EAO-BiTCN-BiGRU-Based Overvoltage Prediction for Variable Frequency Motor Ends in Oil Drilling Platforms

As the power source for oil drilling operations, the operational reliability of variable frequency motors directly affects drilling stability. Insulation faults, as potential hidden hazards in motor operation, are characterized by strong concealment and slow progression, making them a critical factor influencing long-term stable motor operation. Therefore, research on insulation fault diagnosis in variable frequency motors is of significant importance for ensuring efficient drilling production. This study addresses the rapid insulation degradation of the main drive motor on the "Exploration No. 8" drilling platform caused by voltage spikes. We propose a hybrid prediction model based on an Improved Aquila Optimizer (EAO)-optimized Bidirectional Temporal Convolutional Network-Bidirectional Gated Recurrent Unit (BiTCN-BiGRU). By analyzing the generation mechanism of overvoltage and optimizing hyperparameters using EAO, the model leverages BiTCN to extract multi-scale local features and BiGRU to model bidirectional temporal dependencies, aiming to accurately predict the amplitude and trend of motor terminal overvoltage and provide reliable data support for insulation degradation analysis.Experimental results demonstrate that the proposed model significantly outperforms traditional neural network models in prediction accuracy, effectively capturing the temporal characteristics of voltage variations and achieving high-precision prediction of motor terminal overvoltage.