Cycle-Consistent and Uncertainty-Aware Neural Surrogates for Tokamak Edge Plasmas

The boundary and divertor plasma govern how a tokamak exhausts power and particles, setting heat fluxes, target conditions, and the onset of detachment. Traditional edge simulations like SOLPS-ITER are too slow for parameter scans, optimization, or real-time control. This work introduces a cycle-consistent neural surrogate: a conditional U-Net forward model mapping five control parameters (including core fueling rate) to 2D plasma-state fields on the SOLPS-ITER mesh; an inverse method built on the frozen forward network uses optimization and cycle-consistency regularization for self-supervised parameter recovery. An ensemble of MLPs predicts electron temperature and density profiles with uncertainty estimates. The forward model achieves normalized RMSE below 2.6% and Pearson correlation above 0.95. Cycle-consistency raises the average cyclical R² from 0.59 to 0.99. The inverse method with a k-d tree warm start achieves database completion rate above 95%, versus ~30% outright failures when cold-started. With about 4×10⁶ parameters, the model produces full 2D predictions in milliseconds, 5-6 orders of magnitude faster than SOLPS-ITER, enabling real-time control, parameter scans, uncertainty analysis, and digital twins.