Analytical Model and Topology Optimization of Doubly-Fed Induction Generator

Abstract

As the core component of energy conversion for large wind turbines, the output performance of doubly-fed induction generators (DFIGs) plays a decisive role in the power quality of wind turbines. To realize the fast and accurate design optimization of DFIGs, this paper proposes a novel hybrid-driven surrogate-assisted optimization method. It firstly establishes an accurate subdomain model of DFIGs to analytically predict performance indexes. Furthermore, taking the inexpensive analytical dataset produced by the subdomain model as the source domain and the expensive finite element analysis dataset as the target domain, a high-precision surrogate model is trained in a transfer learning way and used for the subsequent multi-objective optimization process. Based on this model, taking the total harmonic distortion of electromotive force, cogging torque, and iron loss as objectives, and the slot and inner/outer diameters as parameters for optimizing the topology, achieve a rapid and accurate electromagnetic design for DFIGs. Finally, experiments are carried out on a 3MW DFIG to validate the effectiveness of the proposed method.

Keywords

• —doubly-fed induction generators
• accurate subdomain model
• surrogate-assisted
• transfer learning.