Volume 10: Sustainable Energy Solutions for Changing the World: Part II

FuelNet: A precise fuel consumption prediction model using long short-term memory deep network for eco-driving Guanqun Wang, Licheng Zhang, Zhigang Xu, Syeda Mahwish Hina, Pengpeng Sun, Haigen Min, Tao Wei, Xiaobo Qu



It has been well recognized that driving behaviors significantly impact fuel consumption of vehicles. In this paper, we propose a FuelNet model based on Long Short-term Memory Neural Network (LSTM NN), which can predict vehicle fuel consumption in a very accurate manner. First, we take the kinetic vehicle parameters and the corresponding fuel consumption parameters to build the FuelNet model, and analyze the correlations between the prediction accuracy and different combinations of input parameters. In addition, our model exhibits the superior capability for fuel consumption prediction (FCP) at different speed, and the comparison with different deep learning models as well as other physics model and data-driven methods suggests that FuelNet can achieve the best prediction performance in terms of both accuracy and stability. Finally, the application of FCP in distinct driving trajectories and abnormal fuel consumption detection performs well, which demonstrates the FuelNet also can provide guidance for eco-driving strategies.

Keywords Fuel consumption prediction (FCP), Long short-term memory (LSTM), Deep network, FuelNet, Eco-driving strategies

Copyright ©
Energy Proceedings