Scenario MPC for fuel economy optimization of hybrid electric powertrains on real-world driving cycles

Optimization based energy management strategies for hybrid electric vehicles require a reliable forecast of the future driver torque demand to yield an appropriate performance in terms of fuel economy and charge sustainability. Apparently, for artificial driving cycles, which are mainly investigated on a testbed, a very accurate preview is available for optimization purposes. In contrast, for real-world driving cycles, which relate to driving on an actual road, this is a much more challenging task. The particular challenges arise with the uncertainties that e.g. originate from surrounding road traffic or even the driver itself. To cope with these uncertainties, we propose a scenario model predictive control scheme. Particularly, we draw independent and identically distributed random samples from an uncertainty interval around the nominal torque profile over the prediction horizon and as such optimize over a sequence of individual scenarios. Simulation results highlight that our control approach is fuel efficient and charge sustaining in spite of torque demand uncertainties while being real-time capable as well. At the same time, we will show that a classical MPC approach is not capable of being charge sustaining when these uncertainties are present.