Route specific Driver Characterization for Data-Based Range Prediction of Battery Electric Vehicles

The development of accurate range prediction algorithms for battery electric vehicles (BEV) is an important requirement for gaining the long-lasting confidence of future users in this technology. The driver, who strongly contributes to the energy consumption of a vehicle, has to be considered for non-automated driving scenarios.This paper introduces a framework for the acquisition and processing of situation-related driving behavior in data-based models, such as support vector machines. The correlation of the driver's behavior to route, environment as well as vehicle characteristics is identified through lazy learning of segmental characteristic values. Suitable parameters are selected by correlation analysis and neighborhood component feature selection. In this context, a distinction is made between several road topologies like traffic lights and crossroads. During the prediction, the expected driver's behavior is determined by searching for the k-nearest-neighbors to the current segment parameters, representing the route, environment and vehicle. The identified driver-specific parameters are used afterwards together with other significant parameters for the training and the application of a data-based model, which allows an accurate prediction of the future demand for energy. The concept provides an online implementable solution, which uses existing information systems and thus could be used for real driving scenarios.