Driving Event Recognition of Battery Electric Taxi Based on Big Data Analysis

Abstract

Personal driving behavior affects vehicle energy consumption as well as driving safety; therefore, driving behavior is key information for electric vehicle (EV) energy management and advanced driver assistance systems. Eco-driving is an efficient way to reduce energy consumption and air pollution. As the basis of driving behavior, limited types of driving event information, as used in several other studies, cannot be used to meet eco-driving evaluation study needs. Complex and inconsistent human-defined rules are not conducive to the establishment of driving events. Hence, it is necessary to establish a driving event classification system with more categories of drive-topics that can present a better linkage between driving behavior and energy consumption. This paper proposes a driving event recognition method. Dynamic Local Minimum Entropy is proposed, and the Latent Dirichlet Allocation algorithm is used to classify different driving events. Drive-topics are proposed which describe driving events more accurately. The data from fifty battery-electric taxis are used to train the algorithm with data collected by the Service and Management Center for EVs, Beijing, in 2018. The relationship between drive-topic and energy consumption is analyzed to demonstrate that driving behavior can be established using drive-topics to support the evaluation of eco-driving for battery-electric vehicles.