<p>This paper introduces a novel decentralized peer-to-peer (P2P) energy trading model leveraging a Proximal Policy Optimization (PPO) driven deep reinforcement learning (DRL) approach, to optimize energy transactions among smart homes within a smart grid environment. The proposed model aims to minimize energy costs while promoting efficient energy consumption patterns through dynamic pricing schemes. A new policy function has been designed to enhance the training and real-time working efficiency of the PPO-based P2P energy trading framework, enabling faster convergence and improved trading strategies. By learning optimal energy trading policies through continuous interaction with the environment, the model integrates historical consumption data and real-time market dynamics to deliver substantial cost savings. Experimental evaluations reveal an average reduction of 45% in energy expenses for participating households compared to conventional methods. Additionally, the proposed framework demonstrates robustness and adaptability across diverse market conditions and consumer preferences, ensuring scalability and practical applicability in real-world scenarios. This study underscores the transformative potential of reinforcement learning in advancing decentralized energy trading systems, offering a sustainable and cost-effective solution for modern energy markets.</p>