Research on Non-Intrusive Load Decomposition Technology Based on Deep Learning Algorithm

Abstract

Load identification is an important part of non-intrusive load decomposition to realize smart electricity consumption. The electricity consumption information of each consumer can be decomposed from the current change information of the main meter, so as to provide more refined and targeted information for electricity consumers. Power consumption management and dispatch services. This paper uses the current effective value and the harmonic component information after Fourier transform to propose a load decomposition algorithm based on a one-dimensional convolutional neural network. It uses similarity comparison to decompose the current information of each consumer and solves the problem of newly added users or the problem of using electrical appliances to retrain. It solves the problems that the current non-invasive load decomposition algorithm using one-dimensional convolution has low decomposition accuracy, new user appliances need to be retrained, and high complexity. It is found through experiments that the method proposed in this paper can also improve the accuracy of load decomposition to a certain extent, and the complexity is low.