Model for Prediction of Energy Consumption in Residential Buildings Based on Transfer Learning

As Climate change has become a major issue worldwide, the importance of building energy management systems (BEMS) is increasing. Because of obligation of public institutions to introduce BEMS and increase in BEMS installation due to the increase in smart buildings, It has become essential to understand and analyze about energy consumption to predict it in order to reduce carbon emissions and strengthen energy-saving policies. Through many previous studies, several analysis models have been proposed that can accurately predict energy consumption, and these models have shown excellent performance in each study. However, to obtain interesting results, a sufficient amount of training data is required. Obtaining a satisfactory amount of data sets generally requires a lot of time and effort, and sometimes it is impossible. This is an obstacle to applying deep learning models to many tasks. In this study, we propose a method for supplementing insufficient performance by transferring knowledge from similar types of data so that deep learning models can perform well even with sparse data sets.