Iterative Learning Control Design for Iteration-Varying State-of-Charge Profiles of Electric Vehicle Batteries

In this research, a battery control method is proposed to handle the repetitive but nonidentical daily state-of-charge (SoC) profiles of electric vehicle (EV) batteries. The proposed method employs an iterative learning control (ILC) framework having a quadratic performance index with iteration-varying weighting matrices. This results in iteration-varying ILC control gains to better cope with iteration-varying SoC profiles. Moreover, input constraints representing the limits on the ranges of the charge and discharge currents are considered, leading to an iteration-varying constrained convex optimization problem. This optimization problem is solved to obtain the ILC control input update via resolving its Lagrange dual problem. Next, a data-driven method based on the dynamic mode decomposition (DMD) approach is proposed to predict the SoC profile in the next weekday based on the SoC profiles in the current and previous weekdays. The predicted SoC profile is then served as the reference for the ILC tracking controller. Finally, the proposed methods are verified through extensive numerical simulations for a synthetic case and for a realistic, benchmark driving pattern. In the simulations, different ways of selecting the iteration-varying weighting matrices are introduced and their control performances are compared. It is also shown that the proposed ILC control design outperforms conventional P-type and adaptive ILC controllers as well as the classical proportional-integral-derivative controller on the tracking of the SoC profile based on the considered realistic driving pattern.