Modeling and forecasting of injected fuel flow using neural network

The aim of this research is to develop an intelligent automated online forecasting of a car fuel consumption using neural network and classified it into classes of driving style. A new online monitoring tool was developed to acquire and analyze data collected from a car for the purpose of fuel consumption modelling and forecasting. The data was transmitted via ECU Can Bus attach to the car to the automotive single board computer. The online monitoring and forecasting tools were developed by using 8-bit Single-Chip Microcontroller as a data acquisition processor. Distance, speed, revolution, fuel flow, fuel consumption and temperature transducer are taped from the experimented car to gain the information. The multilayered perceptron network trained by Levenberg-Marquardt algorithm was selected as a black box model for forecasting purposes. The input variables were taped from car sensors. The data set consists of 2000 data samples. The first 1000 data were used for training and the rest were used in validation and forecasting process. Based on the best network execution, it was found that the best MSE during validation phase is about 0.0804 produced at the 26 hidden neurons. The results of the forecasting during training obviously show that during the first 200 data series the forecasting error is quite high but after 200 data series the neural network model have a tendency to improve quickly and forecast slightly the real value of the injected fuel flow.