Transformer partial discharge source location method and system based on high frequency current and ultra-high frequency method

Abstract

The present invention proposes a method and system for locating a local discharge source of a transformer based on a combination of high-frequency current and ultra-high frequency methods, and relates to the technical field of transformer discharge source locating. It includes obtaining the current waveforms of the first and last terminals of the high / medium / low voltage side of the three-phase winding, obtaining identification parameters and standard parameters, forming an input matrix with the standard parameters, fitting the relationship between the coil number and the input matrix, and obtaining a linear regression relationship; judging the approximate location of the discharge by identifying the parameters, screening out ultra-high frequency sensors near the approximate location of the discharge, and judging the signal-to-noise ratio of the local discharge signal collected by the screened ultra-high frequency sensors; if the signal-to-noise ratio is higher than a set threshold, using the ultra-high frequency method to specifically locate the discharge source; if the signal-to-noise ratio is lower than the set threshold, calculating the coil number based on the linear regression relationship, and at this time, the discharge source is located at the numbered coil. The present invention combines the advantages and disadvantages of the high-frequency current method and the ultra-high frequency method, and can more accurately determine the location of the local discharge source.

Description

Technical Field

[0001] The invention belongs to the technical field of transformer discharge source positioning, and in particular relates to a transformer local discharge source positioning method and system based on a combination of high frequency current and ultra-high frequency method. Background Art

[0002] The insulation reliability of large oil-immersed transformers is a key link in the safe and stable operation of power grids. Partial discharge, as one of the manifestations of insulation degradation of electrical equipment, is an important cause of electrical equipment failure. Detecting and locating partial discharge can help operation and maintenance personnel predict the status of equipment and perform maintenance in advance. However, due to the complex internal structure of the transformer, the problem of low accuracy in locating partial discharge inside the transformer has not been solved.

[0003] High-frequency current method and ultra-high frequency method are currently the commonly used on-site transformer partial discharge detection methods. Among them, the circuit positioning technology based on high-frequency current signals mainly targets defects in windings. According to the time-frequency domain characteristics of the current signal, the discharge phase can be determined and the position of the wire turn where the partial discharge is located can be located.

[0004] The UHF method locates partial discharge in space based on the time difference of arrival (TDOA) principle. When the electromagnetic wave signal propagates only in the transformer insulating oil, the built-in UHF probe can achieve a better positioning effect. However, in reality, there are iron yokes, windings and other structures inside the transformer, which have a significant shielding effect on the electromagnetic wave signal. The specific analysis is as follows:

[0005] (1) When local discharge occurs between phases, part of the signal propagating toward the iron yoke or winding will be severely blocked. At this time, the installation position of the UHF sensor will have a significant impact on the partial discharge results;

[0006] (2) When local discharge occurs inside the winding, the electromagnetic wave signal is refracted and reflected between the windings, reaches the oil gap between the winding end and the iron yoke, and then propagates to the entire space. This process causes serious signal attenuation, and the UHF positioning accuracy is very low at this time. Summary of the invention

[0007] In order to overcome the deficiencies of the above-mentioned prior art, the present invention provides a transformer local discharge source positioning method and system based on the combination of high-frequency current and ultra-high frequency method, which realizes the purpose of accurate positioning as a whole. First, the approximate phase type of local discharge is determined by the high-frequency current method, and the linear regression relationship between the standard parameters of the current waveform and the coil number is fitted, and then at least 4 ultra-high frequency sensors near the phase where the local discharge is located are selected to collect the ultra-high frequency signal of the local discharge, and the following judgment is made: if the ultra-high frequency signal signal-to-noise ratio is higher than the set threshold, the ultra-high frequency signal is used for TDOA positioning; if the ultra-high frequency signal attenuates sharply and the signal-to-noise ratio is lower than the set threshold, it means that the local discharge is likely to occur inside the winding. At this time, the coil number is calculated based on the above-mentioned linear regression relationship, and the discharge source is located at the numbered coil.

[0008] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0009] The first aspect of the present invention provides a transformer partial discharge source location method based on a combination of high frequency current and ultra-high frequency method.

[0010] The transformer partial discharge source location method based on the combination of high frequency current and ultra-high frequency method includes the following steps:

[0011] A simulation model of a three-phase three-winding transformer is established, each phase coil is numbered, and a simulation signal is added to a certain numbered coil to simulate the current waveforms at the beginning and end of the high / medium / low voltage side of the three-phase winding;

[0012] Extract and screen the characteristic quantity from the current waveform obtained by simulation to obtain the identification parameters and standard parameters of the simulated current waveform, form the standard parameters into an input matrix, fit the relationship between the wire cake number and the input matrix, and obtain a linear regression relationship;

[0013] A classifier is trained based on the identification parameters of the simulated current waveform, field test data is obtained, and the trained classifier is used to determine the approximate location where the actual partial discharge occurs through the identification parameters of the field test data, and the UHF sensors near the approximate location where the partial discharge occurs are screened out;

[0014] Determine the signal-to-noise ratio of the UHF signal collected by the selected UHF sensor:

[0015] If the signal-to-noise ratio is higher than the set threshold, the UHF TDOA method is used to locate the discharge source;

[0016] If the signal-to-noise ratio is lower than the set threshold, the wire coil number is calculated based on the above linear regression relationship. At this time, the discharge source is specifically located at the numbered wire coil.

[0017] The second aspect of the present invention provides a transformer partial discharge source location system based on the combination of high frequency current and ultra-high frequency method.

