A Modified Incentive-based Demand Response Model using Deep Reinforcement Learning

In recent years, significant challenges emerge in power grid, such as peak load reducing and renewable incorporating, which can be relieved through demand response program. This paper proposes a demand reduction modified function to better describe the fluctuation of customers’ energy demand, aiming to help the service provider designing an optimal incentive rate more close to the real situation of demand response. In particular, to overcome the complexity of this high dimensional and continuous decision making problem, an algorithm is designed based on reinforcement learning and deep neural network. Simulation results show that this proposed algorithm for the modified incentive-based demand response model encourages the participation of customers, and also enhances the overall profits of both service provider and customers to achieve a win-win result.