Deep Reinforcement Learning-based Energy Management Strategy for a Microgrid with Flexible Loads

Abstract

In this paper, an intelligent deep reinforcement learning -based energy management strategy is investigated for the microgrid to minimize the operation cost. The thermostatically controlled loads are considered in the microgrid to ensure the operation flexibility. The energy management problem is first formulated as a Markov decision process, and the state-of-the-art deep reinforcement learning method, namely proximal policy optimization, is applied to solve the decision-making problem. Meanwhile, the system uncertainties including wind power generations, electricity prices and electricity loads are considered. A comparison study with respect to other methods is carried out to illustrate the effectiveness of the proposed method.