A predictive and adaptive control strategy to optimize the management of integrated energy systems in buildings

The management of integrated energy systems in buildings is a challenging task that classical control approaches usually fail to address. The present paper analyzes the effect of the implementation of a reinforcement learning-based control strategy in an office building characterized by integrated energy systems with on-site electricity generation and storage technologies. The objective of the proposed controller is to minimize the operational cost to meet the cooling demand exploiting thermal energy storage and battery system considering a time-of-use electricity price schedule and local PV production. Two control solutions, a Soft-Actor-Critic agent coupled with a rule-based controller, and a fully rule-based control strategy, used as a baseline, are tested and compared considering various configurations of battery energy storage system capacities, and thermal energy storage sizes. Results show that the proposed control strategy leads to a reduction of operational energy costs respect to the fully rule-based control ranging from 39.5% and 84.3% among different configurations. Moreover the advanced control strategy improves the on-site PV utilization leading to an average increasing of self-sufficiency and self-consumption of 40% among different scenarios. The baseline control strategy results more sensitive to the size of storage whereas the proposed control achieves high savings also when smaller capacities of battery energy storage systems and sizes of thermal energy storage are implemented. The outcomes of the work prove the impact of implementation of advanced control as a way to optimize energy costs with a comprehensive view of the whole integrated energy system considering both thermal and electrical energy storage operation.