Predicting Hourly Energy Consumption in Buildings

Abstract

Predicting energy consumption in residential, commercial, and industrial buildings based on square foot, geometry, load profile, and weather conditions is a challenging task. To effectively manage the energy demand, forecasting has become a key element for operators and buildings’ owners to monitor their energy usage. Predicting the energy demand patterns on a monthly and yearly basis helps improve buildings’ energy management. This research work contains data sets from the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) on building types such as educational, offices and residential users. Based on one-year training data, authors were able to predict the next two-year energy demand of 1500 buildings using three different forecasting models: Light-GBM, Artificial Neural Network, and Linear Regression. The preliminary findings indicate that Light GBM outperforms other models.