As the core electrical equipment in the power system

the high-voltage switchgear in substations is affected by the climate of southern China

leading to equipment malfunction or refusal to operate. The intricate mechanism of moisture condensation poses challenges in predicting the critical temperature of condensation

resulting in low modeling accuracy and a lack of condensation samples.To tackle these challenges

a prediction method for the critical condensation temperature of highvoltage switchgear in substation is proposed

based on enhanced relative generative adversarial networks(ERGAN) and Bayesian optimization bidirectional gated recurrent unit(BO-BiGRU)networks.This method aims to improve the prediction accuracy of the condensation critical temperature of high-voltage switchgear.Firstly

to address the scarcity of condensation samples

ERGAN is designed to expand the training samples.This model introduces a relative loss function and a gradient penalty term

and utilizes one-dimensional convolutional layers and spectral normalization layers to reconstruct the generator and discriminator

effectively improving the quality of generated samples and training stability.Secondly

Bayesian optimization is employed to iteratively optimize the hyperparameters of BiGRU model

constructing BO-BiGRU model to predict the critical value of condensation. Finally

real-world data from 35 kV high-voltage switchgear in substation in Fujian are used for simulation verification. Simulation results indicate that

taking the mean absolute error as an example

the prediction performance of the proposed method is 68.95% higher than that of model-driven methods

laying a foundation for the prevention and control of condensation in high-voltage switchgear in substations.