A Study on the Prediction of Building Energy Consumption using Deep Learning Technique

In this study, the energy use of buildings was compared and analyzed by using weather data predicted with machine running techniques. Python was used as a predictive program to predict weather data and TRNSYS was used to simulate the energy usage of buildings. For weather forecasting, weather data from 1 August to 7 August were studied to forecast ambient air temperature and solar radiation. The lowest error came in seven days, with the outside air temperature standing at 1.8 percent and the solar radiation at 2.4 percent. The energy use of the building was simulated by using weather data predicted through the 7 days learning data with the lowest error. As a result , the error rate of cooling energy use was 1.92%, the sum of cooling energy and lighting energy use was 1.79%, and the building control by using predicted weather data didn’t show a big difference with just control.