Reconstructions of electron-temperature profiles from EUROfusion Pedestal Database using turbulence models and machine learning
This study uses plasma-profile data from the EUROfusion pedestal database, focusing on the electron-temperature and electron-density profiles in the edge region of H-mode ELMy JET ITER-Like-Wall (ILW) pulses. We make systematic predictions of the electron-temperature pedestal, using the density profiles and engineering parameters of the pulses as inputs. We first present a machine-learning algorithm that, given more inputs than theory-based modelling and 80\% of the database as training data, can reconstruct the remaining 20\% of temperature profiles within 20\% of the experimental values, including accurate estimates of the pedestal width and location. The most important engineering parameters for these predictions are magnetic field strength, particle fuelling rate, plasma current, and strike-point configuration. This confirms the potential of accurate pedestal prediction using large databases. Next, we take a simple theoretical approach assuming a local power-law relationship between the gradients of density ($R/L_{n_e}$) and temperature ($R/L_{T_e}$): $R/L_{T_e}=A\left(R/L_{n_e}\right)^α$ with $α\approx 0.4$ fits well in the steep-gradient region. When $A$ and $α$ are fit independently for each pedestal, a one-to-one correlation emerges, also valid for JET-C data. For $α= 1$, $A \equiv η_e$, a known control parameter for turbulence in slab-ETG theory. Measured values of $η_e$ in the steep-gradient region lie well above the slab-ETG stability threshold, suggesting a nonlinear threshold shift or a supercritical turbulent state. Finally, we test heat-flux scalings motivated by gyrokinetic simulations, and we provide best-fit parameters for reconstructing JET-ILW pedestals. These models require additional experimental inputs to reach the accuracy of the machine-learning reconstructions.