Convex Combination Based Reinforcement Learning Algorithm for Dynamic Pricing Demand Response

Demand response (DR) is an effective strategy to correct supply-demand mismatches and achieve energy savings by adjusting the supply side flexible loads promptly. This study balances the profit of service provider (SP) and the cost of expenditure of customers (CUs) after participating in a DR strategy, and formulates the sequential decision problem of dynamic pricing in a hierarchical electricity market (HEM) as a Markov Decision Process. A convex combination based reinforcement learning algorithm is proposed to effectively overcome the overestimation problem and pessimistic underexploration of traditional algorithms and achieve an economical dynamic pricing DR strategy for the HEM.