Speed-stratified, energy-balanced VSP models for high-resolution fuel use

This is the authors' abstract. We don't add key points for this paper.

High-resolution estimation of vehicle fuel consumption is critical for urban energy planning and traffic management, yet traditional vehicle-specific power (VSP) models suffer from coarse binning and weak speed–power coupling. This study develops two improved approaches: a speed-stratified equal-interval model (SSEIM), which applies exponential fitting within 5 km/h speed layers, and an enhanced Motor Vehicle Emission Simulator (MOVES) configuration (EMOVES), which introduces dynamic bin merging based on fuel contribution and inter-bin fuel-rate differences. Using 489 million one-second records from 764 light-duty vehicles, both models were calibrated and validated across speed ranges and roadway types. Results show that SSEIM and EMOVES reduce mean absolute percentage error of the fuel consumption factor from 9.34 % and 9.32 % (traditional equal-interval and default MOVES, respectively) to 2.63 % and 3.14 %. SSEIM achieves highest accuracy at medium–high speeds (e.g., 1.4 % error at 80–100 km/h), while EMOVES performs best in low-speed and complex conditions (e.g., 4.4 % error on local roads versus 11.3 % with default MOVES). These findings confirm that explicit speed stratification and energy-balanced binning recover physical continuity and significantly improve robustness across diverse operating regimes.