Fault Location Of Wind Farm Collector Line Based On Deep Learning

In order to achieve the goal of "carbon neutrality",China will vigorously develop wind power,photovoltaic and other renewable energy sources on the basis of the existing energy structure to facilitate carbon emission reduction.It is expected that wind power will further expand its share in China’s energy structure in the future and provide obvious economic and social benefits.The phenomenon of abandoning wind power in wind farms occurs from time to time due to the long time taken to find the fault point of the collection line of wind farms,which has plagued many wind power enterprises.This paper summarizes and analyzes the advantages and disadvantages of existing fault location technologies for transmission and distribution networks,and discusses the possibility of fault location on collecting lines from different perspectives based on the unique structure of wind farms.Its contents include the following parts:A fault location scheme based on FDM and twin network is designed in this paper.In order to meet the requirements of locating faults in the wind farm operation,and it is difficult to obtain samples without preserving the fault waveform in the operation site,the twin neural network is used to enhance the fault samples and the FDM algorithm is used to efficiently extract fault features.Simulation results of PSCAD/EMTDC show that this method can efficiently identify faults in different sections.In this paper,a fault location scheme based on improved depth autocoding network is designed.It is suggested that the location of installation measuring points should be reasonably selected.Based on this,an autoencoder network is constructed.By analyzing the zero-sequence network,autoencoders are used to perform regression processing on features.PSCAD/EMTDC simulation shows that the proposed scheme has higher accuracy than other schemes.This paper designs a fault location scheme based on time series.Considering the internal harmonics of wind farm and the influence of noise of measuring points,the difficulties of using traveling wave method in wind farm are analyzed.In this paper,the combined multi-terminal method is used to locate the wind farm.CNN-LSTM algorithm is used to filter out the noise part of the voltage waveform to obtain the fault traveling wave component.After that,the second-order Teager energy operator is used to calibrate the arrival time of the wave head and the multi-terminal method is used to complete the location.PSCAD/EMTDC simulation shows that the proposed scheme is feasible in wind farms.