小样本下基于三元组原型网络的电压暂降类型识别方法

Identification Method of Voltage Sag Types Based on Triplet Prototype Network Under Small Samples

区域电网中电压暂降标签样本量有限,传统深度神经网络在小样本下难以进行有效学习,实际工程中应用困难。文章提出了一种小样本下基于三元组原型网络的电压暂降识别模型。使用有限的训练样本构建电压暂降三元组,并对三元组进行关系学习来提取暂降特征。增加高效通道注意力机制捕捉样本的跨通道特征交互信息,使模型在提取的特征分布具有高类内聚合性与类间可分离性。使用原型分类器对样本特征与类原型进行相似度对比来判定样本类别,保证了模型的分类鲁棒性。文章分别使用仿真数据与实测数据进行验证。实验表明,相较于传统模型,该方法能够在小样本下实现准确的电压暂降分类效果。

The number of voltage sag labels in the regional power grid is limited,and traditional deep neural networks are difficult to learn effectively under small samples,and it is difficult to apply in practical engineering.This paper proposes a voltage sag identification model based on triple prototype network under small samples.Voltage sag triples are constructed using limited training samples,and relational learning is performed on the triples to extract sag features.An efficient channel attention mechanism is added to capture the cross-channel feature interaction information of samples,so that the model has high intra-class aggregation and inter-class separability in the extracted feature distribution.The prototype classifier is used to compare the similarity between the sample features and the class prototype to determine the sample category,which ensures the classification robustness of the model.In this paper,simulation data and measured data are used for verification.Experiments show that,compared with the traditional model,the method in this paper can achieve accurate voltage sag classification in small samples.