Enhancing demand response in microgrid economic dispatch: An approach integrating deep deterministic policy gradient with custom activation function

The economic dispatch of microgrids presents significant challenges, particularly in managing demand response. This study addresses the limitations of deep reinforcement learning (DRL) in strictly adhering to demand response constraints within microgrid economic dispatch. We propose a novel custom activation function, named DRActivator. Initially, we develop an economic dispatch model for microgrids that incorporates demand response considerations. DRActivator integrates demand response constraints directly into the Actor-Critic network based on the deep deterministic policy gradient (DDPG) algorithm, enabling the agent to adapt to changes in the state space while meeting constraints. An empirical analysis is conducted on a microgrid to validate the effectiveness of the proposed approach. The experimental results demonstrate that the proposed algorithm highlights the potential of DRActivator in microgrid economic dispatch and offers a novel perspective on the application of DRL under complex constraints.