Developing machine learning facilitated pedestal models

This conference manuscript provides an overview of recent activities in a project developing machine learning (ML) facilitated pedestal models. The project is divided to three branches, consisting of surrogate modelling techniques for pedestal magnetohydrodynamics, development of reduced pedestal transport models with ML methods, as well as data-driven methods to learn corrections for the remaining gap between numerical predictions and experimental observations. A proof-of-principle model for accelerating pedestal MHD stability evaluations has been recently published, and the next step activities to go beyond this proof-of-principle are detailed. First proof-of-principle models are emerging from the part of the project developing surrogate models for local, linear pedestal gyrokinetic evaluations based on GENE simulations for JET and MASTU. The data-driven models are proceeding from purely observations-based models to models that combine both physics models and experimental observations for a combined representation.