Model Predictive Control for Engine Thermal Management System 2021-01-0225

A predictive control method for the cooling system of an engine is developed in order to improve fuel efficiency through the use of vehicle onboard GPS/Navigation system. Conventionally, in an internal combustion engine cooling system, coolant temperature is controlled from predefined maps or models depending on the engine speed, accelerator pedal position, engine torque, and/or fueling rate at that instant. Due to the instantaneous decisions taken to change target coolant temperature, road gradient changes in terrain could cause engine under-cooling on a steep uphill or over-cooling when driving downhill.

The paper presents the concept of predictive coolant temperature control strategy, utilizing GPS/Navigation data to recognize driving conditions by sensing vehicle position, speed limit, and road information like elevation and grade. The information is processed and predictions of future trajectory of vehicle thermal behavior can be identified by using model predictive control (MPC). In order to regulate engine operating temperature as close as possible to the desired levels that may yield the best fuel efficiency, optimal timing and profile of target coolant temperature can be determined based on cost function minimization and system/component constraints. For the prediction, a state-space engine thermal model is developed and calibrated with the simulation data generated by a high-fidelity vehicle model. One of driving scenarios for the application of this method is when the uphill slope of the road is detected and engine load is above a threshold, then the controller will adjust target coolant temperature based on current driving conditions prior to the event of high power demand. As a result, as compared to conventional control, the predictive control demonstrated energy savings due to delayed timing of knock mitigation with spark retard and reduced radiator cooling fan operation during high load demands. Simulations are conducted by using a high-fidelity GT-SUITE model and a driving cycle for proof of concept and to assess the benefit of this technology.