Research On Load Identification Method Based On Non-intrusive Monitoring Technology

Non-intrusive load monitoring(Nonintrusive Load Mointoring,NILM)is to estimate the working and energy consumption status of each electrical equipment based on the electrical parameters at the power inlet,and obtain detailed power status information through load identification technology.Compared with intrusive load monitoring(Intrusive Load Mointoring,ILM)greatly reduces the number of sensors.In recent years,it has received more and more attention in the field of scheduling optimization and load refined management.With the application of deep learning in image recognition,natural language processing and other fields,deep learning algorithms only rely on low-frequency sampled load data to automatically learn features and achieve better load identification results.However,as the cost of complex network training has increased greatly,this article focuses on the application of deep learning algorithms and conducts in-depth research on load identification algorithms under NILM technology.First,on the basis of analyzing and summarizing the source and type of load data,it preprocesses the adopted data set,and gives a comprehensive NILM evaluation standard in terms of accuracy,generalization,scalability,and identification performance of different types of loads.Secondly,a load identification algorithm based on the convolutional block attention model is proposed.The attention mechanism module is applied to the convolutional neural network to improve the feature extraction ability of the convolutional neural network.By convolution based on multiple data sets The load identification experiment of the block attention model proves that the algorithm can further improve the accuracy of load identification without increasing the network parameters,and compares the evaluation standard with other NILM algorithms with the help of the toolkit to confirm the superiority of the algorithm in this paper.Finally,in view of the difficulty in obtaining label data in the actual application of NILM,the adversarial transfer learning network is used to process data sets with large differences in distribution samples,and the common features between different data sets are extracted,and the REDD,UK-DALE and REFIT Experiments are performed on the data set,and t he target domain data is decomposed in an unsupervised manner to achieve high recognition accuracy in cross-domain learning.