基于VAE预处理和RP-2D CNN的不平衡负荷数据类型辨识方法  

Data Type Identification Method for Imbalanced Load Based on VAE Preprocessing and RP-2D CNN

作　　者：黄冬梅 吴志浩 孙园 胡安铎 时帅 孙锦中 HUANG Dongmei;WU Zhihao;SUN Yuan;HU Anduo;SHI Shuai;SUN Jinzhong(College of Electronic and Information Engineering,Shanghai University of Electric Power,Shanghai 201306,China;College of Electrical Engineering,Shanghai University of Electric Power,Shanghai 200090,China;College of Mathematics and Physics,Shanghai University of Electric Power,Shanghai 201306,China)

机构地区：[1]上海电力大学电子与信息工程学院,上海201306 [2]上海电力大学电气工程学院,上海200090 [3]海电力大学数理学院,上海201306

出　　处：《电力系统及其自动化学报》2022年第10期66-72,80,共8页Proceedings of the CSU-EPSA

基　　金：上海市科委地方院校能力建设项目(20020500700)。

摘　　要：针对负荷数据类型辨识中存在的类别不平衡及特征提取不足的问题,提出一种基于变分自编码器预处理和递归图-二维卷积神经网络的不平衡负荷数据类型辨识方法。首先,利用变分自编码器的过采样方法对少数类样本进行平衡化处理。然后,使用递归图算法将负荷曲线图像化。最后,根据二维卷积神经网络求取分类结果。算例分析表明,变分自编码器能有效地改善负荷数据中存在的类别不平衡问题,提高少数类的召回率;同时,相比于序列输入的分类器模型,经过递归图编码后,其图像输入的二维卷积神经网络模型有更高的分类准确度。Aimed at the problems of class imbalance and insufficient feature extraction in load data type identification,an imbalanced load type identification method based on variational auto-encoder preprocessing and recurrence plot two-dimensional convolutional neural network is proposed in this paper.First,the over sampling method of variational auto-encoder is used to balance a few classes.Then,the load curves are imaged using the recurrence plot algorithm.Finally,the classification results are obtained using two-dimensional convolutional neural network.The analysis result of an ex⁃ample shows that variational auto-encoder can effectively solve the problem of class imbalance in load data and improve the recall rate of a few classes.At the same time,compared with the classifier model of sequence input,the two-dimen⁃sional convolutional neural network model of image input which is coded by recurrence plot has a higher classification accuracy.

关 键 词：变分自编码器 递归图 类别不平衡 负荷分类 

分 类 号：TM714[电气工程—电力系统及自动化]