摘要
为提供排放试验所需的车速曲线,基于划分的短行程数据,采用K-均值聚类算法构建了西安市汽车行驶合成工况。首先对采集的原始数据进行短行程划分并进行特征提取,针对提取的高维特征向量之间的冗余性和非线性关系,采用核主分量分析法进行降维。然后基于K-均值的聚类算法,对降维后特征向量进行划分,按照离聚类中心最近的原则选择各聚类的短行程样本,将其合成为平均速度为21.51 km/h、持续时间为1 166 s、距离为6.9 km的西安市汽车行驶工况。对比表明,西安市汽车行驶工况接近于日本J10-15标准,但加速度参数要高很多。
In order to provide the vehicle speed profile for emission test, we use the K-means clustering method to develop a car driving cycle for the city of Xi'an based on the short trip data. Firstly, the original data is divided into micro-trips, then the feature extraction of micro-trips is performed. To eliminate the redundancy and nonlinear relations which exist in the high dimension vector, the Kernel Principal Component Analysis method is used to dimension reduction of the feature vector. Lastly, the reduction vectors are clustered into 3 groups using the K-means clustering method. According to the nearest micro-trip principle, samples to the cluster centers are selected as representative samples, based on which, the driving cycle is developed, which contains a 1166 s speed time series, with an average speed of 21.51 km/h and a distance of 6.9 km in Xi'an city. The characteristics of the developed driving cycle are compared with some other vehicle driving cycles, the comparison results show that the developed driving cycle is similar to J10-15 mode of Japan, but the former has much higher acceleration parameters.
出处
《汽车技术》
北大核心
2015年第8期33-36,共4页
Automobile Technology
基金
长安大学中央高校基本科研项目(310822151029)资助
