微型电脑应用2026，Vol.42Issue(4)：119-123,5.
基于改进CUSUM和CNN的非侵入式负荷检测方法
Non-invasive Load Detection Method Based on Improved CUSUM and CNN
摘要
Abstract
To improve the accuracy of non-invasive load recognition,a non-invasive load detection method based on improved cu-mulative sum(CUSUM)and convolutional neural network(CNN)is proposed.By considering the difficulty of detecting tran-sient features during the switching process of electrical appliances,power change values are introduced into the traditional CU-SUM time detection method to improve the sensitivity of event detection.The classifier of CNN is improved,and the traditional Softmax classifier is changed in to clustering algorithm,after that improving the accuracy of recognizing similar transient fea-tures.The testing results show that the accuracy of the improved CUSUM event detection algorithm on the AMPds2 data set are 91.44％and 97.35％,respectively.On the BLUED data set,the improved event detection method has a lower number of missed and false detections.The load recognition method based on improved CNN achieves recognition precision,recall and F1 value of 92.58％,93.41％and 0.9299 on the AMPds2 data set,respectively,which are higher than the recognition methods of SVM,BP,and standard CNN.From this,it can be concluded that the proposed method has good performance on the recogni-tion of non-invasive load and has certain effectiveness.关键词
非侵入式负荷/AMPds2数据集/电力设备/CNN负荷识别/CUSUM事件检测Key words
non-invasive load/AMPds2 data set/power equipment/CNN load recognition/CUSUM event detection分类
信息技术与安全科学引用本文复制引用
李成学,刘永光..基于改进CUSUM和CNN的非侵入式负荷检测方法[J].微型电脑应用,2026,42(4):119-123,5.基金项目
国家电网公司科技项目(SGXJDK00PJJS2000062) （SGXJDK00PJJS2000062）
