Adaptive Multi-Strategy Grey Wolf Optimizer with Reinforcement Learning for Multi-Objective Precision Irrigation Optimization

Precision irrigation is crucial for sustainable agriculture, yet conventional single-objective optimization methods struggle to balance conflicting demands such as crop yield, operational cost, and environmental sustainability. This study introduces an Adaptive Multi-Strategy Grey Wolf Optimizer with Reinforcement Learning for Multi-Objective Optimization (AMSGWO-RL-MO) to enhance precision irrigation decision-making. AMSGWO-RL-MO integrates four strategies: standard GWO exploitation, Lévy flight exploration, differential evolution-based diversity enhancement, and Stochastic Elite Opposition-Based Learning. A Q-learning mechanism dynamically adjusts these strategies, adapting to real-time search conditions to select the optimal approach. We constructed a comprehensive three-objective framework incorporating soil moisture dynamics, crop growth models, and environmental impact assessments. Experimental simulations over a 40-day growth cycle demonstrate AMSGWO-RL-MO’s rapid convergence by the sixth generation, consistently achieving a high-quality Pareto front across 30 independent runs. The knee-point solution yielded a mean crop yield of 96.96%, outperforming standard GWO and multi-strategy variants by approximately 3.8%. Statistical analysis confirms its superior robustness and well-distributed solutions along the Pareto front. These results indicate that the RL-driven adaptive mechanism effectively balances exploration and exploitation. The proposed method offers a more diverse array of Pareto-optimal solutions, presenting a broader trade-off space for balancing crop yield and environmental sustainability compared to traditional weighted-sum approaches. This enhancement facilitates scientific agricultural decision-making under various operational constraints.