Methodological Study On The Urban Driving Cycle Development Based On GPS/GIS Data

Driving cycle serves as the standardized procedure for testing and certifying the fuel economy and emission of conventional vehicles,as well as the driving range of new energy vehicles.Also it is the basis to evaluate emerging vehicular technologies.Driving cycle with good representativeness can decrease the gap between the fuel economy and emission tested in laboratory and the values measured on roads,which not only provides reliable reference for customers when choosing vehicle products,but also offers guidance in new vehicular technology research and vehicle product development.Besides,a legislative driving cycle that can represent the real driving conditions will certainly contribute to the implementation of the energy saving and emission reduction polices,which is helpful in maintaining the government's credibility.Aiming at the fact that the in use driving cycle fail to represent the real urban driving characteristics,the contents of this thesis are as follows.(1)The pre-process procedure for the raw driving data were designed to converted raw driving data base into raw short trip data base,which includes short trip segmentation,filtration and elimination to get rid of repeated and false data.Taking the fuel consumption of short trips as the dependence variable,correlation analysis and multi linear regression were applied to choose the 10 variables that can best describe the characteristics of the short trips.(2)Then the VHT of the urban road network were estimated by using the GPS/GIS traffic data together with the three phase speed and traffic volume model built according to TPI.The characteristic short trip data base was obtained after different weights given to area and period of time.(3)The 10 characteristic variables of every short trip were calculated to obtain the variable matrix which then got dimensionality reduced to 4 with principal component analysis.K-means clustering was adopted to classify the short trips into 3 category including low speed,middle speed and high speed.The data point with extreme variable value were excluded by checking its silhouette value and the remaining data were used to construct driving cycle by random selection.The Markov method was also adopted to develop a cycle for every speed category.(4)The 2 driving cycles generated by using different methods were compared to real driving conditions and the differences between them were evaluated by using characteristic variables and distribution.The result showed that cycle generated by kmeans got an average relative error of 5.50% in characteristic variables and 1.77% in characteristic distribution.While the same error of the cycle generated by Markov method was respectively 3.6% and 1.35%,which proved that the driving cycle developed using Markov method is more similar to the real driving conditions.This research showed that the driving cycle generated by data weighting and using Markov method can represent the real driving condition in urban area which can provide reference for legislative driving cycle development.