Research Of The Construction And The Oscillation Properties Of The Driving Cycle

Driving cycle is a series of speed-time curves, which can reflect actual roaddriving features of a certain type vehicles in a particular city or area. It is mainlyused to test and reasonably evaluate fuel economy and exhaust emissions capabilityof vehicles and engines, therefore the vehicle can be made further optimization.Because the motor vehicle and non-motorized vehicle are mixed on the road in ourcountry, so the city structure, control measures and driving habits in our country arevery different from that in developed countries, such as European country and USA.Therefore extremely difference exists in the driving cycle. Developing a drivingcycle which is suitable for our national conditions is definitely necessary.Based on passenger car in Hefei, the paper confirms the road test plan for dataacquisition of driving cycle and12characteristic parameters which can representdriving feature are introduced. The combination of multivariate statistical theoryand SOM neural network is used in the construction of driving cycle. The accuracyof the cycles constructed by the improved clustering and the single k-meansclustering are analyzed from the errors in characteristic parameters value, the jointprobability distribution for speed and acceleration, nonparametric test of theacceleration distribution and the simulation of fuel consumption of ADVISORsoftware. The clustering accuracy is increased and the error of the fitting cycledecreases. So it can reflect the real urban traffic conditions comprehensively.As traffic congestion and traffic signal control, the deceleration and stop beforeentering intersection and the acceleration when leaving intersection will make the drivingconditions show obvious fluctuation phenomenon. Such fluctuations can cause the driverless comfortable, slow down the speed and increase more fuel consumption and airpollutants. Therefore, the study of the fluctuation is particularly important. The typicalintersection condition data is dealt with detrended fluctuation analysis and waveletanalysis method, then the dominant fluctuation amplifications of different datapoints before and after intersection are determined by spectrum analysis method.This is important for the study of the energy regulation of the intersection.