Study On Method Of Detecting Power Transformer Winding Deformation Fault Based On Pulse Coupling Injection

As the development of global energy internet,there is an increasingly rapid demand for constructing the strong smart grid,how to ensure the stability and reliability of power system and power equipment is always the hotspot in power technology field.Power transformer is regarded as the most significant power transmission equipment in power system,and winding deformation is one of the main reasons to cause the internal failure inside transformer.How to diagnose the winding deformation of a running transformer,track its status and make a warning before the catastrophic failure are the key scientific and technical issues which should be urgently solved.Frequency response analysis has been developed as the basic method that transformer manufacturers and power utility follow to diagnose winding deformation for a long time,severely deformed failure has been successfully recognized in multiple cases,and major blackouts have been therefore avoided.However,the necessary condition for implementation of this method is the outage of power transformer;offline technique cannot satisfy the developing trend of condition based online monitoring and intelligent diagnosis of power equipment.Through the detection of winding failure when transformers are in service,the available test data can provide important guidance.By obtaining the status of winding with timely and accurate online technique,the informative indicators can be offered,and the overhaul and maintenance of transformer can be performed with specific target,which shows broad application prospects in rapid identification of transformer accident and the assessment of short circuit capacity for transformer,et al.Therefore,studying the online or live detection and diagnosis technique of winding deformation is of significance to promote the development and application of condition assessment for transformer in smart grid of China.This thesis proposes and studies the method for detecting transformer winding deformation fault based on pulse coupling injection,and this method is attempted to being applied in running power equipment for live detecting of failure.The basic principle of method is described;the criteria for determining the key parameters of injected excitation pulse are explored.The equipment for fault detection is developed and their performance testing is conducted.The impact of coupling circuit on online frequency response is explored by establishing the transformer winding equivalent lumped parameter electrical model,in which the bushing coupling circuit is considered.The validation experiment is carried out and the detection equipment is installed on a running electrical device for sampling data.The impact of electromagnetic interference source on the measurement is preliminarily analyzed and the suppression steps are put forward.According to the test feedback,methods of obtaining impulse frequency response signature based on short time Fourier transform and multiscale complex continuous wavelet transform are proposed,which surpass the traditional fast Fourier transform.The simulation experiments of transformer winding faults are performed;the diagnosis and identification method of winding deformation fault based on support vector machine is studied.The main conclusions of this thesis are as follows:(1)The types and key parameters of injected excitation pulses are determined,the waveform of nanosecond square wave pulse doesn't change significantly with the disturbance of the load,thus the pulses are suitable to be the excitation signal for detection.Furthermore,it is noted that the pulse width and pulse rise time of nanosecond pulse not exceed 800 ns and 100 ns,respectively,and the amplitude of nanosecond pulse should be at least 600 V.Increasing the amplitude of pulse can improve the signal to noise ratio.The configuration of bushing is skillfully used to design a bushing capacitive coupling sensor,the finite element simulation and withstand voltage tests are performed to verify that the installation of sensor has no significant influence on normal operation of bushing.A signal online injection/measurement circuit is designed,the signal injection circuit is developed to realize the lossless transmission of signal and filter the power frequency high voltage to protect the detection equipment.The measurement circuit is manufactured to sample the response voltage signal of a transformer when it is in service,and its frequency band satisfies the demand for frequency response analysis.A portable winding deformation fault detector is designed based on FPGA,which has functions of producing adjustable pulse signal,sampling by two channels,processing the data,et al.(2)The transformer winding equivalent online electrical model considering the effect of bushing coupling circuit is built,the impedance relationship is adopt to interpret the morphological characteristics of online frequency response signature.It is recommended that the low frequency band of signature represents the characteristic of bushing coupling circuit,and the middle and high frequency bands of signature still represent the characteristic of winding,and a key frequency point should be determined to guide the actual work.The impact of coupling capacitance variation and bushing insulation breakdown on online frequency response signature are studied,the results show that the frequency response signatures are shift towards to the upper and lower directions induced by above two factors,which can be used to distinguish the bushing coupling circuit failure and internal winding failure.In addition,the shifts of resonance and anti-resonance of signature are recommended to be the diagnostic indicators.(3)The transformer offline experiments verify that the impulse frequency response signatures can reflect the status of health and faulty transformer,and power frequency high voltage is applied to a transformer to validate the safety of the detection method and equipment,repeated experiment of power transformer verifies that the repeatability of detection method is enough high when the lead wires are fixed,which lay the foundation for actual application of the method.In addition,the proposed method is applied to 750 kV high voltage shunt reactors for online detection of internal winding faults,the preliminary offline experiments and fault simulation experiment are carried out to verify the effectiveness and feasibility of the proposed method.Finally,the detection equipment is firstly installed on running three-phase 750 kV high voltage shunt reactors,which realize the application of the proposed method.The impact of electromagnetic interference source on the measurement is briefly analyzed,and some hardware and software measures to inhabit the interference are introduced.(4)Short time Fourier transform is used to process transient detection signal,the obtained impulse frequency response signature shows high definition compared with that of FFT method,however,the essential of short time Fourier transform makes it more suitable for processing the sub-stable signal.The method of obtaining impulse frequency response signature from transient signal based on complex continuous wavelet transform is then studied;the simulation results of constructed transient signal and transformer equivalent electrical model show that the frequency components calculated from wavelet transform are much closer than those of FFT method.The transformer experiments show that the impulse frequency response signatures based on wavelet transform show more similarities with sweep frequency response signatures than those based on FFT method,and the signatures are less affected by the noise,which benefits to the actual winding deformation diagnosis.(5)Experiments of simulating transformer winding deformation faults are carried out,the impact of winding inter disk short circuit fault,radial deformation and disk space variation on the morphological characteristic of impulse frequency response signature are analyzed.The frequency variation of effective extreme point of frequency response and the mean square error indicators in different frequency bands are proposed to construct the characteristic parameters for fault detection,respectively.The support vector machine algorithm is used to identify the types of winding deformation failure.The analysis shows that the average identification accuracy for 10 groups of test samples by using two characteristic parameters are 78.1% and 83.3%,respectively,both satisfactory results are obtained.