An Exponentially Varying Speed Prediction Method Based on SVM Recognition

The accuracy of vehicle speed prediction has an important influence on the energy management strategy of MPC (Model Prediction Control, MPC), and the exponentially varying prediction model has been widely used in the vehicular speed prediction in future driving cycle. Under a certain prediction horizon, the accuracy of the exponentially varying prediction model mainly depends on the selection of the decay coefficient, but the decay coefficient of the traditional exponentially varying prediction model is fixed. The traditional exponentially varying prediction model cannot meet the prediction accuracy of different driving cycles due to the fixed decay coefficient. In the paper, an exponentially varying prediction based on tunable decay coefficient using SVM (Support Vector Machine, SVM) to recognize the driving cycle is proposed. The simulation results show that the proposed SVM-based exponentially varying speed prediction model can effectively decrease the RMSE (Root Mean Square Error, RMSE) of the vehicle speed prediction, which improves the accuracy of vehicle speed prediction, and provides a foundation for the MPC-based energy management.