Towards intelligent management of regional building energy systems: A framework combined with deep reinforcement learning for hybrid energy storage

Abstract

The adoption of renewable energy has been increasingly recognized as a viable solution to the rapid growth in building energy consumption. Integrating energy storage units into building energy systems can effectively mitigate uncertainties associated with renewable energy and enhance the balance between energy supply and demand. Compared to single energy storage systems, hybrid energy storage offers greater regulation potential and flexibility. However, the increased complexity of control strategies due to additional regulation variables presents a significant challenge. Furthermore, existing research often neglects the interactive effects between heat pumps and heat storage units. Addressing these issues, this study examines a regional energy system in Tianjin that integrates renewable energy generation, ground source heat pumps, and hybrid energy storage. The operational characteristics of ground source heat pumps are incorporated to optimize the charging and discharging processes of hybrid energy storage systems. Using a multi-agent deep reinforcement learning algorithm, the study adaptively optimizes the coordinated control of hybrid energy storage with the objectives of enhancing system operational benefits and increasing renewable energy utilization. Compared to a benchmark model, the proposed approach improves net system income by 23.64% and reduces underutilized renewable energy by 27.96%.