Comparison of EPED-NN Predictions to the Large DIII-D Experimental Pedestal Database

We compare pedestal structure predictions of the EPED model to experimental data from DIII-D H-mode discharges. EPED-NN (neural net) and TokSearch tools are used to accelerate the process of EPED runs and experimental data processing, which enables novel comparisons across thousands of discharges. We establish a shape-independent edge localized mode (ELM) detection method for determining the pre-ELM EPED-NN input parameters and gathering pedestal profile measurements from Thomson scattering and charge exchange recombination spectroscopy (CER). We find EPED-NN predictions serve as a reasonable upper-bound for pedestal height and the simple scaling for pedestal width based on the normalized poloidal pressure at the top of the pedestal agrees well with experimental data. For the pedestal height, we find similar dependencies in plasma triangularity and pedestal electron density between measured and predicted results.