Buildings account for roughly 37% of energy-related CO
2 emissions, and space cooling already consumes nearly 10% of global electricity. Cooling demand is rising fastest in tropical cities, where air-conditioning could reach 45% of peak load, especially in India by 2050. This review
[...] Read more.
Buildings account for roughly 37% of energy-related CO
2 emissions, and space cooling already consumes nearly 10% of global electricity. Cooling demand is rising fastest in tropical cities, where air-conditioning could reach 45% of peak load, especially in India by 2050. This review critically examines thermal energy storage (TES) as a flexibility resource across three distinct scales: individual buildings, district heating and cooling networks, and city-level multi-energy systems. Using a Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)-based search of Scopus, Web of Science, and IEEE Xplore with primary and supplementary strings, 4447 records were identified, of which 174 were included. Each quantitative study was classified by validation level (simulation, laboratory, pilot, or operational) and by the centrality of thermal storage. Sensible, latent, and thermochemical storage technologies are compared using energy density (10–500 kWh/m
3), efficiency (40–95%), cycle stability, and technology readiness. The review then evaluates the role of artificial intelligence (AI), machine learning, and Internet of Things platforms in forecasting, predictive control, and operational optimization of TES networks. Thirteen method families, grouped into AI and machine learning methods, optimization methods, control methods, and digital enabling technologies, are assessed against six explicitly defined criteria with evidence-coded scores. Among 47 quantitative studies, 37 (78.7%) are simulation-only, and only four (8.5%) report operational data. Direct TES-AI studies report simulated energy savings of 8–64% and peak load reductions of about 35%, whereas field-validated intelligent control reports 17% energy savings in a single real building experiment. The review also identifies inherent drawbacks of artificial intelligence-based operations, including limited interpretability, high data and computational demands, concept drift, and cyber vulnerabilities that increased peak electric load by 17.4% in a simulated attack. A structural imbalance in the literature is evident: most validated deployments remain at the building-scale, whereas urban-scale evidence is confined to district cooling, aquifer and pit storage, and multi-energy hub studies; no study reports the coordinated operation of distributed TES assets across multiple districts. A conceptual framework and a staged roadmap linking building, district, and urban scales are proposed. Priority research needs include urban-scale pilots in tropical climates, techno-economic assessment, interpretable and drift-robust AI, and interoperability standards that support United Nations’ Sustainable Development Goals 7, 11, and 13.
Full article