Online optimization based energy management of hybrid electric vehicles using direct optimal control

Abstract

Hybrid Electric Vehicle Energy Management (HEV-EM) is being extended beyond the standard power/ torque splitting formulations to accommodate conflicting objectives such as fuel efficiency, battery aging and engine-out emissions. Pareto front based approaches are employed for Design of Experiments (DoE) -based calibration of deterministic rule based HEV-EM strategies using offline optimization [1]. Vehicle connectivity and Plug-In Hybrid Electric Vehicle (PHEV) have changed the system boundary conditions for the EM problem. Increased awareness of real driving efficiency, data availability during operation and advances in computation has motivated the application of online-optimization based strategies [2]. The focus had been on extension of Equivalent Consumption Minimization Strategy (ECMS) with additional dynamics, optimization criteria and constraints. This approach faces challenges such as availability and inaccuracy of the modeled dynamics as well as handling of adjoint state dynamics [3].

This paper proposes an alternative method to optimize engine on/off state and torque split between the energy converters using predictive information within a limited time horizon by extending the Model Predictive Control methodology [4] based on direct optimal control. An important advantage of the method is that it does not require explicit modeling of the additional dynamics in addition to its ability to handle state constraints directly and its real-time capability. Further, the proposed method is flexible in handling change of control objectives as well as variation of control weighting and reduces calibration effort due to reduced number of functional parameters. The advantages of this methodology will be demonstrated exemplarily with a 2-cylinder Combustion Engine Assist (CEA) PHEV powertrain EM as a use-case considering battery stress and aging.