The Impact of Meteorological Data on Machine Learning Model Development for Thermal Energy Consumption Forecasting in District Heating Systems

Buildings are among the largest energy consumers in the world. Predicting energy consumption in buildings has a significant impact on energy control, design optimization, renovation assessment, as well as energy pricing guidelines, thereby laying the foundation for the development of energy management strategies to achieve significant energy savings. The key factors that largely influence energy consumption in buildings include region-specific climatic conditions and individual user expectations regarding indoor temperature, humidity, and air quality. This paper proposes the random forest machine learning method for predicting thermal load as a key element in developing data-driven energy management strategies. The research was applied to a historical dataset obtained from the SCADA system of the heating plant at the Faculty of Mechanical Engineering in Nis, covering the heating season of 2022/23 and spanning the period from October 15, 2022, to December 1, 2022, along with corresponding meteorological data. The primary goal was to compare thermal load predictions obtained using the same machine learning method in both cases, with a difference in one input variable – in the first case, the actual outdoor air temperature was used, while in the second case, the outdoor air temperature adjusted for wind speed, air humidity, and solar radiation (feels like temperature) was used. The proposed method was evaluated based on four performance metrics.