A Hybrid Electric Vehicle (HEV) can achieve a considerably higher overall fuel economy than conventional vehicles. However, the fuel economy of HEVs strongly depends on the control strategy of the hybrid powertrain. Compared to conventional non-predictive hybrid control systems, predictive control methods can significantly further improve the fuel economy by anticipating future driving conditions such as hills, curves, speed limits, traffic flows, and in the future even geo-fenced low emission city zones. However, the selection of the driving mode and power distribution between multiple power sources which achieves the highest overall energy efficiency considering predictive information along the future driving route is a highly complex task. This complexity is further increased when taking into account the feasibility of the control actions, for example considering aspects such as driveability, NVH, emissions and lifetime durability, as well as real-time capability for online implementation on an HCU (Hybrid Control Unit). AVL’s predictive control methodology offers a flexible approach applicable to different hybrid types (e.g. HEV, Plug-in HEV and Range Extender) and powertrain topologies (e.g. parallel and series). It provides a real-time applicable control strategy that reduces the fuel/energy consumption, based on the preview of the future road type and topography and the likely velocity profile along the route, typically received from a digital map system in the form of an eHorizon, which may be enhanced with dynamic information from V2X. The control method consists of a long-horizon estimation that determines an energy efficient battery depletion strategy, which is subsequently tracked by an optimization-based short-horizon control method to select the driving mode and power distribution. The control method provides control requests for the different power sources, which can additionally explicitly consider constraints for feasibility and driveability (e.g. frequency of engine starts) to avoid unacceptable control policies. The performance of the control method is validated towards a globally optimal strategy obtained by an offline optimization using dynamic programming. The concept has been implemented on HCU hardware with limited computational resources, where its real-time capability was evaluated. In this paper, the predictive control methodology was applied on a hybrid passenger vehicle and the effectiveness of the strategy was demonstrated in a simulation environment. The predictive control method shows significant improvements in fuel economy, compared to a non-predictive and heuristic baseline control strategy.

Dr. Arno Huss, AVL List GmbH, AUSTRIA Dr. Stephen Jones, formerly AVL List GmbH, AUSTRIA Mr. Sander Boksebeld, AVL List GmbH, AUSTRIA Mr. Niklas Wikström, AVL List GmbH, AUSTRIA Mr. Gerald Teuschl, AVL List GmbH, AUSTRIA