Road grade and truck weight matter: Investigating link-level energy consumption uncertainty

Modeling heavy-duty truck energy consumption and emissions under real-world conditions is challenged by uncertainties from dynamic road grade and truck weight. This study integrates second-by-second operational, fuel consumption, and NOx emissions data with accurate road grade and weight information across over 12,000 100-meter road links. The data were collected from 48 in-use trucks operating along a 60-kilometer highway with varying topography. This study quantifies link-level modeling uncertainty and analyzes the influence of road grade and truck weight. Multiple regression and machine learning models are applied to evaluate energy consumption prediction performance under different feature combinations. Results show that using only average speed for 5-km segments results in a 25% error, which decreases to 11% when road grade, truck weight, and acceleration are incorporated. These findings underscore the need to integrate road grade and truck weight into predictive models to improve energy and emissions analysis.