The Error Prediction Method for Voltage Transformers Based on the EGO-LSTM Model

The operation management of voltage transformers tends to focus on long-term monitoring, lacking real-time online condition assessment methods. Therefore, this paper utilizes the measurement data from the transformer for data mining and exploration. It studies the combination of the EGO (Efficient Global Optimization) algorithm and LSTM (Long Short-Term Memory), using the prediction function and error function of the parallel EGO-LSTM model as the foundation to derive analytical expressions combining the sample distribution probability density function and the cumulative distribution function. Then, using the PIE (Pseudo Expected Improvement) criterion, new sample points are selected to update the parameters of the prediction model. Finally, the optimized LSTM model is used to predict the error data of the voltage transformer. To verify the superiority of the proposed algorithm, it is compared with other optimization algorithms, and simulation results show that the prediction error of this method is generally lower than that of other optimization algorithms.