Electric Field Simulation and Surrogate-Based Optimization Design for the Insulation Structure of Winding End in 400 kV VSC-HVDC Converter Transformers

Insulation coordination at the transformer terminals constitutes a critical step in the optimized design of converters. A two-dimensional axisymmetric parametric model with four-level refinement was developed to meet the requirements of multi-physics simulations, including electromagnetic and thermal fields. Based on a second-level simplified model, gradient electric potentials were applied as boundary conditions for both the valve-side and grid-side windings to simulate the electric field distribution. To optimize the field distribution at the winding ends, Monte Carlo sampling was employed to generate training data, and a computationally efficient surrogate model was constructed to replace time-consuming finite element simulations. The error between the surrogate model and the full simulation results remains below 1%. After optimization using the surrogate model, the maximum surface electric field strengths at the valve-side and grid-side ends were reduced 30% and 27% respectively.