The role of building energy forecasting in campus sustainability: Tackling climate change with machine learning

Global warming is expected to increase by 1.5 °C between 2030 and 2052. This phenomenon may lead to an escalation in energy consumption within buildings. Given the evolving environmental landscape, it is imperative for university campuses to proactively strategize on mitigating risks through advanced building energy forecasting methodologies. Although numerous scholars have developed building energy models (BEMs), there remains a scarcity of focus on clarifying the importance of BEM, particularly in climate change and its repercussions. Furthermore, despite various review articles on BEMs, a comprehensive guideline outlining the critical factors for clarifying building energy consumption has yet to be established. This study has initiated the development of building energy forecasting techniques utilizing statistical approaches, such as multivariate regression models, multiple linear regression (MLR) models, and analyses of relative significance. The resulting data encompasses electricity (ELC) and steam (STM) usage. The independent variables employed as inputs include building characteristics, time-related factors, and weather conditions. The findings underscored the necessity of categorizing campus buildings by type, identifying equipment power density as the foremost influence on ELC consumption. At the same time, the degree of heating emerged as the key factor affecting STM usage. The laboratory building type exhibited the highest steam consumption, warranting careful oversight. The predictive models furnish valuable insights into the building attributes that are vital and pertinent to campus building policies and execution strategies. The enhancement of STM aspires to elevate awareness regarding the seriousness of climate change through prospective weather simulations.