Fuel economy has always been a key performance indicator of automobiles, which is also an urgent problem to be addressed in environmental protection. At present, there are many methods for fuel consumption measurement. However, most of them only achieve monitoring of automobile fuel consumption, which cannot quickly and accurately evaluate the fuel economy of different models. In this article, based on real-time automotive controller area network bus data, a driving scenario method is proposed to divide the common driving process into short time units, classify and analyze automotive data under different driving states, establish evaluation models under each scenario, and design scoring methods according to the characteristics of each model to comprehensively evaluate the fuel economy of automobiles. The study shows that the evaluation results are not affected by driving environment or driving behavior. The fuel consumption calculation error is only 0. 4 after the prescribed driving time and road test, and the maximum evaluation error of the same car is only 4% . This method can provide car buyers with more accurate fuel economy evaluation reports for their models, avoid the inconvenience of field test drives, effectively improve the purchase rate of desirable vehicles, and can promote the emerging online car market.