Study of the radial requirements for accurate predictions of tokamak profiles and fusion performance

Temperature and electron density profile predictions for tokamak fusion experiments involve simulating profile gradients at a set number of points and integrating to predict kinetic profiles (ne, Te, Ti). Using first-principle simulations to calculate points in this gradient interpolation is very computationally intensive, motivating the use of as few points as possible. 20 years of discharges from Alcator C Mod were used to determine the optimal number of points and locations needed to accurately predict profiles and fusion power. Each discharge was fit with a gaussian process based fitting routine, and the normalized logarithmic gradients were calculated. In the normalized minor radius range of 0.2 to 0.9, for a 3-point gradient interpolation, the best points were at r/a=0.5,0.825,0.9; for a 4-point gradient interpolation, r/a=0.45,0.7,0.85,0.9; and for a 5-point gradient interpolation, r/a=0.35,0.55,0.75,0.875,0.9. With these points, a 3-point interpolation created a mean error of 5% in the fusion power with a 25.5% standard deviation; 4 points had .7% and 10%; while 5 points had .05% and 6.8%. These results suggest that a reduced set of local turbulence simulations for profile prediction in the core of inductive, on-axis heating scenarios can provide sufficiently accurate results.