Energy consumption forecasting and thermal insulator selection with random forest regression

Thermal insulation is one of the most widespread and effective solutions for minimizing energy consumption in buildings. However, selecting the most suitable insulating material for specific climatic conditions remains a major challenge. Moreover, predicting this energy consumption is complex due to the diversity of parameters to be considered as well as the different existing climate zones. In this context, this work proposes a machine learning-based model to predict an energy consumption index in the case of using thermal insulators in various climate zones around the world. The objective is to identify the most optimal insulator among several options. The model used in this study is Random Forest (RF), which belongs to the family of ensemble learning models. The data used in this study come from numerical simulations carried out with Matlab and consist of 1,400 samples, derived from the analysis of 35 thermal insulators distributed across 20 climate zones. Each zone was studied by considering three typical summer days and three typical winter days. After designing this database, a preprocessing step was performed to analyze the distribution of the Φ index to be predicted. In addition, a correlation analysis was conducted to identify the most influential characteristics on energy consumption, thereby enhancing the predictive performance of the model. The results show that the RF algorithm provides remarkable performance in predicting this performance index, with an R² accuracy of 84% and a mean absolute error of 0.02. These performances were evaluated in comparison with other models commonly used in similar studies, confirming the superiority of the RF model in terms of accuracy and reliability. These results highlight the strong predictive capability of the RF model in estimating the value of the Φ index, i.e., energy consumption. The application of this model appears promising for optimizing energy consumption in buildings.