Research on Identification Approach of Risky Driving Behaviors for New Energy Vehicles Based on Internet of Vehicles Data

In order to study the driving style and driving risk of drivers for new energy bus, 14 new energy buses running on urban, suburban and rural roads were selected as the research objects, and 1 month driving data were obtained. The characteristic indexes of acceleration, deceleration, acceleration instantaneous change rate, velocity fluctuation rate and bearing rate were extracted, and the drivers were divided into three categories: aggressive, normal and calm by non-supervised learning threshold classification (K-medoids) method. The results show that aggressive drivers have larger measurement indexes in the five indexes. Aggressive drivers not only frequently use sharp acceleration and deceleration driving behaviors, they also use aggressive driving behaviors such as sharp turn and speed change. Aggressive drivers tend to use the accelerator pedal violently while conservative drivers tend to use the accelerator pedal gently. With the decrease of driving aggressiveness, the data points in the potential risk area are significantly reduced, the distribution of acceleration tends to converge to zero, and dangerous behaviors such as rapid acceleration and rapid deceleration are significantly reduced.