Study On Vehicle Driving Condition Based On Principal Component Analysis And Clustering Algorithm

Nowadays,with the rapid development of economy,the number of cars in China is increasing,which not only brings heavy pressure to the urban traffic,but also makes the automobile exhaust become the main source of urban air pollution.As is known to all,vehicle driving condition is an important basis for quantifying vehicle exhaust emissions and fuel.It can fully reflect the actual driving mode of the vehicle so that the reliable estimates of vehicle emissions and fuel conditions obtained.In 2019 the national graduate mathematical modeling D questions provided by the data as sample,two kinds of combination clustering method is used to build the car working condition diagram,a genetic simulated annealing algorithm and fuzzy C-means clustering with the combination of clustering method,another for self-organizing mapping neural network with particle clustering algorithm combining the clustering method.Firstly,after a series of data preprocessing,493467 sample data were divided into 2348 kinematics segments,and the selected characteristic parameters were calculated.The improved principal component analysis method was used to reduce the dimensionality of the characteristic parameter matrix,and the first three principal components whose cumulative contribution rate was more than 85%were extracted,and the scores of the first three principal components of all fragments were calculated.Then,two combination clustering methods and two traditional clustering methods are used for clustering:(1)Since K-means clustering is easy to fall into local optimum and the initial value selection of particle swarm optimization algorithm is arbitrary,the self-organizing mapping neural network and particle swarm clustering method are used for clustering.(2)Because fuzzy C-means clustering is also easy to fall into local optimum,genetic simulated annealing algorithm is adopted to optimize it and then cluster it.(3)In order to compare with the clustering method combining self-organizing mapping neural network and particle swarm clustering,the traditional K-means clustering is used for clustering.(4)In order to compare with the fuzzy C-means clustering optimized by genetic simulated annealing algorithm,the traditional fuzzy C-means clustering is used for clustering.After clustering with the above method,representative segments are selected according to the method of minimum absolute error to fit the driving pattern.Based on the characteristic parameters and the velocity-acceleration probability distribution,the effectiveness of the synthesis condition was verified.The results show that the average relative errors between the driving conditions and the original driving conditions are 6.87%,6.46%,16.17%,15.02%,respectively,by the self-organizing mapping neural network combined with particle swarm clustering,the genetic simulated annealing algorithm optimized fuzzy C-means clustering,and the traditional K-means clustering and fuzzy C-means clustering.Compared with the traditional method,the driving pattern synthesized by the two improved algorithms has high accuracy and small error,which can better reflect the traffic condition information contained in the original data.Finally,the figure of running condition of a city built in this paper,with the domestic Beijing,Shanghai and foreign typical working condition,Japan 10 and FTP75-15 comparing the NEDC conditions,can be concluded that the city’s road traffic and other areas at home and abroad of road conditions exist significant differences,this also reflected the importance of building in urban road conditions.