Research On Urban Vehicle Driving Cycle Construction Based On Improved Clustering Method

With the advocacy of energy saving and emission reduction concept,automobile emissions and energy consumption testing have become the focus of enterprises,and driving cycle are an important basis for the model construction of vehicle exhaust emissions and fuel consumption.In addition,driving cycle are widely used in vehicle evaluation/selection/design/matching and traffic control/management/planning.China does not have its own working condition system currently,using the new European driving conditions as the official driving cycle for vehicle emission testing.Therefore,the construction of more accurate driving cycle in Chinese cities is meaningful to vehicle emission testing,new vehicle development and urban traffic state analysis.In view of the above problems,this paper research on the construction process of driving cycle,the evaluation of clustering methods and the improvement of K-means clustering method.Using the collected vehicle speed data of Beijing,Shanghai,Guangzhou and Changchun in China for6 months,driving cycle is built based on micro-trips method,principal component analysis,particle swarm optimization algorithm,clustering,etc.In the case of the construction process,the K-means clustering is improved with particle swarm optimization algorithm and its effectiveness was verified by chi-square test.The main contributions and conclusions of this dissertation are summarized as:(1)Research on the Construction Process of Vehicle Driving cycleThe traditional micro-trips method is divided into modules such as data filtering,segmentation,feature extraction,principal component analysis,cluster,driving cycles splicing.Based on the Matlab programming language,the Real-World Driving Cycle construction program is developed and optimized to automatically and quickly construct driving cycles that reflect the actual operation of the vehicle and the driver’s driving habits based on the input data.(2)Analysis of Clustering Methods based on Driving CycleClustering is widely used to aggregate Micro-trips with similar characteristics into groups,and determine the type of traffic conditions corresponding to each segment by analyzing its characteristic parameters,which has an important impact on affect the accuracy of driving cycle.For the clustering process in driving cycle construction,the clustering stability,efficiency,accuracy and sample adaptability of K-means clustering,Kmedoids clustering,fuzzy clustering and Gauss-mixture clustering are compared and analyzed.The results show that the K-means clustering performance is superior,but it is susceptible to the initial center.(3)Improvement of K-means Clustering Based on Driving CycleAiming at the problem that K-means is sensitive of initial value and easy to fall into local optimal solution,the initial value can be optimized according to the characteristics of sample data.The 90% confidence interval is selected as the initial center range,which excludes edge points and isolated points,improves the stability significantly while ensuring the cluster effect.In addition,the particle swarm optimization algorithm is used to initialize the initial clustering center of the K-means clustering.The sample point density is used as the optimization target,and the points in the region with relatively long relative distance and large sample density are optimized as the initial clustering center.The stability and effectiveness of the improved K-means clustering method are verified by calculating the average deviation of the clustering center and the chi-square test value of the velocity distribution frequency.(4)Construction and Analysis of Urban Driving CycleThe improved K-means clustering method is applied to constructing driving cycle of Beijing,Shanghai,Changchun and Guangzhou.The reliability and effectiveness of the improved K-means clustering method in constructing driving cycles are verified by the comparison of eight kinematics parameters and the speed-acceleration joint distribution of the comparative conditions.Comparing the urban driving cycle in China with other typical driving cycles around the world,it shows that Chinese cities have frequent acceleration and deceleration,low running speed,high proportion of idle speed and low speed,which is different from the current domestic NEDC driving cycle.The driving cycles constructed in this study can reflect the actual traffic situation of Chinese cities and have practical value.