In the pursuit of low-carbon transportation, electric vehicles (EVs) are experiencing unprecedented growth, necessitating efficient charging management strategies. Traditional online real-time charging scheduling, often relying on receding horizon policies, overlooks critical challenges posed by the asynchronous nature of EV arrivals and departures. Specifically, these strategies struggle to effectively handle two distinct types of charging sessions: those already connected in real-time with known parameters, and those yet to connect but anticipated within a prediction horizon, characterized by higher uncertainty compromising scheduling efficiency. To fill these gaps, this article proposes Padding-Asynchronous-arrivals, Transition-probabilities-Captured, Heterogenous-horizons-aware (PATCH), a novel framework for EV scheduling. In the first Padding-Asynchronous-arrivals (PA) module, by detecting real-time connected sessions and padding future arrivals, PATCH prepares for optimal charging allocation. In the second Transition-probabilities-Captured (TA) module, the charging scheduling is formulated as a Markov decision process (MDP), leveraging the deterministic arrival time of real-time sessions to accurately model transition probabilities and distinguish uncertainties. The final Heterogenous-horizons-aware (H) module ensures that each session’s charging requirements are met prior to departure, while dynamically adjusting the prediction horizon based on individual session durations. The MDP-based scheduling problem within PATCH is solved using a robust-bonded dynamic programming algorithm, ensuring resilience against various uncertainties while optimizing time costs. Simulations based on real-world EV charging data demonstrate that, under the premise of maintaining a similar computing speed with traditional methods, underscoring its potential to enhance EV charging management.
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号