Research on Dual Model Non-intrusive Load Identification Based on Convolutional Neural Network and BP Neural Network

As an important branch of intelligent electricity use, power load identification realizes accurate identification of power load, which can further improve the intelligent electricity use system. Aiming at the problems of singularity, inaccuracy and unreliability of electric load identification, this paper proposed a method based on deep learning. the change value of starting-stopping load and current was determined after analyzing for collected household appliances. The load data set of household appliance was collected followed by further refining of the superimposed load state which was trained by using current characteristics and load characteristics, respectively, by combining with convolutional neural network CNN and BP neural network. This paper was designed to find the best model suitable for load identification to improve the identification rate, enhance the reliability and accuracy of load identification.