FEM-Circuit co-simulation of superconducting synchronous wind generators connected to a DC network using the homogenized J–A formulation of the Maxwell equations

Abstract

High-temperature superconductors (HTS) are greatly appealing for the development of high efficient, and high energy density power devices. They are particularly relevant for applications requiring light and compact machines such as wind power generation. In this context, to ensure the proper design of the superconducting machines and their reliable operation in power systems, it is then important to develop models that can accurately include their physics but also can describe properly their interaction with the system. Thus, one approach is the co-simulation. The goal of the present work is to put to use this numerical technique when superconducting components are involved. Here, the case study is a 15 MW hybrid superconducting synchronous generator (HTS rotor and conventional stator) coupled to a direct current (DC) network via a rectifier and its associated filter. The case study related to wind power application allows grasping the technical issues when employing co-simulation dealing with HTS machines. The FEM of the generator is done in COMSOL Multiphysics, which interacts with the circuit simulator Simulink through the built-in Functional Mock-up Unit (FMU). For the present study, a new version of the latest J-A formulation combined with homogenization technique is introduced allowing an even faster computation time compared to the T-A formulation. Distributed variables and global variables such as current density, magnetic flux density, and local losses for the former and voltage, current, electromagnetic torque, and power quality for the latter are estimated and compared for both formulations. The idea is to find the best-suited combination FEM-circuit under criteria of computational speed, accuracy, and numerical stability. It is then shown that all formulations generate an error of less than 5% on the machine parameters; and the J-A formulation with first order elements stands out with a significant 4-fold reduction in computational costs.