An Ensemble Machine Learning Model for Enhancing the Prediction Accuracy of Energy Consumption in Buildings

Predicting building energy use is necessary for energy planning, management, and conservation. It is difficult to achieve accurate prediction results due to the inherent complexity of building thermal characteristics and occupant behavior. Machine learning has been recently applied for predicting energy consumption. Improving its predictive accuracy and generalization ability is essential. Therefore, this study proposed a machine learning model for an ensemble approach to forecasting energy consumption in non-residential buildings. Various datasets from non-residential buildings were collected to assess the predictive performance. Artificial neural networks, support vector regression, and M5Rules models were used as baseline models in this study. Evaluation results have confirmed the effectiveness of the ensemble machine learning model in the next 24-h energy consumption prediction in buildings. The mean absolute error (MAE) and mean absolute percentage error (MAPE) obtained by the ensemble machine learning model were 2.858 kWh and 16.141 kWh, respectively. The ensemble machine learning model can improve the MAE by 123.4% and the MAPE by 209.3% as compared to baseline models. This study contributes to highlighting the advantages of machine learning applications for the building sector. Ensemble machine learning models can be proposed as an effective method for forecasting energy consumption in buildings.