AI-Enabled Energy Management for Sustainable Smart Greenhouses: An Integrated Review

Greenhouse based Controlled Environment Agriculture enables year round crop production. Maintaining a specific microclimate requires a significant amount of energy for controlling the atmosphere under the environment. Rising electricity prices, irregular solar energy generation, and specific microclimate requirements create challenges in operations and the economic viability of the system, thus demanding smart energy optimization. This creates a technical need for advanced Demand Side Management (DSM) techniques capable of aligning load operation with forecasted demand, renewable output, and market signals. Traditional strategies such as fixed rules, time slot scheduling, or heuristic controls offer limited adaptability, lack robustness under uncertainty, and do not generalize across climate zones or crop requirements. Recent developments introduce hybrid AI driven DSM frameworks that combine forecasting, optimization, and autonomous control using approaches such as Deep Learning (DL) based predictors, Model Predictive Control schedulers, and exploration Reinforcement Learning policies to enhance renewable utilisation, reduce cost, and maintain microclimate stability. This review systematically analyzes these methodological shifts, comparing forecasting, scheduling, and control approaches in terms of predictive performance, computational overhead, and deployment feasibility. Recurring challenges including limited transferability across climates, insufficient reporting of techno economic outcomes, and weak validation against biological risk constraints are examined alongside potential solutions such as uncertainty quantification, domain adaptation, and assessment of reliability. The outlines research priorities standardised evaluation directions to support robust, scalable, and commercially relevant DSM adoption in smart greenhouse environments.