An Optimal Power Bidding Strategy for Electric Vehicles Via PER-Based Reinforcement Learning

As the number of electric vehicles (EVs) on the road increases, the difficulty of locating charging stations has become a prominent issue. The increase in random EV charging applications poses a significant challenge to the main power grid's power distribution capacity, particularly during peak periods. Given that the optimized second price auction (SPA) mechanism exhibits benefits in resolving resource allocation problems, it is possible to apply optimized SPA to resolve the problem of power allocation for charging stations. This study aims to investigate new methods for optimizing the power bidding strategy of EVs in the trading market using the SPA mechanism. To this end, a reinforcement learning (RL) method based on prioritized experience replay (PER) is intended to assist EVs in formulating power bidding strategies under the optimized SPA mechanism. Experimental results demonstrate that the PER-based RL method outperforms both the deep Q-network (DQN) method and the random method in terms of increasing EV benefit and decreasing EV owners' waiting time.