VMD-CNN-BiGRU Model for Vehicle Fuel Consumption Prediction Considering Spatial-Functional Road Characteristics

To accurately predict the fuel consumption of vehicles, this study proposes a vehicle fuel consumption prediction model based on the VMD-CNN-BiGRU algorithm by considering six road spatial features such as road grades, one-way road attributes and intersection attributes. First, the VMD algorithm is employed to reduce the nonlinearity and nonsmoothness of the raw data by determining the optimal number of VMD decomposition modes. Then, the CNN-BiGRU algorithm is used to predict each modal component after decomposition, and the obtained prediction results are compared and analyzed with the prediction results of existing CNN-BiGRU, EMD-CNN-BiGRU and EEMD-CNN-BiGRU models. The results show that the VMD-CNN-BiGRU model significantly outperforms other models in terms of prediction performance and can accurately capture the trend of vehicle fuel consumption, thus effectively verifying the superiority and feasibility of the model. In addition, this study provides an in-depth analysis of the correlation between vehicle fuel consumption and time as well as spatial-functional characteristics of roads. The results show that the macroscopic distribution of vehicle fuel consumption is consistent with the distribution during peak traffic hours, while there are significant differences in vehicle fuel consumption under different road conditions.