The stellarator is a promising candidate for a commercial fusion reactor because its confining magnetic field is generated entirely by external coils, enabling inherently steady-state and macroscopically stable operation. Since the fusion power density scales with the square of the plasma pressure and a high (arguably, the fourth) power of the magnetic field, next-generation stellarators require superconducting magnet systems capable of producing fields beyond those achieved in present-day experiments. Achieving such fields requires geometrically complex coils with non-planar centerlines, continuously varying curvature, and winding packs whose cross-sections rotate along the conductor path. Modern plasma optimisation codes typically represent these coils only as filamentary centerlines, without information on the conductor technology, winding-pack geometry, or structural design. Furthermore, the magnet design methodologies reported in the literature have generally been developed for individual stellarator concepts and are therefore difficult to transfer to other configurations. This lack of a general design approach is becoming an increasingly severe limitation as reactor-scale stellarator concepts continue to diversify in coil topology and superconducting technology.

This thesis develops a comprehensive approach for the pre-dimensioning of stellarator magnet systems that translates into a parametric workflow starting from the filamentary output of plasma optimisation and delivering a complete preliminary magnet design comprising the winding pack, conductor operating point, and structural casing. The workflow evaluates the proposed magnet design from both electromagnetic and mechanical point of view using a three-dimensional model in Ansys Mechanical APDL, and also performs a one-dimensional thermal-hydraulic analysis in THEA. The process is iterative and requires only a small number of manual iterations to converge to a solution that satisfies the requirements of all the physics models considered.

The workflow is first applied to HELIAS 5-B, for which an independent conductor and coil design is available, enabling verification against established results.

Then, it is used to develop the first magnet system for CIEMAT-QI4X, a reactor-scale quasi-isodynamic stellarator configuration. Two superconducting technologies are considered for the same configuration and compared through electromagnetic, mechanical, and thermal-hydraulic analyses: a low-temperature design based on a Nb3Sn cable-in-conduit conductor operating at 4.5K, and a high-temperature design based on a VIPER-like REBCO conductor operating at 20K. Both designs prove viable at this stage but differ in the structural support they require and the conductor limits they operate against, with the REBCO variant emerging as the more favourable of the two.

Finally, a preliminary magnet design is developed for the Polaris tabletop stellarator based on a non-insulated REBCO winding, including the electromagnetic and thermal design as well as an assessment of screening-current effects for different contact resistances.

The results obtained in the thesis demonstrate that a single parametric workflow can be applied across different stellarator configurations and superconducting technologies without reformulation, while showing that mechanical integrity and conductor strain, rather than electromagnetic performance, drive the design of reactor-scale stellarator magnets.