Power Management in Smart Buildings Using Reinforcement Learning

Abstract

This paper proposes a novel framework of power management system (PMS) using reinforcement learning (RL) in presence of a thermal energy storage system (TESS) and a battery energy storage system (BESS) to achieve peak load shaving and compensate for BESS limitations. This paper focuses on using RL approach to define the optimal charging/discharging schedule of TESS and BESS in PMS. The optimization problem is formulated as Markov decision process (MDP) and then solved by Q-learning algorithm. The efficacy of the proposed PMS framework is demonstrated by using power consumption data of a campus building. Moreover, the optimal solution obtained by RL is validated and compared with metaheuristic optimization approaches such as particle swarm optimization (PSO). Results show the effectiveness of the proposed PMS using RL to define the optimal operation schedule of energy storage systems (ESSs) while reducing 42.2% of the required BESS capacity.