Weather-aware energy prediction for battery electric trucks from fleet telematics and road weather data

Authors:
Preprint
Publisher website
Google Scholar
Abstract
In regional freight operation, route, payload, and charging plans for battery electric trucks depend on trip energy. In Nordic winters this demand varies with temperature and road-surface state, which vehicle energy models omit by treating auxiliary demand as constant and rolling resistance as independent of weather. This study predicts the trip-level energy consumption of a battery electric truck in commercial service in Finland and Sweden from 18 months of telematics, matched along the route to road-weather analyses of temperature, water film, and surface snow. A calibrated longitudinal-dynamics model is combined with a gradient-boosting correction and evaluated on held-out trips, months, and an unseen winter. The physics residual rises from slightly below zero above 15◦C to 50 kWh/100km below -15◦C, each 1◦C drop adding 1.17 kWh/100km, and neither a constant-power auxiliary load nor a load-proportional rolling loss reproduces it alone. A 0.15 mm water film adds 8.4 kWh/100km and snow cover 6.8 kWh/100km plus 7.8 kWh/100km per centimeter, independently of temperature, so consumption peaks near 0◦C. On unseen trips thecorrection raises the R2 of total consumption from 0.40 to 0.87 and recovers 53% of the residual variance, 41% for an unseen winter, twice what a temperature-dependent auxiliary load achieves. A wet, snowy day at -20◦C shortens the range of a fully loaded truck from about 260 km to 185–195 km. A diesel truck analyzed as a reference shows that fleet fuel-level signals are too coarse for such prediction.

This publication has 15 references indexed in Scilit: