Model predictive energy management for induction motor drives and all-wheel-drive battery electric vehicles : a flatness based approach

Abstract

Since emission and fuel economy standards have risen significantly, car manufacturers find themselves forced to invest in new battery and related technologies. Research on methods and technologies that improve the efficiency of both the battery and the electric powertrain, pose the greatest challenges of this technological transition. Therefore, software solutions for energy management and motion control as well as economic driving strategies are becoming more and more the focus of future developments. Software solutions allowing for potential economic savings are particularly appealing, since these do not require any structural or mechanical design changes. With the main objective of increasing economic savings, this present work investigates analytical models of the electric powertrain for a battery electric vehicle with two drive modules on the front and rear axle. The modeling approach focuses on loss processes associated with the energy conversion of the voltage source inverters and induction motors. Due to a wide spectrum of involved time constants in the range of seconds to a few milliseconds, efficiency analyses of electric vehicles rarely follow model-based approaches and instead rely on characteristic loss maps, which neglect dynamic effects and physical limitations, for example, those resulting from the limited battery voltage. This widespread spectrum comes from both the longitudinal dynamics of the vehicle and the voltage and current harmonics that result from high frequency switchings of the inverters’ semi-conductors. New dynamical models are thus proposed that can be efficiently integrated into vehicle simulations and also be implemented online on embedded systems, such as the motor control unit of the investigated vehicle. In doing so, an average value model of the voltage source inverter is derived, based on a double Fourier integral analysis of the semi-conductor switching signals. Furthermore, a widely used model of the induction motor, applied for motor analysis and control design, is reformulated into an equivalent differential flat system based on the definition of a new flat output. Both component models are integrated into a vehicle simulation of a Mercedes Benz EQC prototype and are thoroughly validated through extensive simulative studies and experimental test series. With the help of the newly introduced models and with the assistance of modern vehicle sensor systems, control strategies of the electric powertrain are investigated that aim for the most energy efficient operation. In a first step, decentralized optimal control approaches are proposed that improve the efficiency of the electric drive module, not only during stationary operation, but also during transient torque conditions. This improvement is achieved by an appropriate field oriented control method. In a second step, optimization-based torque allocation strategies are investigated and evaluated experimentally. Finally, a centralized predictive control approach is presented that exploits all operational degrees of freedom, which are the variable torque allocation, the front and rear axle magnetic flux, and the adjustment of the vehicle speed according to topographical and traffic dependent conditions. Significant economic savings are demonstrated for the decentralized control methods as well as for the centralized control approach. The high level of accuracy and performance that is achieved by the proposed model-based framework and predictive operational strategies are only made possible by exploiting the positive structural properties of the newly introduced differential flat system representation of the induction motor.