Optimal scheduling of BIES with multi-energy flow coupling based on deep RL

Building integrated energy systems （BIESs） can enhance energy efficiency ratio （EER） and reduce carbon emissions while meeting diverse user-side load demands. To further improve the energy dispatch capability of BIES， this paper proposes a low-carbon economic and optimal dispatch method for BIES with multi-energy flow coupling based on deep reinforcement learning （deep RL）. Firstly， a mathematical model of a photovoltaic-storage integrated BIES with multi-energy flow coupling is established to fully characterize energy interaction and coupling characteristics. Secondly， the state space， action space， and reward function for the operational dispatch strategy are designed using deep RL， and a low-carbon economic and optimal dispatch framework is constructed using the ​soft actor-critic （SAC） algorithm.​ Finally， the proposed method is validated in typical daily load scenarios in summer and winter. Results demonstrate that， compared to similar methods， the proposed method achieves faster convergence， more stable optimization effects， and effectively reduces both daily energy costs and carbon emission costs in IES operations.