摘要
为解决现有异常辨别方法误识率高的问题,研究基于深度神经网络的输电线路异常自动辨识方法。通过小波多尺度分解、重构和奇异值分解处理输电线路信号,利用向量机分类并获得信号模态函数,通过描述信号序列间的关系获取最优特征分量值,建立深度神经网络异常辨识模型,使用卷积操作学习不同空间内的关联特征,最后利用长短期记忆网络(Long Short-Term Memory,LSTM)处理时间序列数据,传输至模型识别异常。通过与阈值比较判断输电线路运行状态,实验证明该方法误识率为0.16%,能准确识别输电线路异常。
Due to the high error rate of existing anomaly identification methods,a deep neural network-based automatic identification method for transmission line anomalies is studied.Transmission line signals are processed by wavelet multiscale decomposition,reconstruction and Singular value decomposition.The vector machine is used to classify and obtain the signal modal function.Obtain the optimal feature component value by describing the relationship between signal sequences.Establish a deep neural network anomaly identification model and use convolutional operations to learn association features in different spaces.Use Long Short-Term Memory(LSTM)to process time series data and transmit it to the model for anomaly recognition.By comparing with the threshold to determine the operating status of transmission lines,experiments have shown that this method has a false recognition rate of 0.16%and can accurately identify anomalies in transmission lines.
作者
曹成顺
CAO Chengshun(Enshi Power Supply Company of State Grid Hubei Electric Power Co.,Ltd,.Enshi Hubei 445000,China)
出处
《信息与电脑》
2023年第15期165-167,共3页
Information & Computer
关键词
深度神经网络
输电线路
异常辨别
辨别方法
deep neural network
transmission line
anomaly discrimination
discrimination method
