电力电缆沿线温度的实时在线监测能够有效避免电缆过热导致的安全事故发生,分布式光纤温度传感技术由于具有耐高温、灵敏度高和抗电磁干扰等优点在电缆温度监测中得到了广泛应用.然而对长距离的电力电缆进行分布式测温时,温度信号的信噪比随距离的增长而降低,影响电缆温度测量的准确度.针对此问题,本文设计了一种基于卷积神经网络的降噪方法,在大量先验数据的基础上对神经网络的参数进行优化更新,将其应用于长距离分布式测温信号进行噪声的滤除.实验结果表明,本文的消噪方法能够将长度为 11 km 的分布式测温信号的噪声水平从原始的±17.5℃抑制到±1℃内,有效抑制了噪声,提高了测温准确度.
Research on noise reduction method of distributed fiber optic temperature sensing system for power cables based on convolutional neural network
Real-time online monitoring of temperature along power cables can effectively avoid the occurrence or ex-pansion of cable fires.Distributed fiber optic temperature sensing technology has been widely used in cable temper-ature monitoring by virtue of its advantages of anti-electromagnetic interference and high temperature resistance.However,the signal-to-noise ratio of temperature signals decreases with the growth of distance when distributed temperature measurement is performed on long-distance power cables,which affects the accuracy of cable tempera-ture measurement.To address this problem,a noise reduction method based on convolutional neural network is de-signed in this paper,and a large number of a priori data sets are used to train the neural network noise elimination model,which is applied to the long-distance distributed temperature measurement signal for noise filtering.The ex-perimental results show that the noise reduction method in this paper can suppress the noise level of the distributed temperature measurement signal with a length of 11 km from the original±17.5℃to within±1℃,effectively sup-pressing the noise and improving the accuracy of temperature measurement.
power cablesdistributed temperature measurementconvolutional neural networksdenoising method
NETL
NSTL
万方数据
维普
