网络,用双向特征金字塔网络(bi-directional feature pyramid network,简称BiFPN)取代YOLOv5s的常规特征提取网络.实验结果表明:相对于其他3种算法,该文算法有更高的负荷识别准确度.因此,该文算法具有有效性.
Load identification is the key link of non-intrusive load monitoring.Aiming at the problem of limited selection of original voltage and current track features and low recognition accuracy,a non-intrusive load identification algorithm based on improved YOLOv5s(YOLOv5s is the model with the smallest pre-training structure in YOLOv5(the fifth version of you only look once)series)was proposed.The coordinate attention(CA)module was added to the backbone network of YOLOv5s,and the bi-directional feature pyramid network(BiFPN)was used to replace the conventional feature extraction network of YOLOv5s.Experimental results showed that compared with the other three algorithms,the proposed algorithm had higher accuracy of load recognition.Therefore,the algorithm of this paper was effective.
作者
李悦
程志友
程安然
姜帅
胡杰
LI Yue;CHENG Zhiyou;CHENG Anran;JIANG Shuai;HU Jie(School of Internet,Anhui University,Hefei 230039,China;Power Quality Engineering Research Center,Ministry of Education,Anhui University,Hefei 230601,China;School of Electronic and Information Engineering,Anhui University,Hefei 230601,China)
出处
《安徽大学学报（自然科学版）》
CAS
北大核心
2023年第5期51-57,共7页
Journal of Anhui University(Natural Science Edition)
基金
国家自然科学基金资助项目(61672032)
安徽省科技重大专项(18030901018)。
