On December 12, 2025, the TRANS Research Group invited Dr. Xu Hao from the University of Science and Technology Beijing to deliver an academic lecture titled “Research on Integrated Transportation and Energy Systems Considering Heterogeneity in New Energy Vehicle Usage Patterns.”
To address challenges such as the heterogeneity in electric vehicle (EV) travel behavior and the uncertainty of renewable energy generation, the report proposed a “Travel Characteristic-Oriented Reinforcement Learning Charging-Discharging Control Framework.” This framework utilizes the DQN algorithm to achieve joint energy scheduling for electric vehicles, buildings, and photovoltaic systems (V2B-PV), delivering system performance that closely approaches that of offline optimization algorithms. In tests using real-world data, the proposed reinforcement learning control strategy demonstrated significant improvements over traditional control algorithms: reducing charging costs by 55%, decreasing carbon emissions by 11.6%, and increasing photovoltaic utilization to 95%. Furthermore, the system quantifies the Value of Information (VoI) of EV travel time and building load data for charging-discharging management strategies. With high computational efficiency at the millisecond scale, the strategy is well-suited for near real-time control scenarios, providing a new paradigm for intelligent transportation-building-grid synergy.