Probabilistic Optimal Control for Energy Management of Connected Hybrid Electrical Vehicles

Energy management has an important impact on fuel consumption and CO
₂
emissions of hybrid vehicles. Nowadays, connected hybrid vehicles are able to predict valuable information to the energy management task. Predictive time-global model based methods has been proven to be efficient. In this paper, a predictive optimal control formulation is proposed, where the prediction of the future power demands are modelled as random variables, and a probabilistic shooting algorithm is developed. Numerical simulations with normally distributed predictions on real driving cycles are performed. The results show a satisfactory robustness with prediction horizons of several minutes.