These slides were presented at the NAFEMS World Congress 2025, held in Salzburg, Austria from May 19–22, 2025.

Abstract

Fast charging of electric vehicles (EVs) presents a critical challenge for the evolution of sustainable electromobility. High charging currents introduce thermal management issues, which can lead to efficiency losses, battery degradation, and increased safety risks. Traditional methods to address these issues rely on highly detailed simulations to analyze and optimize thermal behavior during charging. However, these simulations are computationally expensive and time-consuming, creating a bottleneck in efforts to improve charging efficiency. To address these challenges, integrating artificial intelligence (AI) into model reduction techniques has emerged as a promising solution, enabling a faster and more efficient approach to thermal analysis. Model reduction simplifies complex models into streamlined versions that retain essential information, facilitating faster computation without compromising accuracy. Historically, model reduction has been accomplished using mathematical and engineering methods such as proper orthogonal decomposition (POD) and balanced truncation, which reduce model complexity by identifying and retaining the most significant components. However, these traditional approaches require extensive expertise, are computationally costly, and are often limited when applied to highly nonlinear systems, such as those encountered in EV thermal management. The integration of AI, particularly through machine learning (ML), brings an innovative edge to model reduction, providing an adaptable and data-driven approach. Unlike conventional methods that rely heavily on predefined mathematical frameworks, AI-driven model reduction learns patterns from data, making it especially useful for managing nonlinearities. By training on simulation outputs, ML algorithms identify underlying relationships in the data, capturing complex interactions in the charging process. Neural networks, for instance, excel at autonomously extracting relevant features, eliminating the need for labor-intensive manual feature selection and making the reduction process more efficient. One of the key advantages of AI-based model reduction lies in its ability to optimize complex systems dynamically. For instance, when applied to fast-charging EVs, AI-driven models can predict thermal behavior in real-time and adapt the charging process to manage temperature rise effectively. This reduces the need for extensive simulation recalculations while preserving accuracy. AI algorithms can rapidly generate reduced models that streamline computation, enabling real-time control and decision-making during the charging process, which ultimately improves charging efficiency and battery life. Another notable benefit of AI-driven model reduction is its ability to incorporate innovative thermal management solutions, such as the use of phase change materials (PCMs). PCMs offer effective thermal regulation by absorbing and releasing heat during phase transitions, making them ideal for managing the high temperatures generated during fast charging. By integrating PCMs into the charging cable and optimizing their thermal properties through AI-reduced models, significant thermal losses can be mitigated, resulting in improved temperature control. This reduces the need for external cooling mechanisms, minimizes energy loss, and enhances charging speed without compromising safety or efficiency. This presentation will delve into the practical applications and benefits of AI-assisted model reduction for optimizing EV charging. By reducing model complexity while maintaining predictive accuracy, AI-driven approaches bridge the gap between theoretical simulation and real-world implementation. The use of AI in model reduction transforms traditional simulation workflows, offering a scalable, data-driven solution that enhances the sustainability and efficiency of EV fast-charging technology. This innovative synergy of AI and model reduction presents a valuable pathway toward achieving a more sustainable, efficient, and scalable future for electromobility.