Demand Forecasting for Battery Swapping at a Heavy-Duty Truck Battery Swap Station in Xinjiang

With the rapid development of the new energy vehicle industry, the electrification transformation of heavy-duty trucks, as an important logistics transportation tool, has become a significant measure for optimizing China's energy structure and promoting green logistics development. This paper conducts field research and compiles three sets of data: the historical actual battery swap counts from October 2023 to February 2024 (on working days) within the service area of a battery swap station in Xinjiang, the total energy consumption during various time periods (including energy consumption from vehicle operation, air conditioning, and other sources), and the average temperature. Based on the cumulative graph of battery swap counts during different time periods, it is observed that the swap counts are primarily concentrated between 1 PM and 3 AM, with fewer occurrences in the morning. An analysis of the causes of this phenomenon is provided. A linear regression analysis is performed on the battery swap counts and the daily average temperature, demonstrating the necessity of using temperature as a predictive indicator for battery swap demand. Furthermore, utilizing the historical actual battery swap counts, total energy consumption during various time periods, and average temperature data, a Long Short-Term Memory (LSTM) model is employed to predict battery swap demand. The model's performance is evaluated by calculating the Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE) between the predicted values and the actual values, indicating that the model performs well. Additionally, the battery swap demand for the first ten working days of March 2024 is forecasted. Considering the seasonal fluctuations in battery swap demand for heavy-duty trucks and the potential market changes for future battery swap heavy-duty trucks, the maximum values from the predicted results of battery swap demand during various time periods are taken as the final demand forecast. This paper proposes a new approach to predicting battery swap demand by considering historical actual battery swap counts, total energy consumption during various time periods (including energy consumption from vehicle operation, air conditioning, and other sources), and average temperature using the LSTM model, which can serve as a reference for other studies on battery swap demand forecasting.