Calibration and Demonstration of Vehicle Powertrain Thermal Management Using Model Predictive Control

Control of vehicle powertrain thermal management systems is becoming more challenging as the number of components is growing, and as a result, advanced control methods are being investigated. Model predictive control (MPC) is particularly interesting in this application because it provides a suitable framework to manage actuator and temperature constraints, and can potentially leverage preview information if available in the future. In previous SAE publications (2015-01-0336 and 2016-01-0215), a robust MPC control formulation was proposed, and both simulation and powertrain thermal lab test results were provided. In this work, we discuss the controller deployment in a vehicle; where controller validation is done through road driving and on a wind tunnel chassis dynamometer. This paper discusses challenges of linear MPC implementation related to nonlinearities in this over-actuated thermal system. Specifically, the fan and grill shutter actuators have a nonlinear influence on the individual airflows through the charge air cooler and radiator heat exchangers, and the design choices in dealing with these nonlinearities affects the control performance and controller memory requirements. The memory requirements of the resulting controller are also analyzed and compared to other MPC controllers.