Research on the generalization of resident load Identification based on Hybrid Branch Network

Abstract Household energy consumption has become an important part of energy consumption, and improving its utilization plays an important role in solving the global energy crisis. As an effective means, load identification technology promotes energy saving on the user side by providing detailed power consumption information. In this paper, aiming at the generalization of power supply voltage fluctuation in load identification, electrical appliances are classified into four categories: time-varying load, linear load, constant power load, and other nonlinear loads. A load identification model with many kinds of neural network branches is proposed. According to the characteristic differences of different types of loads, different branches are selected for different types of electrical appliances for identification. Finally, compared with the recognition accuracy of other common models in different environments, it is concluded that the method in this paper has higher recognition ability and generalization ability.