Prediction of Plasma Confinement Indices by Gaussian Process Regression

To optimize the design of a helical fusion reactor by varying the shape of the magnetic coils, several requirements related to the performance of the reactor should be satisfied under various constraints. To address this multi-objective optimization problem, we utilized Gaussian process regression (GPR) for machine learning to develop a surrogate model capable of predicting the dependence of the objective functions on the parameters representing the coil shape. This study demonstrates that the dependence of objective functions, such as plasma volume and the Mercier criterion, on the shape of helical coil windings can be predicted by GPR.