Optimal Power Split for Dual-Motor Powertrain of Electric Vehicles With Bilevel MPC

The power split strategy (PSS) is essential for optimizing the efficiency of dual-motor powertrains (DMPs) in electric vehicles (EVs). This article addresses the online power split optimization problem associated with a DMP based on a Simpson planetary gearset. A computationally efficient PSS is proposed, utilizing model predictive control (MPC) to enhance mode selection and power distribution. To improve the accuracy of vehicle velocity and acceleration predictions, the study constructs an acceleration prediction map using a multistep multivariable Markov chain (MC). The diverse modes and forms of power split in various dual-motor drive configurations create considerable challenges in optimizing the control sequence. Therefore, a bilevel algorithm is introduced to significantly enhance computational efficiency, where the lower level calculates power distribution across different modes using various methods. Compared to a rule-based control strategy, the bilevel MPC reduces energy consumption by 5.82% and decreases brake operation frequency by 9.46%, while maintaining motor speed variations within a reasonable range. Although the bilevel MPC exhibits a slight decrease in performance relative to DP-MPC, it achieves a substantial 46.1% reduction in computation time.