Construction Of Actual Driving Cycle Of Shenzhen Road Based On Clustering And Python

Typical driving cycle as an important indicator of fuel consumption and emission certification testing for automobile have extremely important values.However,our country's standards based on early European driving cycle(NEDC,New European Driving Cycle)are no longer meet the actual traffic conditions in China.In order to build an unified driving cycle standard that meets the actual road conditions in China,the Ministry of Industry and Information Technology commissioned the China Automotive Technology and Research Center to begin collecting data on vehicle traffic in 41 sample cities including Shenzhen since2015.This paper takes the vehicle traffic in Shenzhen as the research object and constructs the driving cycle in Shenzhen.This paper based on the clustering method and Python language to constructed the driving cycle in Shenzhen.First of all,using the autonomous driving method to collecting data for the entire year of the vehicles,a total of 61 vehicles were collected,of which 38 were light vehicles and 23 were heavy trucks.Using OBD interfaces as well as GPS to collecting data.Secondly,selected 16 characteristic parameters,divided speed-time data to short trips segments,and analyze the collected data using the method of “data cleaning – short trips division-dimension reduction-short trips cluster classification-short trips selection”.Next,using the principal component analysis method to reduce dimension and filter short trips,the16-dimensional light short trips parameter data were reduced to 4D and 2D,respectively.Then,using MiniBatchKMeans and AP clustering algorithm to cluster,4 and 137 categories short trips were obtained.Finally,use the correlation coefficient to extract short trips segments in each category to build the driving test cycle in Shenzhen.The main features of construction cycle are: the light vehicle driving cycle o perating time:1800s,acceleration ratio:38.59%,deceleration ratio:39.65%,and constant speed ratio:5.0%,idling ratio:16.76%;bus driving cycle operating time:1800s,acceleration ratio:33.36%,deceleration ratio:30.46%,cruise rate:29.41%,and 6.75% idle speed ratio.In the above research process,Python language was used as an analysis tool,and the clustering method was used for short trips classification.Generally,people use K-Means to build driving cycle,but the K-Means have disadvantage of slower convergence in large-scale data and needs to determine their own clustering categories.This paper combined with using the method MiniBatchKMeans processed large-scale data and AP clustering to process short trips parameter data after dimension reduction.Compared with MiniBatchKMeans,the improved AP clustering is significantly better than MiniBatchKMeans for short tripsclassification of light vehicles,while the classification effect of heavy duty trucks is not obvious.Compared with the original data,it is concluded that the driving cycle constructed in this paper can represent the actual traffic conditions in Shenzhen.Compared with existing driving cycles at home and abroad,there is a large gap between them,indicating that the driving cycles abroad are not applicable to domestic traffic.The research results of this paper have laid a theoretical foundation for the formulation of the new laws and regulations in Shenzhen,and have a high practical value and theoretical reference significance.