Driving Cycle Development for Urban Bus using Principal Component Analysis and DBSCAN Clustering: With the Case of Haikou, China

Abstract

Driving cycles are an important means for new vehicle technology development and emission prediction and evaluation. To establish a representative driving cycle for urban buses in Haikou city, in this paper, the principal component analysis (PCA) and DBSCAN cluster algorithm are applied to develop the driving cycle. Firstly, a large number of vehicle driving data are collected, which comprised of 12 characteristic parameters. Next, the PCA is employed to extract main components from the characteristic parameters of driving data and the DBSCAN cluster is used to select representative micro trips. Subsequently, several most representative micro-trips were picked out to form the driving cycle. The effectiveness and uniqueness of the developed driving cycle are verified via comparing the parameters with the real-world driving data and the existing driving cycles, respectively.