Research on Advanced Control Strategies for Building Energy Management Systems to Optimize Energy Consumption in Commercial Buildings by Using DRL-PC Algorithm

This study investigates the application of Deep Rein-forcement Learning combined with Predictive Control (DRL-PC) algorithms in Building Energy Management Systems (BEMS) for commercial buildings in Dhaka, Bangladesh. The research compares the performance of a DRL-PC algorithm against a traditional rule-based control system in a 10-story office building over a 12-month period. The DRL-PC algorithm demonstrated significant improvements, achieving a 21 % reduction in annual energy consumption, a 25 % increase in occupant comfort sat-isfaction, and a 28% reduction in peak energy demand. The system showed remarkable adaptability to Dhaka's challenging climate, maintaining optimal indoor conditions 92 % of the time. With a payback period of 2.7 years, the proposed system offers a promising solution for enhancing energy efficiency and occupant comfort in commercial buildings, particularly in regions with complex climatic conditions.