Influence of the prediction horizon length of a PHEV energy management on fuel consumption

The use of information about the future vehicle trajectory is especially advantageous for the energy management strategies of Plug-in Hybrid Electric Vehicles. This is based on the fact that for minimal fuel consumption the stored electric energy should be consumed until the end of the trip, if the trip length exceeds the electric range of the vehicle. Therefore, best results are achieved by an optimization of the torque distribution between both electric motor and combustion engine knowing the whole trajectory until the next use of a recharging station. Due to the long recharging times this means usually an optimization until the end of the trip. A drawback of such long predictive horizons is the high computation cost. Another is the increasing model uncertainty due to the use of simplified powertrain models for the prediction algorithm and also the reliability of the predicted trip information. Therefore, one aim is to reduce the prediction horizon as much as possible without increasing significantly the fuel consumption. To save computation cost of the optimization and decrease the influence of model uncertainties, in this paper an energy management for Plug-in HEV calculating the global optimum for the whole trip is compared to optimization with different prediction horizon lengths. To define the desired SOC at the end of the prediction horizon a linear reference SOC function is used. Depending on the chosen prediction length the trajectory is divided into several sections, each one standing for one prediction horizon. At the entrance to every section the energy management calculates the optimal torque set point for the whole next section (prediction horizon). In order to exclude the influence of the optimization algorithm, Dynamic Programming is used to calculate the global optimum.