[0018] The transformer partial discharge source location system based on the combination of high frequency current and ultra-high frequency method includes:

[0019] The simulation module is configured to: establish a simulation model of a three-phase three-winding transformer, number each phase coil, add a simulation signal at a certain numbered coil, and simulate the current waveforms at the first and last ends of the high / medium / low voltage sides of the three-phase winding;

[0020] The fitting module is configured to: extract and screen the characteristic quantity from the current waveform obtained by simulation, obtain the identification parameter and the standard parameter of the simulated current waveform, form the standard parameter into an input matrix, fit the relationship between the wire cake number and the input matrix, and obtain the linear regression relationship;

[0021] The approximate location identification module is configured to: train a classifier based on the identification parameters of the simulated current waveform, obtain field test data, use the trained classifier to determine the actual approximate location of partial discharge through the identification parameters of the field test data, and screen out the UHF sensors near the approximate location of partial discharge;

[0022] The positioning module is configured to: determine the signal-to-noise ratio of the UHF signal collected by the selected UHF sensor:

[0023] If the signal-to-noise ratio is higher than the set threshold, the UHF TDOA method is used to locate the discharge source;

[0024] If the signal-to-noise ratio is lower than the set threshold, the wire coil number is calculated based on the above linear regression relationship. At this time, the discharge source is specifically located at the numbered wire coil.

[0025] One or more of the above technical solutions have the following beneficial effects:

[0026] The high-frequency current method is suitable for locating discharges on windings, while the ultra-high frequency method is more suitable for locating local discharges in oil, and the positioning error is large under the obstruction of the windings. Based on this, the present invention provides a method and system for locating the local discharge source of a transformer based on the combination of high-frequency current and ultra-high frequency methods. Combining the advantages and disadvantages of the two methods, the high-frequency current method is used to determine the approximate phase type of the local discharge, and the linear regression relationship between the standard parameters of the current waveform and the coil number is fitted, and then at least 4 ultra-high frequency sensors near the phase where the local discharge is located are selected, and the ultra-high frequency signals of the local discharge are collected using the selected ultra-high frequency sensors, and the following judgment is made: when the local discharge does not occur in the winding, the ultra-high frequency signal signal-to-noise ratio is high at this time, and the use of the ultra-high frequency signal for TDOA positioning has a high accuracy; when the local discharge occurs inside the winding, the ultra-high frequency signal attenuates violently and the signal-to-noise ratio is lower than the set threshold. At this time, the coil number is calculated based on the above-mentioned linear regression relationship, and the discharge source is located at the numbered coil.

[0027] The method of the present invention can narrow the search space range during positioning, reduce the influence of refraction, reflection, diffraction and the like of ultra-high frequency signals during TDOA positioning, and more accurately determine the location of the local discharge source.

[0028] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0030] Figure 1 This is an overall flow chart of Embodiment 1 of the present invention.

[0031] Figure 2 This is a simulation model diagram of a three-phase three-winding transformer according to an embodiment of the present invention.

[0032] Figure 3 This is a waveform diagram of a simulated PD signal 1 according to an embodiment of the present invention.

[0033] Figure 4 This is a waveform diagram of a simulated PD signal 2 according to an embodiment of the present invention.

[0034] Figure 5 This is a waveform diagram of a simulated PD signal 3 according to an embodiment of the present invention.

[0035] Figure 6 The time corresponding to the maximum value of the STFT spectrum of the current at the first end of the phase A high-voltage winding of the first standard parameter of the present invention is A graph that changes monotonically with the pie number.

[0036] Figure 7 The time corresponding to the maximum value of the STFT spectrum of the first-end current of the medium-voltage winding of phase A of the standard parameter of the first embodiment of the present invention is A graph that changes monotonically with the pie number.

[0037] Figure 8 The time corresponding to the maximum value of the STFT spectrum of the current at the first end of the phase A low-voltage winding of the first standard parameter of the present invention is A graph that changes monotonically with the pie number.

[0038] Fig. 9 The time corresponding to the maximum value of the STFT spectrum of the high-voltage winding end (neutral point) current of the first embodiment of the present invention is A graph that changes monotonically with the pie number.

[0039] Fig.10The time corresponding to the maximum value of the STFT spectrum of the medium voltage winding end (neutral point) current of the standard parameter of the first embodiment of the present invention is A graph that changes monotonically with the pie number.

[0040] Fig.11 The time corresponding to the maximum value of the STFT spectrum of the total current at the first end of the standard parameter of the embodiment of the present invention is A graph that changes monotonically with the pie number.

[0041] Fig.12 The time corresponding to the maximum value of the STFT spectrum of the total current at the end of the standard parameter of the first embodiment of the present invention is A graph that changes monotonically with the pie number.

[0042] Fig.13 is the ratio of the total current frequency band energy ratio at the head end to the total current frequency band energy ratio at the end end, R, of the standard parameter of the first embodiment of the present invention SE A graph that changes monotonically with the pie number.

[0043] Fig.14 The figure is a schematic diagram of the division of transformer space by phase and the arrangement of sensors in an embodiment of the present invention.

[0044] In the accompanying drawings, the components represented by the reference numerals are listed as follows:

[0045] 1. Low voltage winding connection, 2. Medium voltage winding connection, 3. High voltage winding connection, 4. Ammeter, 5. Wire coil, 601. High voltage bushing, 602. Medium voltage bushing, 603. Low voltage bushing, 7. Winding coil, 8. Iron yoke, 9. Clamp. DETAILED DESCRIPTION

[0046] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0047] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.

[0048] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0049] Embodiment 1

[0050] This embodiment discloses a method for locating the local discharge source of a transformer based on the combination of high-frequency current and ultra-high frequency method. On the basis of establishing the transformer equivalent circuit model, the current waveforms of the first and last ends of the three-phase windings A, B, and C are simulated, the waveform identification parameters and standard parameters are extracted, and the input matrix composed of the coil number and the standard parameters is linearly fitted to obtain a linear regression relationship. Then, the identification parameters of the high-frequency current waveform measured on site are used to judge that the local discharge roughly occurs in a certain phase area or between two phases, and then a suitable built-in ultra-high frequency sensor is selected according to the discharge area, and the signal-to-noise ratio of the ultra-high frequency signal collected by the selected ultra-high frequency sensor is judged: if the signal-to-noise ratio is higher than the set threshold, TDOA spatial positioning is performed; when the local discharge occurs inside the winding, the ultra-high frequency signal attenuates violently and the signal-to-noise ratio is lower than the set threshold. At this time, the coil number is calculated based on the above-mentioned linear regression relationship, and the discharge source is located at the numbered coil.

