Towards driving-independent prediction of fuel consumption in heavy-duty trucks

Heavy-duty vehicles are among the major contributors to greenhouse gas emissions in addition to their high energy consumption. Thus, modeling their fuel consumption (FC) is of prime importance to the limitation of these environment-harmful emissions and energy saving. In this paper, we propose a data-driven model based on artificial neural networks (ANN) to predict the average FC in heavy-duty cloud-connected Ford trucks. In particular, we propose a driving-independent model based only on the vehicle weight and road grade. Owing to idling situations, the average FC includes some outliers; we propose to remove these outliers based on the weight-normalized average FC to take the changing vehicle weights into consideration. Initially, the model uses the percent torque, vehicle speed, vehicle weight, and road slope as predictors. In that case, our proposed model achieved an R2 of 0.96 outperforming the results in the literature by a significant margin. Next, we investigate the cases of excluding the torque and vehicle speed in order to assess the model's effectiveness when using only those predictors which are independent of the vehicle dynamics. In these challenging cases, our proposed model still maintains an R2 above 0.8.