基于递归图和MobileViT的复合电能质量扰动识别  

Complex Power Quality Disturbances Recognition Based on Recurrence Plot and MobileViT

作　　者：刘道庆 臧润泽 李鹏 王振克 程晨 LIU Daoqing;ZANG Runze;LI Peng;WANG Zhenke;CHENG Chen(Huaihe Energy Power Group Co.,Ltd.,Huainan 232001,China)

机构地区：[1]淮河能源电力集团有限责任公司,安徽淮南232001

出　　处：《现代信息科技》2025年第1期30-34,共5页Modern Information Technology

摘　　要：针对复合电能质量扰动(Power Quality Disturbances,PQDs)信号在识别过程中出现的特征提取难、识别效率低等问题,文章提出一种高效的可视化方法和轻量级网络模型的电能质量扰动识别方法。首先,利用递归图(RP)方法对一维PQDs信号进行可视化以捕捉信号中的非线性动态特征。其次,根据IEEE Std 1159—2019标准对30种PQDs信号进行仿真生成,并采用RP方法建立图像数据集。再次,采用MobileViT轻量级网络对自建图像数据集进行特征提取与分类识别任务。最后,通过与传统轻量级网络MobileNetV2和MobileNetV3的对比实验和鲁棒性检验,验证了文章采用方法具有更高的识别准确率和更快的推理速度。Aiming at the problems of difficult feature extraction and low recognition efficiency in the recognition process of complex Power Quality Disturbances(PQDs)signals,this paper proposes a high efficiency visualization method and a PQDs recognition method of the lightweight network model.Firstly,the Recurrence Plot(RP)method is utilized to visualize one-dimensional PQDs signals to capture the nonlinear dynamic features in these signals.Secondly,according to IEEE Std 1159—2019 standard,30 types of PQDs signals are simulated and generated,and the image dataset is established by the RP method.Thirdly,the MobileViT lightweight network is used to perform feature extraction and classification recognition tasks on the custom image dataset.Finally,through comparative experiments and robustness tests with the traditional lightweight networks of MobileNetV2 and MobileNetV3,it is verified that the method adopted in this paper has higher recognition accuracy and faster reasoning speed.

关 键 词：电能质量扰动 递归图 可视化 MobileViT 轻量级网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]