Study On Insulation Early Warning Of Distribution Line Based On Ground-fault Information

As people more and more rely on the electrical energy,whatever in their daily lives or in the production process electricity will be needed all the time.For users,the end of the power delivered is the distribution network.If a fault occurs in the distribution lines,it will directly affect customers.Therefore,to protect the safe operation of the distribution line is an important issue.Achieve distribution line insulation status online monitoring and warning fault's signal in time,which has a special significance for the safe transport of electricity.This paper describes the current situation at home and abroad for basic research on the distribution line fault diagnosis,and use neural network combined with phase plane and wavelet packet excellent capability in the processing waveform to forecast the value of insulation deterioration on the distribution lines.Meanwhile,the description of when had to do the early warning were studied.On the problem what algorithm to be used to predict distribution lines' insulation,we use the GA-BP neural network as a theoretical basis for the superiority of GA-BP neural network in forecasting and define the concept of insulation deterioration values,and proposed to describe the current state of temporary ground fault insulation characteristic quantities:the duration of the fault,the fault waveform feature amount,two adjacent fault is within two hours of the like.So that,establishing a mapping between the insulating state the amount with insulation deterioration characteristic value by GA-BP neural network.In this paper,use the small current grounding fault production recording device,fault duration and the time difference between two consecutive failures can be directly extracted.However,fault waveform feature amount can not be obtained through the computer observation,you have to go through signal processing tools.On the problem how to extract fault waveform feature amount,there are two methods.The first is a method of phase trajectory,phase trajectory can be obtained by mapping phase plane transient ground fault wave occurs the moment,partition of the phase plane,calculates the percentage share of discrete points within each region as a fault waveform feature amount.The second is based on wavelet packet transformation,the current temporary ground fault waveform is decomposed to extract the fundamental component amplitude,harmonic distortion and energy entropy to characterize the three characteristic quantities.For the first method of the disadvantages of the predicted range is relatively large due to the extraction of all temporary ground faults caused by the waveform.In the second method is optimized because the amplitude of the fundamental frequency component were be used to remove a partial ground fault transient data,thereby improving the accuracy of the prediction algorithm.Through ground fault historical data research in power enterprises to validate the results show that the algorithm uses two methods can be relatively accurate prediction of deterioration in insulation value on the current ground fault of this section feeder,and kept the value within a certain error range.Setting a threshold of section feeder distribution network can make timely warning,can make good power supply bureau to adjust the time and ready to maintain or replace equipment,can ensure continuity of supply,can improve power supply efficiency.