Model Predictive Iterative Learning Control Design for Battery Optimal Electro-Thermal Management Under Daily-Variant State-of-Charge Patterns

Abstract

This research studies the control of electro-thermal dynamics of cylindrical Li-ion battery packs in electric vehicles (EVs). These dynamics are coupled by the charging-discharging current which generates the Joule heating that directly affects to the operation of battery cells. Hence, to guarantee the battery cells’ temperatures in a desired range for their best operation, a model predictive iterative learning control (MPILC) design is proposed, which composes of an iterative learning controller (ILC) and a model predictive controller (MPC). The former controller with iteration-varying learning gains helps better track slightly variant daily state-of-charge (SoC) patterns of the battery pack. A constant upper bound is derived for the tracking error norm, based on which the iteration-varying learning gains can be designed to make the tracking error converge to zero. The latter controller employs the result of the former as a predicted disturbance to design the cooling-heating temperature input for the battery pack by minimizing its consumed energy while driving the battery cells’ temperatures to a desired range. Simulations are then carried out to illustrate the effectiveness of the proposed MPILC design in tracking daily-variant SoC profiles while guaranteeing battery cells’ temperatures in expected intervals.

Keywords

• Li-ion battery electro-thermal model
• iterative learning control
• model predictive control
• iteration-varying references