Construction Of Driving Cycle For Light-duty Vehicles In Fuzhou Based On Statistical Methods

With the growth of car ownership,the problem of air pollution caused by car exhaust emissions has seriously affected people's daily lives,and the energy shortage caused by the continuous increase in fuel consumption has become more serious.Driving cycle not only is an important basis for testing the energy consumption and emissions of the car,but also the main benchmark for the performance of the car.In order to reflect the driving conditions of light-duty vehicles in Fuzhou,basing on the vehicles driving data of Fuzhou,this paper constructs two driving cycles by using the improved K-means clustering method and Markov method,respectively.The main contents are as follows:(1)Since the K-means clustering method is sensitive to the initial center and the iteration is prone to fall into the local optimal solution,this paper uses Self-organizing Maps(SOM)algorithm to optimize the initial center of K-means clustering.Firstly,the number of clusters is determined by the Davies-Bouldin index and Silhouette index.Secondly,the weight vector trained by the SOM algorithm is taken as the initial center of the K-means clustering.Finally,the most representative result of each cluster is selected to constitute the driving cycle.(2)Based on Markov chain theory,this paper considers the velocity-time series as a Markov process,and constructs the corresponding sample space and state space.The transition probability matrix of the Markov chain is determined by the maximum likelihood estimation.Then a suitable initial state is affirmed and its next state is determined by the Monte Carlo method.At last,the appropriate samples that constituting the driving cycle are selected according to the previous identified states.(3)The driving cycles constructed by two different methods are compared with test data in terms of the characteristic parameter,velocity frequency distribution,acceleration frequency distribution and speed-acceleration combined frequency distribution.In addition,the two driving cycles are compared with other standard driving cycles.The results show that the average absolute errors about characteristic parameter of the SOM-clustering method and the Markov method are 0.8808% and 3.8720% respectively,both within 10%.Besides,the SOM-clustering method is a little better than Markov chain method in other aspects.Furthermore,the two driving cycles constructed in this paper can better reflect the actual operation of light-duty vehicles in Fuzhou more than other standard driving cycles.