[0051] The method proposed in the embodiment of the present invention can narrow the search space range during positioning, reduce the influence of refraction, reflection, diffraction and the like of UHF signals during TDOA positioning, and more accurately determine the location of the local discharge source.

[0052] like Figure 1 As shown, the transformer partial discharge source location method based on the combination of high frequency current and ultra-high frequency method includes the following steps:

[0053] A simulation model of a three-phase three-winding transformer is established, each phase coil is numbered, and a simulation signal is added to a certain numbered coil to simulate the current waveforms at the beginning and end of the high / medium / low voltage side of the three-phase winding;

[0054] Extract and screen the characteristic quantity from the current waveform obtained by simulation to obtain the identification parameters and standard parameters of the simulated current waveform, form the standard parameters into an input matrix, fit the relationship between the wire cake number and the input matrix, and obtain the linear regression relationship;

[0055] A classifier is trained based on the identification parameters of the simulated current waveform, field test data is obtained, and the trained classifier is used to determine the approximate location where the actual partial discharge occurs through the identification parameters of the field test data, and the UHF sensors near the approximate location where the partial discharge occurs are screened out;

[0056] Determine the signal-to-noise ratio of the UHF signal collected by the selected UHF sensor:

[0057] If the signal-to-noise ratio is higher than the set threshold, the UHF TDOA method is used to locate the discharge source;

[0058] If the signal-to-noise ratio is lower than the set threshold, the wire coil number is calculated based on the above linear regression relationship. At this time, the discharge source is specifically located at the numbered wire coil.

[0059] In some embodiments, each phase coil is numbered, and a simulation signal is added at a certain numbered coil, specifically:

[0060] The phase conductors are numbered starting from the bushing side and gradually increasing in number toward the neutral point side;

[0061] At a certain numbered wire cake, a simulation signal is added with a set number of wire cakes as a step interval, and the simulation signal is a simulated PD signal.

[0062] In some embodiments, the current waveforms at the beginning and end of the high / medium / low voltage sides of the three-phase winding are obtained by simulation, wherein the current measured on a certain phase includes:

[0063] High voltage winding first end current , High voltage winding end current , Medium voltage winding first end current , medium voltage winding end current and the low voltage winding first end current ;

[0064] It also includes the total current at the first end calculated based on the measured current. and the total terminal current .

[0065] In some embodiments, feature quantities are extracted and screened from the current waveform obtained by simulation to obtain identification parameters and standard parameters, specifically including:

[0066] From the characteristic quantities extracted from the current waveform, the characteristic quantities capable of distinguishing the phase of the partial discharge are screened and calculated as identification parameters;

[0067] From the characteristic quantities extracted from the current waveform, the characteristic quantities linearly related to the coil number are screened and calculated as standard parameters;

[0068] The characteristic quantities related to the standard parameters include two types: current band energy ratio and current STFT spectrum maximum corresponding time. The standard parameters specifically include: the total current band energy ratio at the head end Ratio of total current band energy to terminal The ratio column vector , the time column vector corresponding to the maximum value of the STFT spectrum of each winding current at the head end , the time column vector corresponding to the maximum value of the STFT spectrum of each winding current at the end , the time column vector corresponding to the maximum value of the STFT spectrum of the total current at the first end , the time column vector corresponding to the maximum value of the STFT spectrum of the total terminal current ;

[0069] The identification parameters include the ratio K of the current energy at the head end of each phase to the current energy at the neutral point, and the ratio M of the maximum amplitude of the total current at the head end to the total current at the end.

[0070] In some embodiments, the standard parameters are combined into an input matrix, and the relationship between the line cake number and the input matrix is ​​fitted to obtain a linear regression equation, which specifically includes:

[0071] The input matrix x for:

[0072] x =[ , , , , , , , ];

[0073] The pie numbers are output as a vector , using the least squares method to fit: , find the weight w and bias b .

[0074] In some embodiments, the classifier is trained based on the identification parameters of the simulated current waveform, specifically:

[0075] Classification of transformer local discharge types ω , the transformer local discharge type ω It includes five types: A-phase discharge, B-phase discharge, C-phase discharge, AB-phase discharge and BC-phase discharge;

[0076] Obtain the identification parameters of the simulated current waveform corresponding to the specific discharge type. The identification parameters of the simulated current waveform are specifically: ψ =[ K Ai , K Bi , K Ci , M ],in K Ai , K Bi , K Ci are the ratios of the current energy at the first end of each phase A, B, and C to the neutral point, respectively, and i represents the coil number of the i-th coil;

[0077] Based on the local discharge type of the transformer and the identification parameters of the simulated current waveform, a training set is constructed. ;

[0078] Based on the training set, the Bayesian classifier is trained to identify that the discharge roughly occurs in a certain phase area or between two phases, and a trained Bayesian classifier is obtained.

[0079] In some embodiments, field test data is obtained, and the actual location of partial discharge is determined by using the trained classifier through identification parameters of the field test data, specifically:

[0080] Extracting current identification parameters from field test data u , the current identification parameters of the field test data u As the sample to be tested;

[0081] The samples to be tested are input into the trained Bayesian classifier, and the class with the largest a posteriori probability is taken as the discharge recognition result, so as to determine the approximate location where the actual partial discharge occurs.

[0082] In some embodiments, selecting the ultra-high frequency sensor near the approximate location where the partial discharge occurs specifically includes:

[0083] The transformer space is divided into three spaces A, B and C according to phases, and at least four UHF sensors are arranged in each phase space including the boundary position;

[0084] When it is determined that the discharge occurs in a certain phase region, that is, when it is determined that the discharge occurs in a certain phase space, at least four UHF sensors in the corresponding phase space are used as UHF sensors near the approximate location where the local discharge occurs;

[0085] When it is determined that the discharge occurs between the AB phase or between the BC phase, at least four UHF sensors in the B phase space are used as UHF sensors near the approximate location where the local discharge occurs.

