Traffic-informed re-ranking of candidate routes for series range-extended electric vehicles based on predicted road-segment equivalent fuel consumption

Reducing road-transport energy use and emissions requires route guidance to consider vehicle energy demand rather than distance or time alone. For series range-extended electric vehicles (REEVs), fuel use, battery-energy change and boundary kinetic-energy change must be expressed on a common basis, and accumulated segment-prediction errors must not destabilise route selection. Existing eco-routing approaches do not fully connect these requirements at the planning stage. This study integrates an additive equivalent-fuel metric, traffic-informed segment prediction and candidate-route re-ranking. Microscopic traffic simulation and a quasi-static REEV model generate segment samples. A seven-model benchmark supports the use of a compact feedforward neural network (FNN), which predicts segment equivalent fuel consumption from length, mean speed and mean vehicle density; repeated A* search generates candidate routes. Test-set root mean square error (RMSE), mean absolute error (MAE) and R ² were 7.37 g, 4.14 g and 0.8544, respectively. Across 18 routes in 3 origin–destination (OD) cases, route-level mean absolute percentage error (MAPE) was 3.60%–4.34%, Spearman correlation was 0.943–1.000 and Kendall correlation was 0.867–1.000. Prediction and simulation selected the same minimum-equivalent-fuel route in all cases, yielding zero selection regret. Relative to minimum-travel-time navigation baselines, the selected routes reduced simulated equivalent fuel consumption by 3.99%–9.60%; actual fuel use in OD1 decreased by 4.10% under the thermostat strategy. Distance–intensity decomposition identified a distance-dominated benefit in OD1, an energy-intensity-dominated benefit in OD2 and combined reductions in OD3. In OD3, candidate distances differed by only 0.09%, whereas simulated equivalent fuel varied by 3.51%, showing that distance alone did not determine route energy. For OD1, fuel savings remained 4.54%–4.82% under three state-of-charge (SOC) trajectory-planning schemes. Instrumented pre-trip processing averaged 15.62 s, and prescribed-route model-in-the-loop (MIL) and hardware-in-the-loop (HIL) responses agreed. The results demonstrate a planning-stage framework for series-REEV route selection using established prediction and search methods.