Heat Consumption Prediction Models Based on Machine Learning for District Heating-Applied Apartment Houses

In district hot water heating supply systems, accurate prediction of heat demand is crucial for stable operation and integration of the thermal energy storage (TES) system. Various prediction models based on machine learning have been developed in the past; however, these models, relying solely on pattern analysis and outdoor temperature, may not guarantee high prediction accuracy. In this study, we developed three categories of heat consumption prediction models for district heating in residential buildings based on various independent variables. First, we collected data on heat consumption in residential buildings and used it as the dependent variable. In terms of the independent variables, we developed three models: model A, which focused on outdoor temperature; model B, which included various climatic factors; and model C, which added holiday and season to the climatic factors. We evaluated the prediction performance of these models. The R² values of models C, B, and A were 0.8144, 0.7814, and 0.7596, respectively.