On modeling microscopic vehicle fuel consumption using radial basis function neural network

Abstract

Energy conservation is one of the central challenges for the transportation system today. A variety of microscopic vehicle fuel consumption models have been developed to support eco-friendly transport strategies. However, most existing models are regression based and are sensitive to the vehicle-specific parameters and the operating conditions, therefore, an expensive and time-consuming calibration procedure is always indispensable in these models' application. In this paper, we propose an artificial neural network-based model to avoid the calibration problem. The main works include: (1) collect extensive field datasets such as large-scale controller area network bus to reflect the local transportation environment's fuel consumption characteristics; (2) conduct correlational analysis to identify the key fuel consumption influence factors; (3) develop a radial basis function neural network-based learning model to capture the nonlinear relationship between the key factors and the corresponding fuel consumption values based on the collected training datasets. The proposed model can give a reasonable prediction of instantaneous fuel consumption without calibration. The effectiveness of the proposed model is validated from a combination of both in-lab and field experiments.