In order to reduce pollutant and CO2 emissions and fulfill future legislative requirements, powertrain electrification is one of the key technologies. In this context, especially 48V technologies offer an attractive cost to CO2 reduction ratio. 48V mild hybrid powertrains greatly benefit from additional electric intake air compression (E-Charging) and direct torque assist by an electric machine (E-Boosting). Both systems significantly improve the transient engine behavior while reducing the low end torque drawbacks of extreme downsizing and downspeeding.
Since E-Charging and E-Boosting have different characteristics concerning transient torque response and energy efficiency, application of model predictive control (MPC) is a particularly suitable method to improve the operating strategy of these functions. MPC requires fast running real-time capable models that are challenging to develop for systems with pronounced nonlinearities. Hence, the focus of this study is on the process modeling of a 48V mild hybrid system with an electric compressor for applying model predictive control algorithms.
Firstly, the problem is formulated by investigation of real world measurements of a 48V mild hybrid demonstrator vehicle with the target powertrain configuration. Thereto, full load accelerations with a rule based and performance oriented implementation of E-Charging and E-Boosting are analyzed. The time response behavior of optimization relevant parameters such as output torque and corresponding energy consumption is identified. Secondly, a detailed co-simulation of a 48V powertrain with a turbocharged gasoline engine, a belt-driven starter generator and an electric compressor is set up. The component models are parameterized by experimental data. Thirdly, this co-simulation plant model is used to analyze various real-time process models, which are designed for MPC purposes. Starting from a semi physical process model containing the nonlinear system dynamics, simplifications for several model parameters are considered to reduce complexity. Finally, a linearization of the nonlinear process model is assessed concerning its applicability in a MPC.