[0086] In some embodiments: if the signal-to-noise ratio is higher than a set threshold, a UHF TDOA method is used to locate the discharge source, specifically including:

[0087] Among the UHF sensors near the approximate location where the discharge occurs, one is selected as a reference sensor, and the time delay between the other sensors and the reference sensor is calculated using the generalized cross-correlation method;

[0088] Substitute the time delay between other sensors and the reference sensor into the TDOA positioning objective function to solve the positioning result;

[0089] or,

[0090] If the signal-to-noise ratio is lower than the set threshold, the coil number is calculated based on the above linear regression relationship. At this time, the discharge source is specifically located at the numbered coil, which includes:

[0091] Determine the discharge phase according to the approximate location of partial discharge;

[0092] According to the standard parameters of the corresponding phase current calculated by the discharge phase, an input matrix is ​​constructed;

[0093] Based on the input matrix and the linear regression equation, the wire cake number is calculated;

[0094] When local discharge occurs between phases AB or BC, the B phase current is used to calculate the standard parameters.

[0095] Taking into account the advantages and disadvantages of the high-frequency current method and the ultra-high frequency method, the embodiment of the present invention provides a transformer partial discharge source location method based on the combination of high-frequency current and ultra-high frequency method, which effectively improves the partial discharge location accuracy. The specific operation steps are:

[0096] Step 1: Establish a simulation model of a three-phase three-winding transformer with YYD connection, number each phase coil, add different types of simulated PD signals to each numbered coil, and then simulate to obtain the current waveforms of each monitoring point on the high / medium / low voltage side of the three-phase windings A, B, and C.

[0097] Step 2: Extract characteristic quantities from the current waveforms obtained by simulation on each phase and voltage side, and select the characteristic quantities that change monotonically with the increase of the coil number, and record them as standard parameters. All standard parameters form the input matrix , the line number is used as the output vector ,Depend on Perform linear regression to obtain weights and bias b .

[0098] Step 3: Select identification parameters from the characteristic quantities of the current waveform obtained by simulation, and judge that the partial discharge occurs roughly in a certain phase area or between two phases based on the identification parameters of the current waveform obtained by field measurement.

[0099] Step 4: Divide the transformer space into three spaces, A, B, and C, according to phase. Set at least four UHF sensors in each phase space (including the boundary) and perform the following screening:

[0100] When it is determined that the discharge occurs in a certain phase space, at least four UHF sensors in the space are used to collect the UHF signals of the partial discharge;

[0101] When it is determined that the discharge occurs between the AB phase or the BC phase, at least four UHF sensors in the B phase space are used to collect the partial discharge UHF signal.

[0102] Step 5: Determine the signal-to-noise ratio of the UHF signal collected by the selected UHF sensor:

[0103] When the UHF signal-to-noise ratio is higher than the set threshold, the selected UHF sensor is used for TDOA positioning;

[0104] When the UHF signal-to-noise ratio is lower than the set threshold, the standard parameters of the phase current are extracted according to the discharge phase to construct the input matrix , obtained from step 2 Calculate and get the wire cake number. At this time, it is considered that the positioning result is at the numbered wire cake.

[0105] As a preferred method, the method of wire pie numbering and PD simulation in step 1 is:

[0106] The numbers start from the casing side (head end) and gradually increase towards the neutral point side (end). a The PD source is added at step intervals with each coil, and the pulse current under various types of PD sources and discharge positions is monitored.

[0107] Preferably, there are three types of simulated PD signals in step 1, and their expressions are: , , ;

[0108] Among them, y1 is the current amplitude of the simulated PD signal 1; y2 is the current amplitude of the simulated PD signal 2; y3 is the current amplitude of the simulated signal 3; A1, A2, A3 are the maximum current amplitudes of the simulated PD signals 1, 2, 3 respectively; t is the signal time; f1, f2 are the signal oscillation coefficients; τ1, τ2, τ3 are the signal attenuation coefficients.

[0109] Preferably, the current waveforms of the monitoring points of the three-phase windings A, B, and C obtained by simulation in step 1 specifically include:

[0110] For a three-phase three-winding transformer with YYD connection, the currents that can be measured at the first end of the bushing of a phase are: , Medium voltage winding first end current , low voltage winding first end current .

[0111] For high voltage winding and medium voltage winding, due to the Y connection method, the neutral point (end) current of the high voltage winding can be obtained. and the neutral point (end) current of the medium voltage winding .

[0112] Based on the measured current, the total current at the head end of all phases can be calculated Total current at the end (neutral point of Y connection) :

[0113]

[0114]

[0115] in, φ Represents separation, Indicates the sum of the currents monitored on all bushings of phases A, B, and C.

[0116] Preferably, in step 2, characteristic quantities that change monotonically with increasing coil number are selected as standard parameters. There are two types of characteristic quantities related to the standard parameters: (a) current band energy ratio; (b) current STFT spectrum maximum value corresponding time.

[0117] The current band energy ratio is defined as the ratio of the current band energy in the 1MHz ~ 3MHz range to the current band energy in the 0.1MHz ~ 3MHz range. The calculation formula is:

[0118]

[0119] in, Represents the frequency spectrum of the current.

[0120] Time-frequency domain (STFT) maximum and the corresponding time The calculation formula is:

[0121]

[0122] is the current value, n represents the sampling point; is the short-time Fourier transform of the current; argmax is the time corresponding to the maximum current.

[0123] For a certain phase and a certain voltage side, inject PD source into each numbered coil, and the standard parameters that can be obtained are:

[0124] Head end total current band energy ratio Ratio of total current band energy to terminal The ratio column vector , the time column vector corresponding to the maximum value of the STFT spectrum of each winding current at the head end , the time column vector corresponding to the maximum value of the STFT spectrum of each winding current at the end , the time column vector corresponding to the maximum value of the STFT spectrum of the total current at the first end , the time column vector corresponding to the maximum value of the STFT spectrum of the total terminal current .

[0125] Input Matrix x =[ , , , , , , , ];

[0126] In this example, the pie number is used as the output vector:

[0127] =[10,20,30,40,50,60,70,80,90] T ;

[0128] Use the least squares method to fit the curve: , find the weight w and bias b .

[0129] Preferably, the identification parameters in step 3 include two types: the ratio K of the current energy at the head end of each phase to the current energy at the neutral point, and the ratio M of the maximum amplitude of the total current at the head end to the total current at the end.

