Machine Learning Assisted Optimization of Climate-Resilient Smart Greenhouse Microgrids

Modern greenhouse horticulture under the Agriculture 4.0 paradigm relies heavily on smart, automated, and energy-intensive climate control, making a reliable and sustainable power supply a core operational requirement. This study presents a data-driven, machine learning (ML)-assisted framework to optimize a smart hybrid microgrid—combining photovoltaic (PV), wind, battery energy storage, and limited backup generation with demand-side load flexibility—for a 5-hectare greenhouse cooperative in Çanakkale, Türkiye. To capture meteorological variability and smart farming load dynamics, five statistically independent hourly realizations (43,800 operating hours) were generated from published climate statistics and coupled with a weather-responsive electricity demand model. Using a 1505-point experimental design (comprising 1500 Latin hypercube samples augmented with 5 benchmark configurations) and chronological dispatch simulations, three surrogate algorithms (random forest, extremely randomized trees, and histogram gradient boosting) were evaluated to predict key socio-economic and environmental metrics: levelized cost of electricity (LCOE), life-cycle greenhouse gas (GHG) intensity, loss-of-power-supply probability (LPSP), and renewable fraction. Multi-objective Pareto screening and subsequent full chronological verification identified an optimal compromise configuration comprising 1040 kW PV, 50 kW wind, 2000 kWh/540 kW battery storage, 105 kW diesel reserve, and 19.9% flexible load capacity. The optimized smart microgrid achieved an LCOE of 0.208 USD/kWh, a GHG intensity of 0.074 kg CO2e/kWh, a worst-realization LPSP of 0.042%, and a mean renewable fraction of 99.3%. Compared to traditional fossil-fuel energy supply, the proposed system reduced energy costs by 65.9%, carbon footprint by 94.7%, and diesel reliance by 99.4%. The findings demonstrate that integrating ML surrogates with smart demand scheduling offers a scalable and transferable digital pathway toward energy-autonomous, decarbonized, and resilient Agriculture 4.0 greenhouse operations.