Improving Velocity Prediction in Electric Vehicles using Hybrid Artificial Neural Network (ANN)

Velocity prediction is an integral part for the development of robust Energy Management System (EMS) of an Electric Vehicle (EV) which essentially enhances the performance and life cycle of the vehicle. In this paper an ANN based approach combining Back-propagation Neural Network (BPNN) and Radial Basis Function Neural Network (RBFNN) is used to forecast velocity on different prediction horizons. These methods are tested on two conventional driving cycles viz. Manhattan and WYU driving cycle and one mixed cycle which is created by combining different random driving cycles. The results are studied in terms of Root Mean Square Error (RMSE) where the proposed network yields the least value in all the cases as compared to conventional BPNN method. The results proved the robustness and adaptability of the proposed method which can be used in practical applications.