Font Size: a A A

Study On High-voltage Cable Fault Location Method In Coal Mine

Posted on:2019-03-09Degree:DoctorType:Dissertation
Country:ChinaCandidate:J F LiFull Text:PDF
GTID:1311330542975855Subject:Power system and its automation
Abstract/Summary:PDF Full Text Request
The continuous and stable operation of coal mine power supply system is one of the important prerequisites to ensure safety of the workers and the production in coal mine.Due to the complex working environment and human factors,short or grounding fault in power cable are serious threat to the reliability of power supply.Once earth fault occurs,it will seriously affect the safe operation of coal mines,and even lead to catastrophic accidents.It is important to cut off the fault line quickly and accurately,troubleshooting is particularly important.High voltage power supply system of coal mine is presented as a multi branched structure with much difference in the length of each branch;arc suppression coil compensation effect to make line fault and non fault line steady characteristics of different small.Besides,much field interference of the fault signal,the low signal-to-noise ratio and fault arc effects lead to the mine fault positioning accuracy and difficulty.In this paper,according to the practical application,new fault location methods are proposed based on the original relay protection.The selection of appropriate line model is the basic of simulation.There are three main kinds of line model,they are lumped parameter model,distributed parameter model and frequency dependent parameter model.The emphasis and application occasions are of difference among these models.The impedance characteristics of transmission and cable lines,different from other power devices,is distributed uniformly in the line,and have the characteristics of distributed parameters.After comparing the modeling mechanism of each model,the propagation characteristics of transient traveling wave signals wiht different models are compared,and then the paper point out that it is necessary to use frequency dependent parameter model to simulate the traveling wave of cable line.Based on the analysis of the traveling wave theory,the concept of virtual power is introduced to explain the sudden change of the signal in the process of traveling wave.Then it is pointed out that both the velocity and wave shape will change when the traveling wave spreading in cable line,which phenomenon is called traveling wave dispersion.The dispersion is more serious in cable line than in transmission line.The dispersion of traveling wave in the cable is analyzed in frequency dependent parameter model,traveling wave velocity and arriving time of traveling wave head are further discussed.The deformation and attenuation of traveling wave propagation in cable lines at different locations are compared by building basic frequency dependent parameter model.It is verified that the traveling wave attenuations are different at different observation points,which brings challenge to determine the arriving time of the traveling wave head.Then the influence of traveling wave with different frequency on the electrical parameters of the cable is studied.The result shows that different frequency traveling wave signals cause the effective size and medium of the cable to change.As a result,the basic electrical parameters of cable R,L,G,C may not be fixed values.The electrical parameters are closely related to the instantaneous frequency of transient traveling wave signals,in my work,not only the corresponding relation between wave velocity and frequency is considered in the calculation formula of traveling wave velocity,but also the changes of the electrical parameters caused by the frequency are taken into account to obtain a more accurate wave velocity.How to get instantaneous frequency of transient signal is a key step in velocity calculating.The phase to mode transformed signal are processed by HHT,and the instantaneous frequency corresponding to the mutation time of the traveling wave head are obtained at the same time.Therefore a new double-teminal fault location is presented,in which the important effect of instantaneous frequency on the final location is fully considered.Simulation shows method in this paper is more accurate than traditional traveling wave location method.At present,there is a lack of high speed communication function among underground power distribution stations.With the development of WAMS technology and the successful application of IEC61850 communication protoco in digital substation on the ground,which brings a great deal of inspiration to the power supply and distribution system in coal mine.With the improvement of WAMS timing accurate and the development of highspeed sampling chip,the double-terminal location method can be applied even in the short lines.Multi layered structure of underground coal mine contains many cable lines,it is an important step to identify the fault section after fault occurred.The relationship between modes of traveling wave and initial traveling wave is researched.On the basis of which,the algorithm of section location is proposed.Singular value decomposition(SVD)has the ability to indicate the abrupt change of the signal.It can not only calibrate the mutation time,but also indicate the amplitude of the abrupt change.The detailed process of SVD and construction of Hankel matrix for a signal are described.The proposed algorithm is based on the comparison of each traveling wave signal’s second SVD components.The largest of which is the fault section or fault line.The method has wild adaptability and can accurately identify the fault section when the system has different fault types,different fault grounding angles and different grounding resistances.There are two crucial factors in traveling wave fault location,they are time and velocity of traveling wave.Once the dimension of the Hankel matrix is determined,the width of the mutation time is determined.The situation is of difference by wavelet transform.With the increase of wavelet transform scale,the width of the position for the signal is also increased.After voltage and current are acquisited at both sides of fault line,forward and backward traveling wave processe by SVD,according to the polarity,first two waveforms of the fault line are identified so as to determine more precise arriving time of waveforms.After wavelet transform,the largest scale of the two waveforms is found out,and the center frequency of that scale is calculated as the frequency of the wave velocity.On the basis of above,the double single-terminal fault location is proposed.Each single-terminal gets a fault location length,then length of two sides processes by error correction algorithm,making the results more accurate.Simulation proves the correctness of this approach.When the fault occurs in the high voltage cable of underground coal mine,,the transient information is different at different distances.If the inference process is regarded as a black box,different voltage and current signals correspond to different fault locations.How to form a black box is a key step.As a rising subject,machine learning has been applied in many fields.Extreme learning machine(ELM)is a kind of machine learning algorithm,who has a strong ability to classify and fit.To form a good generalization of the box,the characteristics of the input data must be representative.Three phase transient voltage and current signals are decomposed by 3 levels wavelet packet and the coefficient of each node is reconstructed.Then the coefficients of each wavelet packet node are reconstructed into 6 statistical features(energy,mean,standard deviation,kurtosis,entropy,skewness),and then the genetic algorithm is used for feature selection.The features which can be more closely linked to the results are selected to establish ELM model.In this paper,the feature selection and the extreme learning machine are combined to establish the model,and the performance of the model is evaluated by the square error value and the determination coefficient.The comparison shows that the model established in this paper is more excellent.After the appropriate model is formed,different fault locations are used to test the model,whose generalization ability is expected.And only in the case of 10 kHz sampling rate can ensure the accuracy.At the end of this paper,a summary of the work done in this paper is given,and the future research directions are also discussed.
Keywords/Search Tags:WAMS, fault location, HHT transform, SVD, feature selection, ELM
PDF Full Text Request
Related items