A Knowledge-Driven Smart System Based on Reinforcement Learning for Pork Supply-Demand Regulation

With the advancement of Agriculture 4.0, intelligent systems and data-driven technologies offer new opportunities for pork supply-demand balance regulation, while also confronting challenges such as production cycle fluctuations and epidemic outbreaks. This paper introduces a knowledge-driven smart system for pork supply-demand regulation, which integrates essential components including a knowledge base, a mathematical-model-based expert system, an enhanced optimization framework, and a real-time feedback mechanism. Around the core of the system, a nonlinear constrained optimization model is established, which uses adjustments to newly retained gilts as decision variables and minimizes supply-demand squared errors as its objective function, incorporating multi-dimensional factors such as pig growth dynamics, epidemic impacts, consumption trends, and international trade into its analytical framework. By harnessing dynamic decision-making capabilities of reinforcement learning (RL), we design an optimization architecture centered on the Q-learning mechanism and dual-strategy pools, which is integrated into the honey badger algorithm to form the RL-enhanced honey badger algorithm (RLEHBA). This innovation achieves an efficient balance between exploration and exploitation in model solving and improves system adaptability. Numerical experiments demonstrate RLEHBA’s superior performance over State-of-the-Art algorithms on the CEC 2017 benchmark. A case study of China’s 2026 pork regulation confirms the system’s practical value in stabilizing the supply-demand balance and optimizing resource allocation. Finally, some targeted managerial insights are proposed. This study constructs a replicable framework for intelligent livestock regulation, and it also holds transformative significance for sustainable and adaptive supply chain management in global agri-food systems.