Impact Factor:7.7

DOI number:10.1016/j.trd.2026.105236

Teaching and Research Group:Q. Zhou, M. Huang, H. Tang, J. Cheng, J. Wu

Journal:Transportation Research Part D: Transport and Environment

Key Words:Battery swapping stations; Intelligent energy scheduling; Electric vehicles; Deep reinforcement learning; Hybrid hierarchical control framework

Abstract:As urban transportation electrification advances, range anxiety has become a key concern for electric vehicles (EVs). Battery swapping stations (BSS) offer a fast and efficient solution, yet traditional scheduling methods struggle to balance profitability with operational risk in dynamic environments. To address this, we propose a data-driven optimization method. First, historical data is used to derive optimal operational decisions, and an EV battery swapping demand forecasting model is built. Based on the forecasted demand and historical strategies, a long short-term memory network predicts the BSS's overall charging and discharging power. A double deep Q-network is then employed to allocate this power to individual batteries, ensuring timely swaps. Validation using real operational data from Chengdu, China, shows the proposed method effectively meets battery swapping demand, enhances scheduling efficiency and station profitability, and reduces peak loads and power fluctuations, demonstrating the potential for practical application in managing smart EV infrastructure.

Indexed by:Journal paper

Discipline:Engineering

Document Type:J

Volume:153

Page Number:105236

Translation or Not:no

Date of Publication:2026-01-24

Included Journals:SCI

Links to published journals:https://www.sciencedirect.com/science/article/abs/pii/S1361920926000295