摘要
低压配电线路可能产生故障电弧引发电路故障,为了区分正常电流和有故障电弧的电流,提出一种基于电流二维图像与卷积神经网络的故障电弧检测方法。首先利用变分模态分解重构电流信号,解决非线性型负载正常和故障电流难以辨识的问题;再使用马尔可夫转移场算法对重构电流信号进行二维图像编码,生成特征图像数据集。为了提高故障电弧检测的准确率和效率,构建了基于卷积神经网络的故障诊断模型,将所提特征图像数据集与未经信号重构的特征图像数据集分别输到所构建的诊断模型进行对比验证,结果表明,所提方法能有效改善非线性负载状态混淆,故障检测的平均准确率达到99%。
Low-voltage distribution lines may produce arc faults,which may cause circuit faults.To distinguish between normal and fault currents,a fault arc detection method based on two-dimensional current images and convolutional neural networks is proposed.Firstly,the current signals are reconstructed using variational mode decomposition to address the challenge of distinguishing between normal and fault currents in nonlinear loads.Then,the Markov transition field algorithm is utilized to encode the reconstructed current signals into two-dimensional images,generating a dataset of feature images.To enhance the accuracy and efficiency of fault arc detection,a CNN-based fault diagnosis model is constructed.The proposed feature image dataset and the dataset of feature images without signal reconstruction are respectively fed into the constructed diagnostic model for comparison and validation.Results indicate that the proposed method effectively mitigates the confusion caused by nonlinear load states,achieving an average detection accuracy of 99%.
作者
董志文
苏晶晶
DONG Zhiwen;SU Jingjing(School of Electronic,Electrical Engineering and Physics,Fujian University of Technology,Fuzhou 350118,China;School of Computer and Data Science,Minjiang University,Fuzhou 350108,China;Zhejiang Institute of Mechanical&Electrical Engineering Co.,Ltd.,Hangzhou 310051,China)
出处
《福建理工大学学报》
CAS
2024年第4期371-378,共8页
JOURNAL OF FUJILAN UNIVERSITY OF TECHNOLOGY
基金
福建省自然科学基金(2020J05170)
闽江学院科研项目(MYK21014)。
关键词
故障诊断
马尔可夫转移场
变分模态分解
卷积神经网络
fault diagnosis
Markov transition field
variational mode decomposition
convolutional neural networks
