Nichtlineare modellprädiktive Regelung von Mild-Hybridantrieben mit elektrischer Zusatzaufladung

• Nonlinear model predictive control of mild hybrid powertrains with electric supercharging

Griefnow, Philip; Andert, Jakob Lukas (Thesis advisor); Abel, Dirk (Thesis advisor)

Aachen : RWTH Aachen University (2021)
Dissertation / PhD Thesis

Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2021

Abstract

In the context of a strongly increasing 48V electrification this thesis takes up the special challenges of the powertrain management of 48V mild hybrid powertrains with electric supercharging and presents a model predictive control concept, which is able to improve the response behaviour and the fuel consumption compared to state-of-the-art heuristic approaches. 48V mild hybrid powertrains with an electrified air path are characterized by a strong interaction between the powertrain and the electrical system. This has significant impact on the degrees of freedom and the complexity of powertrain management. In addition, increasing 48V electrification in the various vehicle domains as well as limited electrical energy and power are further reasons for the importance of an intelligent energy and power management, which makes the best possible use of the limited resources of cost efficiently designed 48V systems. The model predictive powertrain management developed in this work enables an optimization-based control of the belt starter generator as well as the electrified air path via the actuators of the throttle valve, the waste gate and the electric supercharger. It is based on a nonlinear model predictive control (NMPC), which optimizes the drive torque and energy consumption taking into account the battery state of charge. The focus of the work is the conception, development and simulative investigation of the optimization-based control concept. The investigations concentrate on the one hand on the analysis of the controller behaviour in exemplary driving situations and on the other hand on the evaluation of the response behaviour and fuel consumption in dynamic driving cycle simulations. The implemented NMPC is based on a nonlinear differential algebraic equation system to describe the system dynamics. The continuous time optimal control problem is discretized through multiple shooting and solved by sequential quadratic programming (SQP) with a generalized Gauss-Newton method. The implementation is done via the MATLAB-based toolkit ACADO (Automatic Control And Dynamic Optimization). With a discretization time of 40 ms and a prediction horizon of 720 ms the NMPC can be implemented in real-time on the PC in combination with a limitation of the SQP iterations. The controller is able to robustly control the powertrain’s degrees of freedom over the entire operating range, even under the influence of high disturbances. Furthermore, it enables a targeted and fuel saving use of the 48V system without negatively influencing the driving dynamics. Under ideal conditions, the presented NMPC can achieve fuel savings of up to 10.3% in a real world driving cycle compared to a state-of-the-art rule based powertrain management. In principle, the potential increases with increasing knowledge about the future driving demand and decreasing driver influence. The weighting of the NMPC allows a calibration between efficient and dynamic driving behaviour. Overall, the NMPC powertrain management represents a promising method of effectively controlling hybrid powertrains with an electrified air path with regard to driving dynamics and fuel consumption. Since, in contrast to heuristic methods, it does not require application and situation specific sets of rules, the approach can be transferred to similar powertrain concepts and is thus suitable for reducing the development, adaptation and calibration effort in the future.