Predictive Transport Modeling of KSTAR Advanced Scenarios Using COTSIM

Integrated transport modeling of high-performance KSTAR discharges is performed using COTSIM, a control-oriented framework that combines empirical, physics-based, and neural network surrogate models for transport and auxiliary heating. The magnetic equilibrium is computed by using a fixed-boundary solver, while safety factor (q) and electron temperature (T_e) profiles are evolved through coupled 1D magnetic diffusion and electron heat transport equations. Anomalous transport is modeled using empirical, Coppi-Tang, Bohm/Gyro-Bohm, and machine-learning-based MMMnet models. Neutral Beam Injection (NBI) and Electron Cyclotron (EC) heating and current drive are modeled using empirical scaling laws or fast surrogates (NUBEAMnet, TORAYnet) trained on high-fidelity simulations. Benchmarking exercises against experimental measurements and TRANSP outputs assess the predictive performance of different model combinations. The simulated evolution of key plasma profiles, including T_e and q, demonstrates good agreement with these benchmarks and validates the accuracy of the COTSIM platform across diverse modeling choices. This modular framework supports efficient scenario analysis and advances efforts toward model-based control and scenario optimization on KSTAR.