A Data-Driven Control Strategy for Trip Length-Conscious Power Management of Plug-In Hybrid Electric Vehicles

Abstract

This paper presents a novel data-driven control strategy for the computationally efficient power management of plug-in hybrid electric vehicles (PHEVs). The proposed method relies on a set of real-time control policies trained through a linear regression process based on a large set of optimal powertrain decisions obtained from dynamic programming. The control policies receive the real-time powertrain system information such as the demanded propulsion force, vehicle speed, battery state-of-charge, etc. to compute the required torque values for the engine and the electric drivetrain system. The proposed controller makes near-optimal decisions when it is evaluated for the same test conditions as trained. When the test and training settings are different, however, the controller decisions deviate from optimality. We show that this deviation can be mitigated by including future drive cycle information such as trip length in the control computations.