Deep Reinforcement Learning-Based Charging Price Determination Considering the Coordinated Operation of Hydrogen Fuel Cell Electric Vehicle, Power Network and Transportation Network
Bei Li, Jiangchen Li, Mei Han
Abstract
Currently, hydrogen fuel cell electric vehicles (HFCEVs) are becoming more financially accessible as an alternative to petroleum-powered vehicles, while also decreasing carbon dioxide emissions. However, the coordinated scheduling of HFCEVs refueling with the operations of hydrogen refueling stations, electrical power network (PN), and transportation networks (TN) represents an integral challenge. This complex problem encompasses a combinatorial mixed-integer nonlinear optimization problem with a sizeable number of decision variables. Existing methods struggle to adequately solve this problem. In this article, deep reinforcement learning (DRL) is deployed to determine the refuelling price to guide the HFCEV refuelling in the transportation network. First, HFCEV traffic flow model based on the refuelling price in real-world TN is presented. Then, the HFCEVs hydrogen demands in the microgrid is presented. After that, an IEEE 30 nodes utility grid exporting electricity to microgrids is presented. At last, DRL (DDPG, TD3, SAC, PPO) is deployed to determine the price based on the traffic condition of the TN and the voltage condition of the PN. The simulation results demonstrate that through the DRL price agent, the total travel time of the TN and the total operation costs of the PN are all reduced, and multiagent DDPG and TD3 algorithm have the best performance.