Rapid Prediction of Transient Stress Field in Transformer Windings Based on Extended Dynamic Mode Decomposition

ABSTRACT

To accelerate transformer winding dynamic stress field analysis and reduce calculation errors of existing fast methods, this paper proposes a rapid prediction method based on extended dynamic mode decomposition (EDMD). First, a transient structural dynamic finite element model incorporating the disc turn‐by‐turn stiffness effect was established. Dictionary functions were then constructed based on the nonlinear ‘electromagnetic‐structural’ physical characteristics under short‐circuit conditions to map raw data into a high‐dimensional space. Dynamic modes and time coefficients were extracted from historical simulation data to rapidly reconstruct the full‐field transient winding stress. Multiple short‐circuit tests on a 10 kV fibre‐optic integrated transformer validated the method. Results show that the proposed method significantly outperforms standard dynamic mode decomposition (DMD) in accuracy. The maximum and relative errors against commercial software are within 13.5 MPa and 4.6%, respectively, with a 7.30% relative error against test measurements. Crucially, the computation takes only 24.69 s, which is 3.4 times faster than commercial software. This study resolves the pronounced calculation errors of DMD in strongly nonlinear predictions, achieving fast and accurate calculation of winding dynamic stress fields and providing critical technical support for the digital transformation of power equipment.