Superconducting (SC) electrical machines have higher power density and efficiency than permanent magnet (PM) electrical machines

making them advantageous for the development of offshore wind turbine capacity beyond 20 MW. To solve the problem that the pole pair number of SC excitation cannot be increased due to the size of the modular dewar

this paper proposes a double-stator superconducting magnetic field modulation electrical machine (DSFMM) adopting integrated dewar. To explore how to achieve optimal coupling between the harmonic components generated by the modulators such as the stator teeth and rotor magnetic blocks and the harmonic components caused by the armature winding with different pole-slot combinations in the DSFMM

this paper analyzes the effects of nine sensitive parameters such as pole ratio

pole-slot combination

pole arc coefficient

and rotor magnetic block ratio of the DSFMM on electromagnetic performance

and the mathematical model based on these parameters is established. To solve the issues of slow speed and high resource consumption of the traditional optimization methods for large capacity DSFMM

this paper uses the sample library generated by the mathematical model to train the back propagation (BP) neural network and the mapping relationship is obtained between the nine sensitive parameters and performance of electrical machine. Then

the multi-objective optimization design of a 20 MW DSFMM is completed by using the improved non-dominated sorting genetic algorithm Ⅱ(NSGA-Ⅱ).