Evaluating and Forecasting the Aggregate Flexibility of Large-Scale Electric Vehicle Charging and Swapping Facilities

Evaluating and forecasting the flexibility in the large-scale electric vehicle (EV) charging and battery swapping systems is crucial for optimizing day-ahead bidding and operational control. This paper proposes a novel framework that integrates evaluation and forecast of the aggregate flexibility, addressing the challenges of computational infeasibility and privacy concerns in tackling individual EVs or batteries. In the evaluation stage, charging and discharging flexibility is estimated using historical backtesting. A modeling method based on order chains is developed to estimate the flexibility of battery swapping facilities. Considering EV user reservation, the feasible region constrained by the energy and power boundaries is modified for both V1G and V2G scenarios. Explicit expressions are derived to improve computational efficiency and reduce flexibility loss. In the forecast stage, a de-stationary Informer-based algorithm is designed to predict the flexibility parameter time-series, which can effectively capture the long-sequence dependencies using an enhanced attention mechanism. Simulations on real-world charging and swapping record datasets demonstrates that the proposed method achieves a superior balance between computational efficiency and evaluation accuracy, while attaining lower forecasting errors compared to baseline models, including Informer, MLP, and LSTM.