Lang Lao and Orso Meneghini attended the 3rd IAEA Technical Meeting on Fusion Data Processing, Validation, and Analysis in Vienna, Austria May 27-31, 2019. Lang Lao gave an invited presentation on “Recent Advances in Equilibrium Reconstruction for Fusion Data Analysis and Plasma Control and Prospect for ITER Applications” and Orso Meneghini gave two oral presentations on “Interfacing Physics Codes with ITER IMAS via OMAS” and “Neural-Network Accelerated Modeling of ITER Plasmas Compatible with IMAS”.

A new workflow has been developed for predict-first equilibrium modeling. The motivation here is to predict equilibria prior to analyzing experimental profiles. The workflow includes a cold start tool, Modelprofiles, which constructs all the profiles and power flows needed by transport codes starting from only global quantities. Modelprofiles has been coupled to the VMOMS equilibrium code to obtain an initial approximate geometry, as well as a pencil neutral beam heating code. The workflow also couples the EFIT Grad-Shafranov code with a validation code which computes various local and global figures of merit to assess the accuracy of a given equilibrium prediction. Using an empirical pressure model in Modelprofiles, we obtain a remarkably small RMS error in the predicted q-profile of only 10% for a database of 31 DIII-D H-mode discharges. Similar results are found for DIII-D L-modes. The automated workflow is fast, with typical run times of less than 30 seconds on a local workstation. The ultimate goal of predict-first modeling is to use theory-based transport model predictions for the pressure profile in predictive EFIT calculations. As a concrete application, it was shown that using TGLF and EPED to predict the pressure profile in the workflow results in good agreement with the temperature and q-profiles for a DIII-D H-mode current scan.

The TGLF-EP+Alpha energetic particle transport model has been validated against four DIII-D H-mode discharges: 153071 (qmin=1), 153072 (qmin=2), 161401 (ITER steady-state relevant), and 171322 (super H-mode). Measured neutrons in the four discharges fall below classical predictions to varying degrees indicating anomalous losses. This and other TGLF-EP+Alpha experimental validation studies use the predicted EP diffusion coefficient from TGLF-EP+Alpha in TRANSP/NUBEAM to obtain the neutron rate and pressure profile. Simulated EP pressure profiles and neutron rates compare favorably (within 20% or less) to experiment in all but discharge 161401, the ITER steady-state relevant case. A substantial under-prediction of EP transport in this case (and over-prediction of EP pressure and neutrons) can likely be attributed to an unstable 3/2 tearing mode unaccounted for in the model. Ongoing validation studies have been facilitated by incorporation of standardized input formats into the TGLF-EP and Alpha codes.

Modeling with the M3D-C1 extended-MHD code indicates that the rotation profile could be used to optimize the plasma response in double-null (DN) plasmas in order to achieve ELM suppression. The SEGWAY workflow, which links together EFIT, EPED-NN, and NEO to modify self-consistently the shape, pressure, and current of experimental equilibrium reconstructions, was used to generate a variety of single-null and double-null equilibria for this study. M3D-C1 plasma-response modeling of these SEGWAY equilibria confirmed that the high-field-side (HFS) plasma response is significantly reduced in DN plasmas, as seen in previous modeling and in experiments. This is true regardless of toroidal mode number (n = 1, 2, or 3) or perturbation-coil location (inboard or outboard side). The trend was found to vanish, however, for modeling performed without rotation. Thus, the rotation profile could be optimized to increase the HFS plasma response in DN plasmas since this is expected to facilitate ELM suppression.

An improved GACODE input framework has been developed and is available for testing on a git development branch. The new framework provides additional profiles and species data for the legacy EXPRO interface used by GYRO, CGYRO, NEO, TGYRO, profiles_gen and vgen, but with numerous new features and benefits. The read and write functions have been streamlined, the presence of MPI for parallel applications is automatically managed, specific profiles are now optional, and most importantly, the interface is cross-language (fortran and python). Identical read/write/interface functions are available in both languages, reducing or eliminating the need for code duplication and significantly accelerating the OMFIT-GACODE data exchange speed. On the GACODE side, the framework builds with the GACODE build system, and on the OMFIT side the framework can be installed by a single invocation of python pip.