A review of artificial intelligence data-driven methods in building energy management: A comprehensive application-based of recent advances

Artificial intelligence (AI) has become a fundamental technology for improving building energy management through data-driven prediction, intelligent optimization, and adaptive control. Although numerous reviews have examined individual AI techniques, a comprehensive application-oriented synthesis of recent advances remains limited. This study presents a systematic review of AI-driven building energy management research published between 2022 and 2024. Following a PRISMA-based screening procedure, approximately 150 peer-reviewed studies were selected and analyzed using a unified classification framework based on AI methodologies, application domains, building types, optimization objectives, datasets, and evaluation metrics. The quantitative analysis shows that residential buildings accounted for approximately 45% of the reviewed studies, reflecting the increasing availability of residential energy datasets and smart home applications, whereas research on office buildings exhibited a declining trend during the selected period. Machine learning and deep learning remained the dominant techniques for energy prediction and forecasting, while reinforcement learning and metaheuristic optimization algorithms were increasingly adopted for intelligent control and multi-objective optimization. Among optimization methods, NSGA-based algorithms appeared slightly more frequently than PSO-based approaches, highlighting the growing emphasis on balancing energy efficiency, thermal comfort, operational cost, and environmental performance. The review further demonstrates a clear transition from isolated prediction models toward integrated AI frameworks that combine forecasting, optimization, and real-time control. Despite these advances, several challenges continue to hinder practical implementation, including limited benchmark datasets, insufficient model transferability across buildings and climates, high computational complexity, limited interoperability with existing building management systems, and the scarcity of large-scale real-world validation studies. By integrating quantitative trend analysis, an application-based classification, a comparative evaluation of AI methodologies, and a critical assessment of current research gaps, this review provides practical guidance for selecting AI techniques and identifies promising directions for the next generation of intelligent, explainable, and scalable building energy management systems.