Robust Thermal Management of Electric Vehicles Using Model Predictive Control With Adaptive Optimization Horizon and Location-Dependent Constraint Handling Strategies

The thermal management system (TMS) in electric vehicles (EVs) consumes a considerable amount of energy for maintaining the battery and cabin temperatures within the desired range. This energy consumption can significantly impact the vehicle's driving range. In this article, a model predictive control (MPC) is applied to minimize the energy consumption of the TMS while enforcing power and thermal constraints. The MPC-based thermal management strategy relies on a control-oriented model that captures the dynamics of the powertrain and thermal subsystems of an EV, as well as the coupling between these subsystems at different timescales. The relatively slow dynamics of the thermal systems call for a long prediction horizon to achieve the best performance. However, large uncertainties associated with speed prediction and preview information significantly impact the performance and robustness. In this study, an extensive sensitivity analysis is conducted to: 1) determine the key traffic and speed features over a long prediction horizon with a significant influence on the EV optimal performance and 2) study the impact of uncertainties associated with predicting these key traffic and speed features on EV performance in terms of energy efficiency and constraint enforcement. The MPC-based thermal management strategy is evaluated using real-world traffic data. To improve the robustness of the algorithm in the presence of uncertainties, a location-dependent constraint is proposed and integrated into the MPC-based thermal management strategy. The simulation results demonstrate that the location-dependent constraint enhances the capacity to enforce the battery temperature constraint, resulting in improved algorithmic robustness against uncertainties in preview information.