[0130] The ratio of the current energy at the first end of each phase to the current energy at the neutral point K , the calculation formula is:

[0131]

[0132] in, H / M / L Indicates high / medium / low pressure side; S Indicates the casing side (head end), E Indicates the neutral point side (end); Represents the magnitude of the current; n Represents the sampling point.

[0133] Total current at the first end and the total terminal current The maximum amplitude ratio M is calculated as follows:

[0134]

[0135] Preferably, the method for determining the type of partial discharge according to the identification parameters of the current in step 3 is:

[0136] Partial discharge of transformer windings can be divided into five types: A-phase discharge, B-phase discharge, C-phase discharge, AB-phase discharge and BC-phase discharge. i The PD source is injected at the point, and the simulation calculation obtains the ratio of the current energy at the first end of each phase A, B, and C to the neutral point. K Ai , K Bi ,K Ci , and the total current at the first end and the total terminal current The maximum amplitude ratio M is recorded as the current identification parameter obtained by this discharge simulation ψ =[ K Ai , K Bi , K Ci , M ]. The current identification parameter is used as a training sample, and its corresponding discharge type ω is the label of the training sample. Among them, the A phase discharge, B phase discharge, C phase discharge, AB phase discharge and BC phase discharge ω They are recorded as 1, 2, 3, 4, and 5 respectively.

[0137] According to the five conditions of A phase discharge, B phase discharge, C phase discharge, AB phase discharge and BC phase discharge, different coil numbers are i Inject PD source at each discharge point to obtain various discharge types ω The corresponding large number of training samples ψ , and then construct the training set: .

[0138] The Bayesian classifier is trained based on the training set to obtain a trained Bayesian classifier.

[0139] The current identification parameters extracted from the field measured data u As the sample to be tested, Bayesian estimation is performed on the sample to be tested based on the trained Bayesian classifier, and the class with the largest a posteriori probability is used as the discharge phase difference result of the final judgment.

[0140] The specific implementation steps of the Naive Bayes method are:

[0141] Step 3.1: Calculate prior probability

[0142] From the training set, count a certain discharge type ω i ( i =1,2,3,4,5, indicating the number of occurrences of the i-th discharge type) n i And the total number of training set samples N , and obtain the prior probability of each discharge type :

[0143]

[0144] Step 3.2:

[0145] ① Find each type of discharge ω i The mean of the training samples :

[0146]

[0147] in k Represents the feature number, j Represents the sample serial number; is the kth current identification parameter value of the jth training sample.

[0148] ② Find each type of training sample The covariance matrix of , and the covariance matrix The inverse matrix and the determinant | |;

[0149] ③ Assume that the local discharge current identification parameter measured on site is u， Identify the parameters u As the sample to be tested, find:

[0150] ;

[0151] is the natural logarithm of the conditional probability under each type of discharge.

[0152] Step 3.3: Determine the discriminant function

[0153] Assume that the local discharge current identification parameter measured on site is u Belongs to the discharge category ω i The discriminant function for:

[0154]

[0155] In the formula, Indicates the current identification parameter to be measured u Belongs to the discharge category ω i The posterior probability of is taken as the natural logarithm.

[0156] Step 3.4: G =[ , , , , Find the largest element in , the type of partial discharge can be determined ω = ibest , ibest=1 or 2 or 3 or 4 or 5, where 1, 2, 3, 4, and 5 represent phase A discharge, phase B discharge, phase C discharge, AB phase discharge, and BC phase discharge, respectively.

[0157] Preferably, the TDOA positioning method in step 5 is:

[0158] First, collect at least 4 UHF sensor signals, use one of the UHF sensors as the reference sensor, and use the generalized cross-correlation method to calculate the time delay between other sensors and the reference sensor. The generalized cross-correlation function is:

[0159]

[0160] in, represents the reference sensor signal, represents another UHF sensor signal, represents the displacement, m represents the sampling point; represent and displacement After size The generalized cross-correlation value of .

[0161] Seek difference Under value Value, please The maximum corresponding Signal and Delay .

[0162] Substitute the time delay between each sensor and the reference sensor into the following TDOA positioning objective function to obtain the positioning result:

[0163]

[0164] in, N is the number of UHF sensors; c is the propagation speed of UHF electromagnetic waves; It is the time when the reference sensor receives the UHF signal generated by partial discharge; For the The arrival time difference between the UHF signal generated by the partial discharge received by the sensor and the reference sensor; For the The spatial coordinate position of each sensor; is the spatial coordinate position of the local discharge source.

[0165] This embodiment takes the installation of eight UHF sensors inside a true-type transformer with a size of 148 cm×72 cm×100 cm as an example to specifically explain the implementation method of the embodiment of the present invention.

[0166] See also Figure 1 This embodiment provides a method for locating the local discharge source of a transformer based on the combination of high-frequency current and ultra-high frequency methods. First, measure the current waveforms at the beginning and end of the three-phase windings A, B, and C, extract the waveform identification parameters and standard parameters, perform linear fitting on the coil numbers and standard parameters, and determine whether the discharge occurs in a certain phase area or between two phases based on the identification parameters. Then, select a suitable built-in ultra-high frequency sensor for TDOA spatial positioning based on the discharge area. When the discharge occurs inside the winding, the ultra-high frequency signal attenuates sharply, and at this time, the high-frequency current electrical positioning method is mainly used.

[0167] The specific operation steps are:

[0168] Step 1: Establish a three-phase three-winding transformer equivalent circuit simulation model, such as Figure 2 shown.

[0169] exist Figure 2 In the figure, the low-voltage winding connection 1 adopts the Δ connection method, and a, b, and c are used to represent the three-phase low-voltage winding respectively; the medium-voltage winding connection 2 adopts the Y connection method, and Am, Bm, and Cm are used to represent the three-phase medium-voltage winding respectively; the high-voltage winding connection 3 adopts the Y connection method, and A, B, and C are used to represent the three-phase high-voltage winding respectively. The ammeter 4 is connected to the bushing head end of each phase and each voltage side, and the Y connection neutral point of the high-voltage side and the medium-voltage side to monitor the high-frequency current. In the figure, as an example, only three coils 5 are drawn on each phase. The bushings are high-voltage bushing 601, medium-voltage bushing 602, and low-voltage bushing 603 in order from high to low.

