Supply chain optimization is essential for enhancing efficiency, reducing costs, and improving customer satisfaction. This paper explores the application of reinforcement learning (RL) in supply chain management, particularly in demand forecasting and logistics planning. We discuss RL frameworks, methodologies, and advantages over traditional methods. Empirical studies demonstrate how RL-based models dynamically adapt to uncertainties, improving demand prediction accuracy and logistics efficiency. The paper also outlines challenges and future research directions. Furthermore, we present real-world case studies demonstrating the successful deployment of RL in various industries and discuss future advancements in AI-driven supply chain systems.