1. D. McCune 1 PTRANSP Predictive Upgrades for TRANSP

2. D. McCune 2 US Predictive Modeling Effort R. Budny, S. Jardin, C. Kessel, L. P. Ku, D. McCune (PPPL). H. St. John (GA). D. P. Grote, L. Lodestro, L. D. Pearlstein, T. D. Rognlien (LLNL). G. Bateman, F. Halpern, A. Kritz (Lehigh). J. Carlsson (Tech-X).

3. D. McCune 3 PTRANSP Plan Leverage TRANSP: –Well validated source models (NBI, alphas, ICRF, LH, ECH/ECCD). –Strong connection to experimental data. –Fusion Grid production facility. Add predictive capabilities to TRANSP: –Robust transport equation solver. –Free boundary equilibrium. –Connection to edge model. Reuse existing software to extent possible.

4. D. McCune 4 Design Principles - 1 Reuse TRANSP and Fusion Simulation Project (FSP) software (to minimize costs). Two driver configurations: –Free boundary: (TRANSP computes sources; analyzes free boundary code results). –Prescribed boundary: traditional TRANSP with: New transport solvers (FSP Solver, GCNM-P). New MHD equilibrium solvers (FSP, TEQ).

5. D. McCune 5 Design Principles - 2 Modular design: interchangeability of critical parts (create/use NTCC modules): –Transport solvers. –MHD equilibrium solvers. –Sources. Leverage TRANSP archives: –Access to experimental data for validation. –NTCC module provided for data access.

6. D. McCune 6 SolverEquilibriumSources GCNMP FSP-Sol Controller TEQ FSP-Equ ESC TRANSP Plasma State TRDATBUF (access to experimental data) FSP-Src XPLASMA (FSP upgrade in progress) Edge AnalysisStability Analysis Postprocessing (initially) Sawtooth Edge Pedestal Bootstrap Curr NCLASS Hirsh-Sig. Lehigh PPPL Porcelli-L Porcelli-P UEDGE DEGAS-2 DCON PEST-2 PTRANSP Schematic TRANSP-based controller FSP-based controller

7. D. McCune 7 Transport Solver Dilemma Current predictive transport models (e.g. GLF-23) are very stiff. Standard numerical integration methods suffer severe oscillations and instability. Attempts to “smooth” GLF-23 directly significantly changes prediction results. Therefore: serious solver upgrade effort. –GCNM-P (General Atomics) & FSP (PPPL).

8. D. McCune 8 Transport Solvers GCNM – Globally Convergent Newton Method – ONETWO Solver (St. John, GA). –Very general stiff PDE integrator. –Use of Jacobian, O(n**2) execution cost. FSP Solver (Jardin & Ku, PPPL). –“Local” Newton method– forward implicit use of dependence of transport on grad(Ti,Te,…). –O(n) but may not be as stable as GCNM.

9. D. McCune 9 The PTRANSP FSP Solver - 1 This has been implemented in the full solver in the FSP: Without linearizationWith linearization ITER simulation Linearization of dependence of GLF-23 fluxes on temperature gradients. Behavior reproducible in simplified single-T analytic transport model. Caveat: DIII-D experimental data validation attempt– not yet fully stable. S. Jardin / L. P. Ku

10. D. McCune 10 The PTRANSP FSP Solver - 2 Convergence Tests: 3 Newton iterations per timestep Double # of zones Reduce timestep by 3 Base case: 1 Newton iteration per timestep S. Jardin / L. P. Ku

11. D. McCune 11 Results for a 500s ITER run: Profiles at 250s Chi Values for entire run Powers vs time ions electrons Chi vs radius at 250s ions electrons S. Jardin / L. P. Ku

12. D. McCune 12 PTRANSP Progress - 1 Predictive Solver Improvement (as shown). –Both FSP solver and GCNM at GA. TRANSP Improvements: –Export of source calculation results. –Accommodation of free boundary equilibrium. –Modification of internal loop structure to allow import of stiff transport solver results. Trdatbuf_lib NTCC module– access to TRANSP input data (experimental data).

13. D. McCune 13 PTRANSP Progress - 2 LLNL’s TEQ free boundary solver module in TRANSP. –NTCC module standards with error handling enhancement. –Time dependent NSTX test results look good. UEDGE/TRANSP coupling: –LLNL design and prototype in place. –Includes TRANSP/UEDGE data exchange schema.

14. D. McCune 14 PTRANSP Progress - 3 NTCC PEDESTAL module– predictive boundary condition option. –Lehigh University team making direct modifications to TRANSP (in progress). –Prototype installation in BALDUR. –Experience with L to H transition dynamics.

15. D. McCune 15 PTRANSP’s Next Step – APS Drive TRANSP with ITER free boundary simulation: –TRANSP provides heating and current drive. –TRANSP uses free boundary simulation predicted temperatures and equilibria. Architecture compatible with density prediction but testing of this capability likely to be postponed. –TRANSP archive produced: Available as input to UEDGE, linear stability solvers, etc.

16. D. McCune 16 Summary The PTRANSP project will provide a community predictive transport code with state-of-the-art capabilities. Like TRANSP itself, it will run as a Fusion Grid production service with world wide access. Control options will be provided for prescribed boundary or free boundary operation. Example of integrated ITER simulation with realistic sources by APS-2006.