[0170] Each phase wire coil is numbered, starting from the bushing side (head end) and gradually increasing toward the neutral point side (end), with a step interval of 10 coils. PD sources are added at coils 10 to 90. The waveforms of various types of signal sources are as follows: Figure 3 , Figure 4 and Figure 5 Then the current waveforms of the bushing side (head end) and the neutral point side (end) of the high / medium / low voltage sides of the three-phase windings A, B, and C are measured.

[0171] It can be understood that the meaning of the coil in this embodiment is: the transformer winding is in a pancake structure, the turns are continuously wound along its radial direction to form a coil, and multiple coils are arranged along the axial direction to form the winding of the transformer.

[0172] Step 2: Extract characteristic quantities from the current waveforms of each phase and voltage side, and select the characteristic quantities that change monotonically with the increase of the wire coil number, and record them as standard parameters. The standard parameters include two categories: (a) current band energy ratio; (b) current STFT spectrum maximum corresponding time. The curve diagram of each standard parameter changing monotonically with the wire coil number is shown in the figure below. Figure 6-Figure 13 shown.

[0173] exist Figure 6 The figure shows the time corresponding to the maximum value of the STFT spectrum of the current at the first end of the A-phase high-voltage winding. A curve that changes monotonically with the number of the wire cake;

[0174] exist Figure 7 The figure shows the time corresponding to the maximum value of the STFT spectrum of the current at the first end of the medium voltage winding of phase A. A curve that changes monotonically with the number of the wire cake;

[0175] exist Figure 8 The figure shows the time corresponding to the maximum value of the STFT spectrum of the current at the beginning of the low-voltage winding of phase A. A curve that changes monotonically with the number of the wire cake;

[0176] exist Fig. 9 The time corresponding to the maximum value of the STFT spectrum of the current at the end of the high-voltage winding (neutral point) is shown in A curve that changes monotonically with the number of the wire cake;

[0177] exist Fig.10 The time corresponding to the maximum value of the STFT spectrum of the medium voltage winding end (neutral point) current is shown in A curve that changes monotonically with the number of the wire cake;

[0178] exist Fig.11 The time corresponding to the maximum value of the STFT spectrum of the total current at the first end is shown in A curve that changes monotonically with the number of the wire cake;

[0179] exist Fig.12 The time corresponding to the maximum value of the STFT spectrum of the total terminal current is shown in A curve that changes monotonically with the number of the wire cake;

[0180] exist Fig.13 The figure shows the ratio of the total current-band energy ratio at the head end to the total current-band energy ratio at the end. SE A curve that changes monotonically with the pie number.

[0181] The current band energy ratio is defined as the ratio of the current band energy in the 1MHz ~ 3MHz band to the current band energy in the 0.1MHz ~ 3MHz band. The calculation formula is:

[0182]

[0183] in, Represents the frequency spectrum of the current.

[0184] Time-frequency domain (STFT) maximum and the corresponding time The calculation formula is:

[0185]

[0186] is the current value; is the short-time Fourier transform of the current; argmax is the time corresponding to the maximum current.

[0187] For a three-phase three-winding transformer with YYD connection, the currents that can be measured at the first end of the bushing of a phase are: , Medium voltage winding first end current , low voltage winding first end current .

[0188] For high voltage winding and medium voltage winding, due to the Y connection method, the neutral point (end) current of the high voltage winding can also be obtained. and the neutral point (end) current of the medium voltage winding .

[0189] Based on the measured current, the total current at the head end of all phases can be calculated Total current at the end (neutral point of Y connection) :

[0190]

[0191]

[0192] in, φ Represents separation, Indicates the sum of the currents monitored on all bushing sides of phases A, B, and C.

[0193] For a certain phase and a certain voltage side, inject the PD source into each coil number, and the standard parameters that can be obtained are: the total current band energy ratio of the head end Ratio of total current band energy to terminal The ratio column vector , the time column vector corresponding to the maximum value of the STFT spectrum of each winding current at the head end , the time column vector corresponding to the maximum value of the STFT spectrum of each winding current at the end , the time column vector corresponding to the maximum value of the STFT spectrum of the total current at the first end , the time column vector corresponding to the maximum value of the STFT spectrum of the total terminal current .

[0194] All standard parameters form the input matrix: x =[ , , , , , , , ];

[0195] Piece numbers as output vector: =[10, 20, 30, 40, 50, 60, 70, 80, 90] T ;

[0196] Use the least squares method to fit the curve: , find the weight w and bias b .

[0197] Step 3: Select identification parameters from the characteristic quantities of the current waveform. There are two types of identification parameters. Among them, the ratio of the current energy at the head end and the neutral point on a certain phase K , the calculation formula is:

[0198]

[0199] In the formula, H / M / L Indicates high / medium / low pressure side; S Indicates the casing side (head end), E Indicates the neutral point side (end); Represents the magnitude of the current; n Represents the sampling point.

[0200] Total current at the beginning of all phases and the total terminal current The maximum amplitude ratio M is calculated as follows:

[0201]

[0202] According to the current identification parameters, it is judged that the partial discharge occurs roughly in a certain phase area or between two phases. The specific method is:

[0203] According to the five conditions of A phase discharge, B phase discharge, C phase discharge, AB phase discharge and BC phase discharge, different coil numbers are i When different types of PD sources are injected, the ratio of the current energy at the first end of each phase to the neutral point is calculated. K Ai , K Bi , K Ci , and the total current at the first end and the total terminal current Maximum amplitude ratio M, record the current identification parameters of this discharge ψ =[ K Ai , K Bi , K Ci , M ], the corresponding discharge type is ω . Among them, the discharge of phase A, phase B, phase C, AB phase discharge and BC phase discharge ω They are recorded as 1, 2, 3, 4, and 5 respectively.

[0204] The simulation results under different conditions are used as training sets: ,The naive Bayes method is used to train the classifier.

