Abstract
This study aims to develop fuel consumption (FC) simulation models for motorcycles (MC) based on the known data of MC’s actual FC and driving features. These data were gathered using a data logger that was directly connected to the test MC to record data continuously with the one-second time step. The collected data was preprocessed to minimise random errors before using them to establish predicting models of FC for MC. A strong correlation between the vehicle’s FC and real-world driving features of vehicle-specific power and speed was recognised. Three FC prediction model types were developed based on the processed data for both on-peak and off-peak hours. Developed prediction models have good reliability with obtained all R2 in the range of 0.95–0.99. Total errors that appear when using the developed prediction models to estimate the average FC per kilometre of MC are from 3.2% to 18.9%. Among them, the average speed-based FC prediction model is the best, with a deviation of FC ranging from 3.2% to 5.4% for both on-peak and off-peak hours. This implies that the vehicle’s average speed could reflect its FC feature in terms of litre per kilometre rather than the vehicle’s VSP.
Acknowledgements
The authors would like to thank the Research Center for Propulsion Systems and Autonomous Vehicles, SME, HUST, Hanoi, Vietnam for supporting the laboratory test.
Disclosure statement
No potential conflict of interest was reported by the author(s).