贝叶斯优化辅助分类GAN的变压器故障诊断方法

Transformer fault diagnosis method based on improved AC-GAN of Bayesian Optimization

针对变压器故障数据难以获取且数据量不均衡性显著的问题,提出一种基于BO-AC改进GAN的变压器故障诊断方法。针对局放超声信号的噪声干扰问题,采用小波阈值降噪提升了数据的信噪比;以一维时域超声信号作为数据源,基于GAN算法进行样本生成和故障诊断,通过在经典GAN架构中添加辅助条件层,优化模型的泛化性能,采用BO策略优化模型超参数,提升识别性能。通过与多种方法在不同方案下识别效果的对比,验证了文中方法的有效性和优越性。文中方法基于不均衡实测数据训练后的测试准确率均在95%以上,解决了样本数据不均衡条件下变压器故障类型的准确诊断问题,为保障变压器的可靠运行提供了新方法。

In order to solve the problem that transformer fault data is difficult to obtain and the amount of data is unbalanced,a transformer fault diagnosis method based on BO-AC improved GAN is proposed.Aiming at the noise interference problem of partial discharge ultrasonic signal,wavelet threshold denoising is adopted to improve the signal-to-noise ratio of the data;With one-dimensional time-domain ultrasonic signal as the data source,sample generation and fault diagnosis are carried out based on GAN algorithm.By adding auxiliary condition layer in the classic GAN architecture,the generalization performance of the model is optimized,and the BO strategy is used to optimize the model super parameters to improve the recognition performance.The effectiveness and superiority of the proposed method are verified by comparing the recognition results with various methods in different schemes.The test accuracy of the method based on unbalanced measured data training is above 95%,which solves the problem of accurate diagnosis of transformer fault types under the condition of unbalanced sample data,and provides a new method to ensure the reliable operation of transformers.