Abstract

Predictive optimization of core plasmas is fundamentally constrained by the cost of repeatedly solving nonlinear transport equations under varying assumptions and constraints. This talk presents PORTALS, a surrogate-assisted workflow that leverages fast statistical models to emulate high-fidelity transport simulations and guide the iterative solution of the transport problem. By enabling efficient sampling of parameter space and targeted refinement of model evaluations, the approach accelerates convergence while maintaining consistency with first-principles physics. This framework provides a practical pathway toward rapid scenario optimization and design.