METHOD AND APPARATUS FOR RENEWABLE ENERGY ALLOCATION BASED ON REINFORCEMENT LEARNING

The disclosure provides a method and an apparatus for renewable-energy allocation based on reinforcement learning for an energy aggregator having an energy storage system (ESS) to coordinate renewable-energy allocation between energy suppliers and energy demanders. In the method, historical power-generation data of each supplier is collected and adapted to generate a renewable-energy index representing uncertainty of renewable energy. Multiple market indexes related to renewable-energy market dynamics are collected and integrated with the renewable-energy index and electricity information of the ESS to generate multiple states of a Markov decision process. The states are inputted to a reinforcement learning model to determine a price for the suppliers and the demanders. According to a supply amount and a demand amount proposed by the suppliers and the demanders in response to the price, the ESS coordinates the supply and demand of the renewable energy, and the reinforcement learning model is updated.