Machine learning and metaheuristic optimization for chiller energy prediction in commercial buildings research trends and future directions

Chiller systems account for a large share of energy use in commercial building HVAC, and accurate prediction supports energy management, demand response, and decarbonization. Although machine learning-based chiller energy prediction has received increasing attention, existing studies have primarily focused on individual prediction models, while the intellectual structure and evolution of research combining machine learning, metaheuristic optimization, and feature selection remain insufficiently mapped. This study conducts a bibliometric analysis of 419 journal articles indexed in the Web of Science Core Collection from 2016 to 2026 using VOSviewer version 1.6.20. The dataset records 20,097 citations, an average of 47.96 citations per paper, and an h-index of 63, with annual output rising from 9 papers in 2016–75 in 2025. Bibliographic coupling of 60 highly cited papers identifies four clusters: ANN-based model predictive control for HVAC, deep and transfer learning for load forecasting, benchmarking of prediction models, and hybrid metaheuristic ANN approaches. Keyword co-occurrence across 56 terms reveals themes in HVAC optimization, machine learning-based load prediction, building energy modeling, and simulation. China leads with 166 publications, representing 39.6 % of total output. The findings establish a comprehensive research map of machine learning and metaheuristic optimization for chiller energy prediction and identify future priorities including chiller-specific feature selection, model interpretability, real-time deployment, climatic generalization, and integration of metaheuristic optimization with deep learning architectures.