An AI-powered Smart Building Energy Management System (SBEMS) designed to autonomously optimize HVAC and Battery Energy Storage Systems (BESS). Instead of relying on traditional reactive thermostats, this system utilizes Deep Reinforcement Learning (PPO) and predictive Machine Learning (LightGBM) to forecast human occupancy and execute proactive "peak shaving" strategies. By intelligently pre-cooling the building and trading energy during off-peak hours, the system drastically reduces grid electricity costs while strictly maintaining occupant thermal comfort and indoor air quality. The solutio