文献类型：期刊文献

英文题名：A Comparative Approach on Predicting Energy Consumption in HVAC Systems Using Boosting Algorithms

作者：Wang, Yetong[1]; Liu, Xiaoyu[2]

第一作者：Wang, Yetong

机构：[1] Hainan Engineering Research Center for Virtual Reality Technology and Systems, Hainan Vocational University of Science and Technology, Hainan, Haikou, 571126, China; [2] College of Applied Science and Technology, Beijing Union University, Beijing, 100012, China

第一机构：Hainan Engineering Research Center for Virtual Reality Technology and Systems, Hainan Vocational University of Science and Technology, Hainan, Haikou, 571126, China

年份：2025

卷号：44

期号：3

起止页码：617-642

外文期刊名：Strategic Planning for Energy and the Environment

收录：EI(收录号：20261920686855)

语种：英文

外文关键词：Adaptive boosting - Architectural design - Artificial intelligence - Behavioral research - Benchmarking - Decision making - Decision support systems - Energy efficiency - Energy management - Energy utilization - Forecasting - Intelligent buildings - Learning systems - Optimization

摘要：Boosting the precision of energy usage prediction in Heating, Ventilation, and Air Conditioning (HVAC) systems is essential in elevating energy efficiency and lowering operating expenses in contemporary building environments. This study introduces a suite of hybrid predictive schemes that integrate advanced machine learning (ML) tactics – Extreme Gradient Boosting (XGBoost) and Categorical Boosting (CatBoost) – with state-of-the-art metaheuristic optimization algorithms, namely the Salp Swarm Algorithm (SSA), Grasshopper Optimization Algorithm (GOA), and Forensic-Based Investigation Optimization (FBIO). Leveraging a high-resolution database obtained from a commercial building in Turin, Italy, six blended schemes were developed and rigorously examined utilizing a comprehensive set of statistical performance indicators, including R2, RMSE, MAPE, MBE, RAE, and VAF, along with convergence rate and runtime analysis. Among the proposed schemes, the XGBoost-SSA hybrid consistently outperformed its counterparts, achieving superior predictive accuracy (R2 = 0.91), the highest variance accounted for (VAF = 91.35%), and reduced computational time (6.20 s). The outcomes demonstrate that incorporating SSA into XGBoost significantly enhances model robustness, convergence behavior, and generalization capability. These findings underscore the potential of the recommended hybrid modeling framework as a robust decision-support tool for intelligent energy management in building systems and offer promising directions for future research in energy analytics and smart infrastructure design. ? 2025 River Publishers