Dynamic pricing of demand response based on elasticity transfer and reinforcement learning

In this paper, we study the dynamic pricing of electricity service provider in the day-ahead spot market. Since the wholesale electricity price that the electricity service provider obtains from the utility grid is constantly changing, and the user's response behavior is unknown, setting a suitable retail price is a big challenge for the electricity service provider. In response to this problem, we propose a method based on elasticity transfer and reinforcement learning, which transfers the elasticity of the implemented demand response region to the region where the user elasticity is unknown, as the initial reference for dynamic pricing, and then uses the SARAS learning algorithm for practical exploration and learning. The simulation results show that the elasticity transfer to the new region as the initial reference can significantly improve the learning rate compared to the system without the initial reference. Therefore, the proposed method can maximize the price of the electricity service provider and the user by setting the optimal price at a faster rate without prior knowledge of the user.