基于运动学片段的纯电动出租车行驶特征模式挖掘

Driving mode mining of pure electric taxi based on kinematic segments

在逐步推行出租车全面电动化的背景下,针对目前对纯电动出租车行驶状态评估的不足,建立一种基于运动学片段的纯电动出租车行驶特征模式挖掘方法,研究纯电动出租车行驶状态特征.首先,基于行驶轨迹GPS数据,从速度特征、加减速和行驶状态3个方面,确定超速比例、加减速频率、行驶速度、怠速时间占比等13个特征指标刻画运动学片段,建立纯电动出租车运动学片段提取方法,研究纯电动出租车行驶状态特征.然后,根据行驶特征指标主成分的特征值大小及累积贡献率,确定关键特征指标,结合K-均值聚类算法,生成多时空场景下的纯电动出租车行驶特征模式,综合评价车辆行驶状态.最后,以深圳市共计9天采样间隔为1 s的700万条纯电动出租车GPS行驶轨迹数据为驱动,提取了1 757条纯电动出租车运动学片段.根据安全性、效率性和舒适性8个关键特征指标进行聚类分析,生成包含主干路、次干路和支路在早高峰、平峰和晚高峰9种时空场景下27类纯电动汽车行驶状态的特征模式库.研究结果表明:综合安全性、效率性、舒适性3方面,早高峰期间的纯电动出租车综合行驶状态优于平峰和晚高峰时段;基于运动学片段、主成分分析及多时空场景聚类分析的纯电动出租车行驶特征模式挖掘方法,能够有效反映并评估纯电动出租车行驶状态,并向驾驶员提供合理的驾驶建议.

In the context of the gradual transition to comprehensive electrification of taxis,and address⁃ing the current shortcomings in evaluating the driving state of pure electric taxis,a method for mining driving characteristic patterns based on kinematic segments of pure electric taxis is established.This study aims to explore the driving characteristics of pure electric taxis.Firstly,utilizing GPS data from driving trajectories,13 feature indicators,including overspeed ratio,acceleration and deceleration fre⁃quency,driving speed,and idling time ratio,are determined from 3 aspects in terms of speed charac⁃teristics,acceleration and deceleration,and driving conditions to characterize kinematic segments.This establishes a method for extracting kinematic segments of pure electric taxis and subsequently studying their driving characteristics.Subsequently,based on the eigenvalues and cumulative contribu⁃tion rates of principal components derived from driving characteristic indicators,key feature indicators are identified.Through integration with the K-means clustering algorithm,a method is proposed for mining driving characteristic patterns based on kinematic segments of pure electric taxis,allowing the identification of driving characteristic modes in various time and space scenarios.Finally,leveraging 7 million GPS tracking data points from pure electric taxis in Shenzhen,with a sampling interval of 1 second over 9 days,1757 kinematic segments of pure electric taxi are extracted.Employing eight key feature indicators related to safety,efficiency,and comfort for cluster analysis,a feature pattern li⁃brary is generated,encompassing 27 classes of driving states for pure electric taxis on main roads,sec⁃ondary roads,and local roads during morning peak,normal hours,and evening peak periods.The re⁃search findings indicate that,combining the three aspects of safety,efficiency and comfort,pure elec⁃tric taxis travelled better during the morning peak than during the flat peak and evening peak hours.A pure electric taxi driving feature pattern mining method based on kinematic segmentation,principal component analysis and spatial-temporal scenario clustering analysis can effectively reflect and evalu⁃ate the driving status of pure electric taxis and provide reasonable driving suggestions to drivers.