Construction and Study of Automotive Working Conditions Based on Improved K_means Clustering

Abstract

AbstractTo address the problem that the random selection of the initial clustering centers when using K-means clustering for the construction of working conditions cannot guarantee the quality and stability of clustering, a method of selecting the initial clustering centers based on the data's own characteristics is proposed. Taking the actual operating conditions of passenger cars in Zhengzhou city as the research object, a processing method combining short-trip division of test data, principal component analysis (PCA) and improved K_means clustering is realized by using MATLAB programming to construct the cyclic working conditions of passenger cars in Zhengzhou city. The results of the comparison with the typical working conditions constructed by traditional K_means clustering show that the error of the characteristic parameters of the working conditions constructed by the improved K_means clustering has been reduced, and the average error of the joint velocity-acceleration distribution and the actual full working conditions has been reduced from 0.55 to 0.43, which is a 21.8% improvement and proves the effectiveness of the improvement, and the accuracy of the constructed working conditions is higher and more comprehensive to reflect the actual traffic conditions in Zhengzhou. The accuracy of the constructed conditions is higher, which can reflect the actual traffic conditions of Zhengzhou city.KeywordsPassenger car driving conditionsShort trip classificationPrincipal component analysisImproved K_means clustering