Optimization of grading rings for 1000 kV dry-type air-core shunt reactor based on hybrid RBFNN–Kriging surrogate model

The 1000 kV dry-type air-core shunt reactor, as one of the key components of ultra-high voltage AC transmission systems, is still in the preliminary stage of research regarding its grading ring configuration. Aiming to reduce the maximum electric field strength of the reactor, this paper proposes a hybrid surrogate model that combines Radial Basis Function Neural Network (RBFNN) and the Kriging model to optimize the configuration of grading rings. First, the sparrow search algorithm is used to optimize the hyperparameters of the RBFNN. To enhance the surrogate model’s fitting accuracy further, the Kriging model is utilized to adjust the deviations produced by the RBFNN. The RBFNN–Kriging hybrid model demonstrates superior fitting accuracy compared to the standalone Kriging model, exhibiting enhanced statistical consistency between simulated and predicted values of maximum electric field strength. This improvement is quantitatively validated through reduced absolute errors and lower root-mean-square error metrics. After the optimization, the maximum electric field strength is substantially reduced, proving the effectiveness of the hybrid surrogate model. Consequently, the proposed method not only ensures the hybrid model’s fitting accuracy but also significantly reduces the reactor’s electric field strength, aligning with engineering application standards and providing a valuable reference for configuring grading rings for the 1000 kV dry-type air-core shunt reactor.