Economic dispatch based on reinforcement learning using proximal policy optimization

Economic dispatch is facing challenges of uncertainties due to increasing penetration of renewable energies. This paper proposes a data-driven, model-free approach for economic dispatch based on deep reinforcement learning (DRL). The objective is to provide cost-efficient dispatch schedule under security constraint with consideration of uncertainties in future operation conditions. Firstly, the economic dispatch is formulated into a sequential decision-making model by defining the action, state and reward. Secondly, proximal policy Gradient algorithm is utilized for the agent to learn policy through interaction between agent and power system. Finally, case studies are performed to validate the effectiveness and advantages of the proposed model.