Integrated Thermal Management of Electric Vehicles Based on Model Predictive Control With Approximated Value Function

The thermal management system (TMS) in electric vehicles (EVs), including climate control and battery thermal regulation, consumes more energy than any other auxiliary components. Therefore, optimizing TMS control is crucial for enhancing EV driving range. However, the complexity of the TMS, described by a differential algebraic system, poses challenges for real-time optimal control. This study proposes model predictive control (MPC)-based solutions for integrated TMS operation in EVs. An optimal thermal management problem is formulated using economic nonlinear MPC (NMPC), and its performance is evaluated. To reduce computational load, an approximated value function (VF) is introduced based on the economic NMPC results. A linear-time-varying MPC (LTV-MPC) with the approximated VF is proposed for real-time implementation using quadratic programming, and through simulations it is compared with the baseline NMPC controller and a rule-based (RB) controller. Results reveal that the LTV-MPC with an approximated VF performs similarly to NMPC while offering slightly compromised cooling performance. It also significantly reduces the computational time by a factor of 10 ⁴ compared with NMPC owing to the short prediction horizon enabled by the approximated VF. Furthermore, when compared with the RB controller, the proposed LTV-MPC achieves energy savings in the range of 22.3% to 29.8%.