基于数据增强和特征注意力机制的灰狼优化算法−优化残差神经网络变压器故障诊断方法
摘要：为提高变压器故障诊断的准确性，提出一种基于数据增强和特征注意力机制的灰狼优化算法（grey wolf optimization，GWO）−残差神经网络（residual neural network，ResNet）故障诊断方法。针对变压器不平衡数据集对故障诊断模型产生的影响，利用带梯度惩罚的Wasserstein生成对抗网络（generative adversarial network with gradient penalty ，WGANGP）对变压器数据进行数据增强。其次，在诊断模型的输入侧引入特征注意力机制，提升模型对平衡数据集中关键特征的敏感性。然后，为加速模型的收敛性，在训练的早期利用GWO-ResNet。最后基于某实测变压器数据集对所提出WGANGP-ATT-GWO-ResNet故障诊断模型的有效性进行验证。
Abstract:To improve the accuracy of transformer fault diagnosis, this paper proposes a GWO-ResNet fault diagnosis method based on data augmentation and feature attention mechanism. Aimed at the effect produced by imbalanced dataset on the power transformer fault diagnosis model, the Wasserstein generative adversarial network with gradient penalty (WGANGP) is utilized to make data augmentation for the power transformer dataset. Secondly, to enhance the sensitivity of the model to the key features in the augmented dataset, a feature attention mechanism is introduced into the input side of the model. Thirdly, in order to accelerate the convergence of the model, the grey wolf optimization algorithm (GWO) was used to optimize the residual neural network (ResNet) in the early stage of training. Finally, the validity of the proposed WGANGP-ATT-GWO-ResNet fault diagnosis model is verified based on a measured power transformer dataset.
标题：基于数据增强和特征注意力机制的灰狼优化算法−优化残差神经网络变压器故障诊断方法
英文标题：GWO-ResNet Power Transformer Fault Diagnosis Method Based on Data Augmentation and Feature Attention Mechanism
