Augmented Lagrangian methods produce cutting-edge magnetic coils for stellarator fusion reactors

Abstract

Finding feasible coils for stellarator fusion devices is a critical challenge of realizing this concept for future power plants. Current design efforts struggle to navigate the highly nonconvex optimization landscape, spend considerable resources scanning the parameter space, and may produce suboptimal coils. In this work, we present an augmented Lagrangian approach to tackle the ill-posed problem of coil optimization. We illustrate its effectiveness and versatility by generating coils for five stellarators with very different symmetries and magnetic field shaping. In all cases, we find Pareto-optimal coil solutions that in various ways outperform published coil sets.