Combing physics-based and data-driven predictions for quantitatively accurate models that extrapolate well; with application to DIII-D, AUG, and ITER tokamak fusion reactors

Abstract

Methodologies for combining the accuracy of data-driven models with extrapolability of physics-based models are described and tested, for the task of building transport models of tokamak fusion reactors that extrapolate well to new operational regimes. Information from multiple physics simulations (the ASTRA transport code with gyro-Bohm and TGLF estimates for turbulent diffusion) as well as multiple distinct experiments (DIII-D and AUG tokamaks) are considered. Applications of the methodology to the task of commissioning and controlling a new reactor such as ITER are discussed.

Abstract: arXiv descriptive metadata (CC0). Thank you to arXiv for use of its open access interoperability. This service was not reviewed or approved by, nor does it necessarily express or reflect the policies or opinions of, arXiv.