Non-intrusive Load Monitoring Based on Convolutional Neural Network Mixed Residual Unit

Abstract

Non-intrusive Load Monitoring(NILM) is designed to determine the type of electrical appliance after decomposing the total power into each appliance in order to analyze how much electricity was consumed. In the traditional pattern recognition, relying on artificial extraction features, there is a large amount of unreliability, and then the use of convolutional neural networks to automatically extract features, but in the traditional CNN structure, when the network level is deepened, training error and test error will increase. Therefore, this paper proposes a network structure with residual unit added to the traditional convolutional neural network, in order to reduce the probability of gradient explosion. The dataset of the paper is the data collected from 22 different electrical appliances, and its current and voltage trajectory map is used to identify the accuracy rate of 97.9%.