A Jaya-driven quasi-predictive optimization strategy with adaptive window scheduling for EV thermal management

Grappling with the trade-off between long-range planning, driven by the battery’s sluggish thermal response, and low computational cost remains a core challenge for model-based optimization methods in electric vehicle (EV) thermal management. This study proposes a lightweight quasi-predictive optimization strategy driven by the non-model-based Jaya algorithm, designed to emulate predictive control while ensuring real-time responsiveness. A dual-mode coordinated cooling model, integrating air and liquid circuits, is developed to capture multi-source thermal interactions between the battery and cabin. To adapt to real-time thermal dynamics, an entropy-guided adaptive window scheduling mechanism with dynamic objective weighting is proposed, enabling balanced trade-offs among window length, control accuracy, and computational efficiency. Simulation results indicate that the proposed strategy reduces energy consumption by 11.48% while maintaining a millisecond-level computation time (0.4 ms per step) over conventional fixed-window approaches. These findings confirm that the proposed strategy offers a computationally efficient, adaptive, and real-time capable solution for scalable thermal management in connected EV applications, supporting future deployment in embedded control systems.