The increasing complexity of hybrid and electric vehicles presents significant challenges for efficient thermal management. This work introduces a digital prototype approach for modeling a hybrid vehicle thermal management system. The method combines low-dimensional, physics-based models with machine learning techniques to achieve faster-than-real-time simulations over complete driving cycles while maintaining high accuracy in predicting component and fluid temperatures as well as pressure losses.

At the core of the approach is a self-developed model library, where each component includes a parameter mask that enables rapid configuration and adaptation to different system layouts. The modular structure allows fast assembly and parametrization of new architectures, significantly reducing development time and measurement requirements. All models have been validated against test bench and vehicle data, ensuring high fidelity and robustness in real-world conditions.

The framework employs a model-based system architecture with standardized interfaces, allowing flexible integration of control strategies and seamless incorporation of data-driven models for enhanced prediction and optimization. Providing FMU-compatible components supports modular and scalable configurations and facilitates integration with other vehicle subsystems or digital twins. By combining physics-based and data-driven modeling, the proposed methodology reduces testing effort, accelerates development, and supports the optimization of energy efficiency and control strategies for hybrid and electrified vehicles.