[0205] Step 4: Divide the transformer space into three spaces A, B, and C according to phase. There are at least 4 UHF sensors in each phase space (including the boundary), for a total of 8 UHF sensors. Fig.14 shown.

[0206] Fig.14 The figure shows the transformer space divided by phase and the sensor arrangement from a top view, showing the winding coil 7, iron yoke 8 and clamp 9 inside the transformer. Fig.14 In the figure, the internal space of the transformer is divided into A-phase area, B-phase area and C-phase area according to the phases. Among them, S1 and S2 are respectively installed on the front and back areas of the box wall facing the AB winding phases, S3 and S4 are respectively installed on the front and back areas of the box wall facing the BC winding phases, and S5, S6 and S7, S8 are respectively installed on the box walls on both sides of the transformer.

[0207] Take one corner of the transformer as the coordinate origin, and the inner length, inner width and inner height of the transformer as x , y , z The positive direction of the axis, the coordinates of the UHF sensor are:

[0208] S 1(49.3, 72, 60), S 2(49.3, 0, 60), S 3(91.8, 72, 50), S 4(91.8, 0, 50), S 5(0,48, 50) , S 6(0, 24, 50), S 7(148, 48, 50), S 8(148, 24, 50).

[0209] Using air gap defects as a source of partial discharge P (70, 10, 90)cm, current identification parameters extracted from field measured data u Bayesian estimation is performed, and the class with the largest a posteriori probability is taken as the final judgment of the discharge phase difference result.

[0210] When it is determined that the discharge occurs in the B phase space, the four UHF sensors in the space are used to collect signals for TDOA positioning. S 1 is used as the reference sensor, and the generalized cross-correlation method is used to calculate the time delay between the other three sensors and the reference sensor. , , Substitute the time delay between each sensor and the reference sensor into the following TDOA positioning objective function to solve the positioning result:

[0211]

[0212] in, c is the propagation speed of UHF electromagnetic waves, which is taken as 0.2; It is the time when the reference sensor receives the UHF signal generated by partial discharge; For the The arrival time difference between the UHF signal generated by the partial discharge received by the sensor and the reference sensor; For the The spatial coordinate position of each sensor; is the spatial coordinate position of the local discharge source.

[0213] The particle swarm algorithm is used to solve the problem, and the final positioning result is (70.485881, 15.798885, 75.014411) cm. It can be seen that the field verification result is close to the actual defect position, and the positioning error is small, which is comparable to the laboratory measurement result.

[0214] Embodiment 2

[0215] This embodiment discloses a transformer partial discharge source location system based on a combination of high frequency current and ultra-high frequency methods.

[0216] The transformer partial discharge source location system based on the combination of high frequency current and ultra-high frequency method includes:

[0217] The simulation module is configured to: establish a simulation model of a three-phase three-winding transformer, number each phase coil, add a simulation signal at a certain numbered coil, and simulate the current waveforms at the first and last ends of the high / medium / low voltage sides of the three-phase winding;

[0218] The fitting module is configured to: extract and screen the characteristic quantity from the current waveform obtained by simulation, obtain the identification parameter and the standard parameter of the simulated current waveform, form the standard parameter into an input matrix, fit the relationship between the wire cake number and the input matrix, and obtain the linear regression relationship;

[0219] The approximate location identification module is configured to: train a classifier based on the identification parameters of the simulated current waveform, obtain field test data, use the trained classifier to determine the actual approximate location of partial discharge through the identification parameters of the field test data, screen out UHF sensors near the approximate location of partial discharge, and determine the signal-to-noise ratio of UHF signals collected by the screened UHF sensors;

[0220] The judgment and positioning module is configured to: if the signal-to-noise ratio is higher than a set threshold, use the ultra-high frequency TDOA method to specifically locate the discharge source;

[0221] If the signal-to-noise ratio is lower than the set threshold, the wire coil number is calculated based on the above linear regression relationship. At this time, the discharge source is specifically located at the numbered wire coil.

[0222] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0223] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A method for locating the local discharge source of a transformer based on a combination of high frequency current and ultra-high frequency method, characterized in that: The following steps are involved: A simulation model of a three-phase three-winding transformer is established, each phase coil is numbered, and a simulation signal is added to a certain numbered coil to simulate the current waveforms at the beginning and end of the high / medium / low voltage side of the three-phase winding; Extract and screen the characteristic quantity from the current waveform obtained by simulation to obtain the identification parameters and standard parameters of the simulated current waveform, form the standard parameters into an input matrix, fit the relationship between the wire cake number and the input matrix, and obtain the linear regression relationship; A classifier is trained based on the identification parameters of the simulated current waveform, field test data is obtained, and the trained classifier is used to determine the approximate location where the actual partial discharge occurs through the identification parameters of the field test data, and the UHF sensors near the approximate location where the partial discharge occurs are screened out; Determine the signal-to-noise ratio of the UHF signal collected by the selected UHF sensor: If the signal-to-noise ratio is higher than the set threshold, the UHF TDOA method is used to locate the discharge source; If the signal-to-noise ratio is lower than the set threshold, the wire coil number is calculated based on the above linear regression relationship. At this time, the discharge source is specifically located at the numbered wire coil.

2. The transformer partial discharge source location method based on the combination of high frequency current and ultra-high frequency method according to claim 1 is characterized in that: Each phase coil is numbered and a simulation signal is added to a certain numbered coil. Specifically: The phase conductors are numbered starting from the bushing side and gradually increasing in number toward the neutral point side; At a certain numbered wire cake, a simulation signal is added with a set number of wire cakes as a step interval, and the simulation signal is a simulated PD signal.

3. The transformer partial discharge source location method based on the combination of high frequency current and ultra-high frequency method as claimed in claim 1 is characterized in that: The simulation obtains the current waveforms at the beginning and end of the high / medium / low voltage sides of the three-phase winding, where the current measured on a certain phase includes: High voltage winding first end current , High voltage winding end current , Medium voltage winding first end current , medium voltage winding end current and the low voltage winding first end current ; It also includes the total current at the first end calculated based on the measured current. and the total terminal current .

