Power Cable Fault Location Technology Based On Deep Learning

In the process of rapid urban development,the problem of land shortage is increasing.Due to its small footprint,buried cables have gradually replaced overhead cables as the main way of laying cables.When buried cables are put into operation,regular cable safety maintenance can effectively reduce the failure rate.However,when cables are laid,operated,and maintained,cable faults still occur due to operating errors or external force damage.If the failure point of the cable cannot be found and processed in time,it will cause inconvenience and economic loss to production and life.Therefore,this thesis deeply researches the cable fault detection method in non-operating state,and the processing process of the cable discharge sound signal in the magnetic sound synchronization is analyzes.Aiming at the phenomenon that there is a lot of environmental noise in the collected discharge sound,a sound noise reduction algorithm based on deep learning is adopted,and a cable fault location system was designed.First of all,the background and research status of cable fault location technology are studied,and the development of machine learning in the field of sound noise reduction is explained.The causes of cable faults and the entire process of power cable fault detection are analyzed.In order to accurately locate the cable fault,a high-voltage pulse generator is used to apply a high-voltage DC pulse signal to the faulty cable.A breakdown of the cable fault point will generate a discharge sound signal and a magnetic field signal.These two signals are received through the sensor,and the distance between the measurement point and the actual failure point is determined according to the magnitude of the discharge sound obtained by the system.Secondly,a large amount of environmental noise is mixed in the collected discharge sound signal,which will affect the accurate judgment of the size of the discharge sound.It is necessary to perform noise reduction processing on the collected signal.Therefore,this thesis comprehensively explores the advantages and disadvantages of traditional sound noise reduction algorithms in practical applications.Based on the Log-Spectral Amplitude Minimum Mean Square Error,it is proposed that sound noise reduction algorithm based on deep learning.The noise reduction algorithm includes two parts: network training and sound noise reduction.In the network training part,the logarithmic power spectrum characteristic value of the signal is used as the input of the network,the network structure is built according to the actual characteristics of the signal,the learning algorithm and transfer function of the network are determined,and the network is trained using a large number of sample sets.In the sound denoising part,the characteristic value of a noisy discharge sound signal is input into the trained network to obtain the estimated characteristic value of the clean sound signal,and then the time domain signal of the clean cable discharge sound is obtained by the waveform reconstruction method.Finally,according to the characteristics of cable discharge sound and magnetic signals,a cable fault location system is designed.The system mainly includes system hardware design,system software environment construction and application program design.The cable fault location system uses piezoelectric sensors and electromagnetic induction coils to collect discharge sound signals and magnetic field signals.The collected signals are amplified,filtered,and A/D transmitted to the embedded processor through a high-speed SPI interface.The processor analyzes the data and reduces the noise.The noise-reduced sound signal is output to the headphones after D/A conversion.The sound signal transformation trend and the strength of the magnetic field signal are displayed on the LCD to help judge the fault point.After field verification,the functions of each part of the system can run normally,and the fault points of the cable can be located effectively.