Volume 4: Innovative Solutions for Energy Transitions: Part III

Predicting Energy Consumption in Mixed-Use Buildings Using Machine Learning Techniques Aaron Jules R. Del Rosario, Aristotle T. Ubando, Alvin B. Culaba

Abstract

The rise of mixed-use buildings has contributed to the sustainable development of cities, but they still lack energy design guidelines. Energy consumption forecasting models have been crucial to the improvement of energy efficiency and sustainability of buildings, but their application to mixed-use buildings have been challenging and less tackled in literature. This study presented a novel forecasting model to predict energy consumption in mixed-use buildings using machine learning techniques. The model integrated kmeans clustering algorithm and support vector regression to improve predicting performance. The model was demonstrated to a case study considering mixed-use buildings in a tropical area. Clustering results found major differences in the consumption behavior of building clusters, especially on peaking characteristics. The proposed forecasting model was able to capture these variations due to clustering, leading to an increase in predicting performance. The model also performed within building modeling standards and better than statistical approaches in the literature.

Keywords Energy consumption forecasting, mixed-use buildings, machine learning, energy conservation in buildings, energy modeling, urban energy systems