4. The transformer partial discharge source location method based on the combination of high frequency current and ultra-high frequency method as claimed in claim 3 is characterized in that: Extract and filter the characteristic quantities from the simulated current waveform to obtain identification parameters and standard parameters, including: From the feature quantities extracted from the current waveform, a feature quantity capable of distinguishing the phase of the partial discharge is selected and calculated as an identification parameter; From the characteristic quantities extracted from the current waveform, the characteristic quantities linearly related to the coil number are screened and calculated as standard parameters; The characteristic quantities related to the standard parameters include two types: current band energy ratio and current STFT spectrum maximum corresponding time. The standard parameters specifically include: the total current band energy ratio at the head end Ratio of total current band energy to terminal The ratio column vector , the time column vector corresponding to the maximum value of the STFT spectrum of each winding current at the first end , the time column vector corresponding to the maximum value of the STFT spectrum of each winding current at the end , the time column vector corresponding to the maximum value of the STFT spectrum of the total current at the first end , the time column vector corresponding to the maximum value of the STFT spectrum of the total terminal current ; The identification parameters include the ratio K of the current energy at the head end of each phase to the current energy at the neutral point, and the ratio M of the maximum amplitude of the total current at the head end to the total current at the end.

5. The transformer partial discharge source location method based on the combination of high frequency current and ultra-high frequency method as claimed in claim 4 is characterized in that: The standard parameters are combined into an input matrix, and the relationship between the fitting line number and the input matrix is ​​obtained to obtain the linear regression equation, which includes: The input matrix x for: x =[ , , , , , , , ]； The pie numbers are output as a vector , using the least squares method to fit: , find the weight w and bias b .

6. The transformer partial discharge source location method based on the combination of high frequency current and ultra-high frequency method as claimed in claim 4, characterized in that: The classifier is trained based on the identification parameters of the simulated current waveform, specifically: Classification of transformer local discharge types ω , the transformer local discharge type ω It includes five types: A-phase discharge, B-phase discharge, C-phase discharge, AB-phase discharge and BC-phase discharge; Obtain the identification parameters of the simulated current waveform corresponding to the specific discharge type. The identification parameters of the simulated current waveform are specifically: ψ =[ K Ai , K Bi , K Ci , M ],in K Ai , K Bi , K Ci are the ratios of the current energy at the first end of each phase A, B, and C to the neutral point, respectively, and i represents the coil number of the i-th coil; Based on the local discharge type of the transformer and the identification parameters of the simulated current waveform, a training set is constructed. ; Based on the training set, the Bayesian classifier is trained to identify that the discharge roughly occurs in a certain phase area or between two phases, and a trained Bayesian classifier is obtained.

7. The transformer partial discharge source location method based on the combination of high frequency current and ultra-high frequency method as claimed in claim 6, characterized in that: Obtain field test data and use the trained classifier to determine the approximate location of actual partial discharge based on the identification parameters of the field test data. Specifically: Extracting current identification parameters from field test data u , the current identification parameters of the field test data u As the sample to be tested; The samples to be tested are input into the trained Bayesian classifier, and the class with the largest a posteriori probability is taken as the discharge recognition result, so as to determine the approximate location where the actual partial discharge occurs.

8. The transformer partial discharge source location method based on the combination of high frequency current and ultra-high frequency method as claimed in claim 6, characterized in that: Screen out the UHF sensors near the approximate location where partial discharge occurs, including: The transformer space is divided into three spaces A, B and C according to phases, and at least four UHF sensors are arranged in each phase space including the boundary position; When it is determined that the discharge occurs in a certain phase region, that is, when it is determined that the discharge occurs in a certain phase space, at least four UHF sensors in the corresponding phase space are used as UHF sensors near the approximate location where the local discharge occurs; When it is determined that the discharge occurs between the AB phase or between the BC phase, at least four UHF sensors in the B phase space are used as UHF sensors near the approximate location where the local discharge occurs.

9. The transformer partial discharge source location method based on the combination of high frequency current and ultra-high frequency method as claimed in claim 1, characterized in that: If the signal-to-noise ratio is higher than the set threshold, the UHF TDOA method is used to locate the discharge source, including: Among the UHF sensors near the approximate location where the discharge occurs, one is selected as a reference sensor, and the time delay between the other sensors and the reference sensor is calculated using the generalized cross-correlation method; Substitute the time delay between other sensors and the reference sensor into the TDOA positioning objective function to solve the positioning result; or, If the signal-to-noise ratio is lower than the set threshold, the coil number is calculated based on the above linear regression relationship. At this time, the discharge source is specifically located at the numbered coil, which includes: Determine the discharge phase according to the approximate location of partial discharge; According to the standard parameters of the corresponding phase current calculated by the discharge phase, an input matrix is ​​constructed; Based on the input matrix and the linear regression equation, the wire cake number is calculated; When local discharge occurs between phases AB or BC, the B phase current is used to calculate the standard parameters.

10. A transformer partial discharge source location system based on a combination of high frequency current and ultra-high frequency method, characterized in that: include: The simulation module is configured to: establish a simulation model of a three-phase three-winding transformer, number each phase coil, add a simulation signal at a certain numbered coil, and simulate the current waveforms at the first and last ends of the high / medium / low voltage sides of the three-phase winding; The fitting module is configured to: extract and screen the characteristic quantity from the current waveform obtained by simulation, obtain the identification parameter and the standard parameter of the simulated current waveform, form the standard parameter into an input matrix, fit the relationship between the wire cake number and the input matrix, and obtain the linear regression relationship; The approximate location identification module is configured to: train a classifier based on the identification parameters of the simulated current waveform, obtain field test data, use the trained classifier to determine the actual approximate location of partial discharge through the identification parameters of the field test data, and screen out the UHF sensors near the approximate location of partial discharge; The positioning module is configured to: determine the signal-to-noise ratio of the UHF signal collected by the selected UHF sensor: If the signal-to-noise ratio is higher than the set threshold, the UHF TDOA method is used to locate the discharge source; If the signal-to-noise ratio is lower than the set threshold, the wire coil number is calculated based on the above linear regression relationship. At this time, the discharge source is specifically located at the numbered wire coil.

Images