Facilitating Battery Swapping Services for Freight Trucks with Spatial-Temporal Demand Prediction

TL;DR

The paper tackles the electrification of heavy-duty trucking by evaluating battery-swapping services through a predict-then-optimize framework. It combines spatial-temporal demand forecasting using Temporal Graph Convolutional Networks and Attention-based GCNs with a rolling-horizon integer program to dynamically allocate batteries between fixed and mobile BSS on a 2,500-mile highway network. Key findings show that short-horizon predictions favor T-GCN, while longer horizons benefit from A3T-GCN, and that mobile versus fixed deployment depends on traffic volatility, with mobile preferred early on and fixed preferred as patterns stabilize. The work provides actionable guidance for infrastructure planners and demonstrates how data-driven forecasting can improve the feasibility and efficiency of large-scale electrification of freight transport.

Abstract

Electrifying heavy-duty trucks offers a substantial opportunity to curtail carbon emissions, advancing toward a carbon-neutral future. However, the inherent challenges of limited battery energy and the sheer weight of heavy-duty trucks lead to reduced mileage and prolonged charging durations. Consequently, battery-swapping services emerge as an attractive solution for these trucks. This paper employs a two-fold approach to investigate the potential and enhance the efficacy of such services. Firstly, spatial-temporal demand prediction models are adopted to predict the traffic patterns for the upcoming hours. Subsequently, the prediction guides an optimization module for efficient battery allocation and deployment. Analyzing the heavy-duty truck data on a highway network spanning over 2,500 miles, our model and analysis underscore the value of prediction/machine learning in facilitating future decision-makings. In particular, we find that the initial phase of implementing battery-swapping services favors mobile battery-swapping stations, but as the system matures, fixed-location stations are preferred.

Figures (6)

• Figure 1: Lost demand under different shifting hours.
• Figure 2: Table \ref{['table:ml_perf']} depicts the performance measure of T-GCN and A3T-GCN. The column Hour stands for the $\hat{\bm{y}}_s$ such that $s \in [1, h]$. We report performance metrics such as root mean squared error (RMSE) and mean absolute error (MAE). Figure \ref{['fig_graph_acc']} shows the predicted traffic volume compared with the actual traffic volume in a service station.
• Figure 3: For \ref{['fig_varying_inventory']}, the horizontal axis stands for the ratio of the total inventory of batteries to the average demand, and the vertical axis stands for the relative lost demand compared to that of the oracle (hindsight) scheduling policy. For \ref{['fig:diff_planning_len_bar']}, the horizontal axis stands for the value of $h$. Based on different $h$, the scheduling policy outputs different unmet demand, and the vertical axis depicts the ratio of unmet demand to the total demand.
• Figure 4: Illustration of shifting. The red line represents the actual traffic of a service station whose traffic volume is not shifted. The blue lie represents a shifted one. The dotted line represents the traffic after shifting forward (4 hours and 8 hours). The traffic pattern of the road network will change after such a shift.
• Figure 5: Demand-supply visualization of 3 stations under mobile BSS to fixed BSS ratio $=0.3$. The green dotted line denotes the ML-predicted demand, the red line stands for the actual demand, and the blue lines shows the supply of the BSS controlled by the scheduling policy. From the plots, we can observe that the mobile BSS is allocated prior to the traffic surge, which demonstrates the effectiveness of the planning policy.