The 20th and 21st joint COTA International Conference of Transportation Professionals (CICTP 2020-21)

In real world, on-road vehicle energy consumption is affected by various elements. In order to better understand the impact of these elements on fuel consumption, this study has built two mesoscopic fuel consumption prediction models to compare the performance between nonlinear fitting method and neural networks and to investigate the influence of speed and acceleration on the calculation results of fuel consumption factor based on Vehicle Energy Dataset, which were collected from 383 personal cars in Ann Arbor, USA. The time-snipping based segregation method is used to divide vehicle data into smaller time-snipping with fixed time interval. To obtain more detailed parametric analysis, the effects of speed and acceleration on fuel consumption factors have been evaluated under different speed ranges, respectively. The results show that the performance of neural network is better than nonlinear fitting method and acceleration has a greater impact on fuel consumption factor at low speed compared with high speed.

fuel consumption;mesoscopic;nonlinear;neural network;time-snipping