Embedding Dynamic Microkinetic Modeling Information Into Reduced‐Order Models Using Gaussian Processes

ABSTRACT In this work, a novel generalizable framework is proposed for obtaining dynamic discrepancy reduced‐order models (DD‐ROMs) that balance the differences between high‐fidelity models (HFMs) and reduced‐order models (ROMs) using Gaussian Processes (GPs). The proposed framework encompasses fundamental criteria for addressing missing underlying physics and is the first‐of‐its‐kind to offer a comprehensive insight guided by sensitivity and correlation analyses into where the discrepancy terms must be incorporated. The proposed framework is employed to correct dynamic mismatches between a reduced‐order model and a high‐fidelity microkinetic model of the steam methane reforming (SMR) reactions. The validation results demonstrate that with the discrepancy function added to the equilibrium constant, the DD‐ROM is capable of mimicking the dynamic trajectories of the microkinetic model with high accuracy, exhibiting an of 97.86% and an of 0.123, while obtaining a significant computational gain, being 104 times faster per model execution than integrating the HFM model.