PREDICTION OF PASSENGER CAR FUEL CONSUMPTION USING ARTIFICIAL NEURAL NETWORK: A CASE STUDY IN THE CITY OF NI？

The reduction of CO 2 emission which is in direct relationship with fuel consumption is of prime importance for the future sustainable use of passenger cars. For the given passenger car, the fuel consumption in urban areas is mostly affected by the conditions related to traffic and driving behavior. In this paper, an artificial neural network model for the prediction of passenger car fuel consumption in the City of Ni？ was developed based on experimentally measured data recorded through on-board diagnostics equipment. Fuel consumption was assumed to be a function of car speed, a city zone, an hour of day and a day of week. A comprehensive preliminary investigation revealed that single hidden layer artificial neural network model having ten neurons can be efficiently trained with Levenberg-Marquardt algorithm to provide satisfactory prediction accuracy. Finally, the analysis of effects of the selected independent variables on the fuel consumption was discussed based on twelve 3D surface plots