Non-intrusive Load Monitoring Using Inception Structure Deep Learning

Abstract

Non-intrusive load monitoring is an important technology to realize energy efficiency tracking and smart electricity consumption, in which the load identification method is very important. In order to realize non-intrusive load monitoring, we pro pose a method based on deep neural network. The dataset is derived from the total energy consumption, sub-consumption and weather data collected from the buildings in the commercial building park in Shanghai, China. First, we perform data cleaning to mi ne the basic characteristics of the commercial building load. Second, we propose a deep learning model based on the Inception structure that combines a Multi-Layer Perceptron, Convolutional Neural Networks, and Long Short-Term Memory to address the problems of overfitting and long computation time of the deep learning model. Finally, a comparison with a common deep learning network model is made on a test data set to verify the effectiveness and accuracy of the proposed method.