The authors frame greenhouse operation as a Controlled Environment Agriculture (CEA) challenge involving multiple interdependent targets: air temperature and humidity, CO₂ enrichment, photoperiod-constrained lighting, and irrigation under dynamic and limited energy availability. We propose a knowledge-driven, multi-objective Model Predictive Controller whose cost function integrates expert priorities elicited via an online Analytic Hierarchy Process (AHP) survey; these AHP-derived weights parameterize the controller’s objectives and are solved over two 72 h seasonal episodes, so the MPC can anticipate renewable availability and coordinate HVAC, (de)humidification, CO₂ dosing, LED lighting, and irrigation alongside dispatch from photovoltaic and wind sources, battery storage, and the grid. By embedding the physical interdependence of climate variables directly into the decision layer, the controller schedules energy-intensive actions around renewable peaks and avoids counterproductive actuator conflicts. Seasonal case studies (summer/high solar and winter/low solar) demonstrate robust performance: temperature tracking errors of SMAPE 2.25%/3.05% and CO₂ SMAPE 3.72–3.92%; humidity control with SMAPE 7.04–8.56%; lighting and irrigation following setpoints with low NRMSE (0.08–0.14). Summer energy was 59% renewable; winter was only 13%, increasing grid reliance to 77.5% (peaks: 4.57 kW/6.92 kW for 197.7/181.5 kWh). Under water or energy scarcity, the controller degrades gracefully, protecting high-priority agronomic variables while allowing bounded relaxation on lower-priority targets. This expert-informed, predictive, and resource-aware orchestration offers a scalable route to precision greenhouse control within the food–water–energy nexus.