Abstract

<p>To achieve green and low-carbon goals in the building energy sector, precise energy management strategies are essential to support user comfort and energy-saving needs during operation. However, the thermal comfort requirements of building users may conflict with societal demands for low-carbon and energy efficiency. This creates a challenge between the precision of energy use models and the speed of energy management strategies. It is necessary to combine the advantages of building physical models and deep reinforcement learning to develop faster and more accurate energy management strategies. This paper proposes a smart energy management optimization for buildings, considering indoor heat transfer. First, a third-order heat transfer model for rooms is constructed to quantify the heat transfer between them. Next, a detailed model of the central air conditioning system is developed, considering the relationships between its internal components. To achieve green and low-carbon building operations while maintaining user comfort, a multi-objective optimization algorithm based on deep policy gradient decision-making is proposed. The method is validated on actual building energy systems using real data with 15-minute resolution. We find significant differences in heat transfer between different rooms within a building, and the proposed intelligent energy management optimization method effectively balances low-carbon, energy-efficient operation with user comfort.</p>