3rd International World Energy Conference, Kayseri, Turkey, 4 - 05 December 2023, pp.595-605
Research conducted on building energy consumption forecasting contributes significantly to reducing energy costs, minimizing environmental impact, enhancing occupant comfort, and optimizing building performance. In this context, this study aims to enhance energy consumption forecasting in university campus buildings by integrating time series and physics/engineering-based datasets to develop a predictive model. Methodology: The research integrates machine learning techniques by utilizing time series energy consumption data obtained from existing buildings alongside building physics/engineering data. Time series data for heating/cooling and lighting are combined with physics/engineering data, such as external environmental factors specific to the building, including outside air temperature, relative humidity, building floor area, floor height, and material type. Results: The study discusses the success of energy consumption forecasting through a comparative analysis of time series and hybrid models. In contrast to models typically preferred in existing literature that focus on either time series or physical information, this research explores the relationship between the accuracy of predictions and the amalgamation of both data types. Conclusions: This study represents a significant advancement in the field of energy consumption forecasting. The findings offer valuable guidance to energy managers and building owners in developing more effective strategies to increase energy efficiency and achieve sustainability goals. This methodology holds broader applicability, particularly in the context of complex structures like university campuses, within the realm of energy management.