Abstract: Aiming at the problem that the current driving cycles were difficult to reflect the real driving conditions of vehicles, taking Xian—a typical large city in China as an example, the construction method of driving cycle of electric vehicles was studied. Firstly, according to the road layout of Xian, the data acquisition scheme of urban road driving condition was designed. Then a semi-supervised classification model combining K-means clustering and support vector machine was proposed to construct the Xian driving cycle. Finally, the Xian driving cycle was compared with the original test data and the international standard driving cycles. The research results show that the relative error between the characteristic parameters of the Xian working condition and actual road driving data is less than 5%, and the average relative error is only 2.66%, which indicates that the constructed driving cycle can truly reflect the motion characteristics of vehicles in Xian. Moreover, because of the difference of power system, the driving cycle of electric vehicle is more radical than that of internal combustion engine vehicle.

Key words: vehicle engineering, electric vehicle, driving cycle, K-means clustering, support vector machine

摘要： 针对现有行驶工况难以反映车辆真实驾驶情况的问题，以国内典型大中型城市——西安市为例，对电动汽车行驶工况构建方法进行研究。根据西安市道路布局，设计了城市道路行驶工况数据采集方案；提出了一种K-均值聚类和支持向量机相结合的半监督分类模型，构建了西安工况；最后将西安工况与原始试验数据和国际标准行驶工况进行对比。研究结果表明：西安工况与实际道路行驶数据特征参数的相对误差均小于5%，平均相对误差仅为2.66%，构建的行驶工况能够真实反映西安市车辆的运动特征；且由于动力系统的差异，电动汽车工况比内燃机车工况更为激进。

关键词: 车辆工程, 电动汽车, 行驶工况, K-均值聚类, 支持向量机