基于生成对抗网络的电能质量信号压缩重构方法  

Compression and Reconstruction Method for Power Quality Signals Based on Generative Adversarial Networks

作　　者：简献忠[1] 王绪涛 王如志[3] JIAN Xian-zhong;WANG Xu-tao;WANG Ru-zhi(School of Optical-electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200090,China;School of Mechanical Engineering,University of Shanghai for Science and Technology,Shanghai 200090,China;School of Materials Science and Engineering,Beijing University of Technology,Beijing 100020,China)

机构地区：[1]上海理工大学光电信息与计算机工程学院 [2]上海理工大学机械工程学院,上海200090 [3]北京工业大学材料科学与工程学院,北京100020

出　　处：《控制工程》2021年第8期1654-1661,共8页Control Engineering of China

基　　金：国家自然科学基金资助项目(11774017)。

摘　　要：针对压缩感知在电能质量信号压缩重构方面存在前期稀疏处理过程繁琐、观测矩阵设计困难、压缩重构速度慢等缺点,首次提出一种基于生成对抗网络模型的电能质量信号压缩重构方法。该网络模型由生成器和鉴别器组成。生成器学习样本分布的特性,经过训练后应用到电能质量信号的压缩和重构过程中。鉴别器与生成器相互对抗以提高彼此性能。此外,该方法在原损失函数中加入重构损失和频域损失,进一步提升重构效果。实验结果表明,提出的压缩重构方法不仅避免了前期对信号进行稀疏处理,而且具有重构效果好、重构速度快、稳定性更强的优势。In view of the disadvantages of compressive sensing in compression and reconstruction of power quality signals,such as tedious sparse processing in the early stage,difficult observation matrix design,and slow compression and reconstruction speed,this paper proposes a compression and reconstruction method for power quality signals based on generative adversarial network model for the first time.The network model is composed of generator and discriminator.The generator learns the characteristics of sample distribution and is applied to the compression and reconstruction of power quality signals after training.The discriminator fights against the generator to improve each other.In addition,the reconstruction loss and frequency-domain loss are added into the original loss function to further improve the reconstruction effect.The experimental results show that the method proposed in this paper not only avoids sparse processing of signals in the early stage,but also has the advantages of good reconstruction effect,fast reconstruction speed and stronger stability.

关 键 词：电能质量信号 压缩感知 生成对抗网络 稀疏度 重构算法 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]