Research On Fault Diagnosis Method Of Optical Fiber Composite Submarine Power Cable

Optical fiber composite submarine power cable has a widely used in engineering applications because it can transfer the information and power simultaneously. In order to reduce the time of cable fault repair and ensure the normal operation,reduce the economic losses, it is necessary to study fault diagnosis method of optical fiber composite submarine.This project carried out its research relying on the cable of online monitoring optical fiber composite submarine power cable project of an electric power company. Firstly, submarine cable monitoring by distributed optical fiber sensing technology are summarized, the wavelet packet analysis and neural network theory are introduced, a combination of wavelet packet and BP neural network method for fault diagnosis method of optical fiber composite submarine are used,the classifications and the reasons for electrical faults of submarine cable are analysised, submarine cable faults data are simulated by according long-term accumulation of cable fault measure experience and submarine optical fiber sensor of temperature and strain change characteristics;Secondly, the wavelet analysis algorithm is used to deal with the fault data,the useful information of original signal are reduction,and the energy, variance and the three kinds of fault characteristics of Shannon entropy of each fault are extracted respectively by using wavlet packet analysis to be diagnosis basis for pattern recognition; Thirdly, for electrical and mechanical faults, two kinds of neural networks, PNN and BP are established respectively. The best parameters of each network are determined. The feature vectors are extracted from two kinds of networks, and the fault zones of three kinds of fault characteristics are compared. Finally, the fault diagnosis platform of submarine cable is set up by using LabVIEW virtual instrument, and the data can be read, analyzed, diagnosed and displayed.Simulation results show that three kinds of feature vectors can be characterized in different degree. For the electrical fault, the standard deviation is the best. For the mechanical fault, the energy of each frequency band as the feature vector is the best. In the selection of neural network diagnosis, overall performance of PNN network is higher than that of BP network, more suitable for the cable fault diagnosis classifier.The results of this study for cable fault diagnosis provides a new way of economical and reliable, in order to provide theoretical basis for the submarine cable line fault monitoring, on submarine cable fault diagnosis and has important reference value.