Comparison of Velocity Forecasting Strategies for Predictive Control in HEVs

Abstract

The performance of model predictive control (MPC) for energy management in hybrid electric vehicles (HEVS) is strongly dependent on the projected future driving profile. This paper proposes a novel velocity forecasting method based on artificial neural networks (ANN). The objective is to improve the fuel economy of a power-split HEV in a nonlinear MPC framework. In this study, no telemetry or on-board sensor information is required. A comparative study is conducted between the ANN-based method and two other velocity predictors: generalized exponentially varying and Markov-chain models. The sensitivity of the prediction precision and computational cost on tuning parameters in examined for each forecasting strategy. Validation results show that the ANN-based velocity predictor exhibits the best overall performance with respect to minimizing fuel consumption.