Development of the Artificial Intelligence Models for Estimating the Building Heating Performance with the Implementing of the Design-Architecture Parameters

Abstract

This study investigates the application of machine learning techniques, specifically Multivariate Adaptive Regression Splines (MARS) and Model Trees (MT), in estimating building heating performance. Accurate heating performance estimation is crucial for improving energy efficiency, reducing operational costs, and achieving sustainability goals. By leveraging real-world datasets that incorporate variables such as weather conditions, building characteristics, and energy consumption patterns, the study aims to evaluate the effectiveness of these two advanced modeling approaches.

The results indicate that both MARS and MT models provide reliable and accurate predictions of building heating performance. However, the MARS model (RMSE=0.247, R=0.993) demonstrates superior performance compared to the MT approach (RMSE=9, R=0.947). The MARS model’s flexibility in capturing nonlinear relationships and interactions among variables contributes to its enhanced predictive accuracy. In contrast, the MT model, which relies on classification-based and formula-driven methods, exhibits limitations in handling complex variable interactions.

This study highlights the advantages of using MARS for heating performance estimation, emphasizing its potential as a robust and adaptable tool for energy management in buildings. The findings underscore the importance of selecting appropriate machine learning methods tailored to specific predictive tasks, ultimately advancing the state-of-the-art in energy modeling and building performance optimization.