Framework for Automated Data-Driven Model Adaption for the Application in Industrial Energy Systems

Increasing flexibility and efficiency of energy-intensive industrial processes is generally seen as a big lever towards a decarbonized energy system of the future. However, to leverage these potentials, the accurate prediction of unit behavior is essential to be able to close the gap between supply and demand. Not only pose nonlinear relations a serious challenge in thermal systems engineering and optimization but real-world unit behavior furthermore changes during operation due to wear, fouling and other effects. In the present work, a novel framework for automated data-driven model adaption is presented which is capable of automating fast and accurate predictions of current system behavior. The framework is based on open protocol bidirectional live communication and mechanistic grey box modeling. While especially thermal energy storage is considered a solution to increase flexibility, it is very challenging for operation optimization. A packed bed thermal energy storage operated under severe conditions leading to continuous fouling acts as proof of concept of the proposed framework. The obtained results indicate major improvement for storage output prediction with the novel framework compared to a conventional approach without readjustment. Furthermore, the presented framework is perfectly suitable and an essential foundation for live condition monitoring, fault prediction, predictive maintenance, and operation optimization.