OVERVIEW AND COMPARISON OF COMMON LOAD IDENTIFICATION MODELS FOR NON-INTRUSIVE LOAD DETECTION

Due to the increasing shortage of energy, people pay more attention to the energy conservation and environmental protection. By providing consumers with monitoring of individual device consumption, consumers can adjust their consumption habits to achieve energy conservation and emission reduction. One way to provide this capability is non-intrusive load monitoring model (NILM). The main challenge of NILM is to select a suitable identification model for load identification and to solve the low accuracy problem of some equipment identification. This paper implements a variety of common load identification models. By comparing the accuracy of various identification models, we obtain the optimal load identification model for multiple equipment combinations prediction. At the same time, we discuss two situations separately due to the different effect of load recognition mode between single load operation scenario and multi-load operation scenario. By comparing the recognition effect between different load recognition models and the recognition effect of various equipment, we provide suggestions for the training method of load recognition model to make the model training effect better.