Transformer winding fault diagnosis method, terminal and storage medium

Abstract

The present invention belongs to the technical field of power equipment maintenance, and specifically relates to a transformer winding fault diagnosis method, terminal, and storage medium. The method comprises obtaining frequency response curves of the transformer winding in normal and abnormal states through a frequency sweep test. During the test, the secondary winding is short-circuited and grounded, the primary winding's head end is connected to a frequency sweep signal, and the tail end is left floating. The number of units in an equivalent network is determined based on the number of characteristic frequencies of the normal winding's frequency response curve, and an equivalent network of a normal winding composed of multiple RLC units is reconstructed. By constructing a high-precision equivalent network and performing frequency response residual analysis, combined with the Pearson correlation coefficient to screen highly correlated components, the method can accurately locate the specific location of winding deformation or faults and distinguish the fault type, thus completely overcoming the limitation of traditional frequency response methods that can only determine the overall degree of deformation.

Description

Technical Field

[0001] The present invention belongs to the technical field of power equipment maintenance, and in particular relates to a transformer winding fault diagnosis method, a terminal and a storage medium. Background Art

[0002] Transformers are critical equipment in power systems, and their safe and stable operation is crucial. However, transformer winding faults are one of the main factors affecting transformer reliability, potentially leading to performance degradation or even damage, threatening the safe and stable operation of power systems. Therefore, accurate and efficient diagnosis of transformer winding faults is of great significance.

[0003] Frequency response analysis (FRA) is currently widely used in power systems to diagnose winding deformation. This method applies a swept sinusoidal signal across a specific frequency range to the winding and measures the steady-state sinusoidal response at the winding end to obtain the winding's frequency response curve. By comparing the current FRA curve with the factory-installed FRA curve, a preliminary assessment of winding deformation can be made. However, traditional FRA methods can only diagnose the overall severity of winding deformation; they cannot precisely determine the exact location of the deformation, the fault type, or the severity of the deformation at each location. Summary of the Invention

[0004] In view of the defect in the prior art that the traditional frequency response analysis method can only preliminarily determine whether the winding is deformed, but cannot accurately obtain the exact location of the winding deformation, the fault type and the severity of different parts, the present invention provides a transformer winding fault diagnosis method, terminal and storage medium to solve the above technical problems.

[0005] In a first aspect, the present invention provides a transformer winding fault diagnosis method, comprising:

[0006] Step S1: Obtain frequency response curves of the transformer winding under normal and abnormal conditions through a frequency sweep test, wherein during the test, the secondary winding is short-circuited and grounded, the first end of the primary winding is connected to a frequency sweep signal and the end is left floating, and the test frequency range is 1 kHz to 2 MHz;

[0007] Step S2: determining the number of units of the equivalent network according to the number of characteristic frequencies of the frequency response curve of the normal state winding, and reconstructing the normal winding equivalent network composed of multiple RLC units;

[0008] Step S3: Calculate the voltage and current parameters of all branches based on the reconstructed equivalent network, and construct a Jacobian matrix for characterizing the sensitivity of the frequency response function to changes in network parameters;

[0009] Step S4: Calculate the frequency response residuals of the normal frequency response curve and the abnormal frequency response curve, decompose the Jacobian matrix and the frequency response residual respectively, and reconstruct them respectively to obtain the corresponding reconstructed real frequency response residual and real Jacobian matrix, and construct a real relationship equation between the reconstructed real Jacobian matrix and the real frequency response residual;

[0010] Step S5: Analyze the Pearson correlation coefficient between each column vector of the reconstructed real Jacobian matrix and the real frequency response residual, sort the column vectors from high to low according to the correlation, and record the corresponding network element numbers;

[0011] Step S6: Select the network components corresponding to the column vectors with the highest correlation, construct a least squares equation to solve the parameter correction value of the faulty component, update the equivalent network parameters and recalculate the frequency response curve;

[0012] Step S7, iteratively execute steps S3 to S6 until the residual between the updated frequency response curve and the measured abnormal curve is less than a preset first residual threshold, and determine the fault location, type and severity based on the final corrected network component number and parameter changes.

[0013] A further improvement of this technical solution is that step S1 specifically includes:

[0014] Step S11: The test frequency range is 1kHz to 2MHz. The sweep frequency signal is generated by a signal generator and input through the first end of the primary winding, with the end remaining suspended.

[0015] Step S12: The secondary winding of the transformer under test is grounded via a short-circuit wire to eliminate interference of the secondary side distributed parameters on the frequency response curve;

[0016] Step S13: Calculate the frequency response function by measuring the spectrum data of the primary side head end current and terminal voltage , the calculation formula is:

[0017] ;

[0018] in, is the frequency response function, which characterizes the impedance characteristics of the transformer winding under high-frequency signals; is the voltage amplitude at the end of the primary winding of the transformer at frequency f, in volts; is the current amplitude at the beginning of the primary winding of the transformer at frequency f, in amperes; Is the phase factor, reflecting the phase difference between voltage and current ; is an imaginary unit, that is, , characterizing the phase shift characteristics of inductance and capacitance.

[0019] A further improvement of this technical solution is that step S2 includes:

[0020] Step S21: jointly analyze the amplitude-frequency characteristics and phase-frequency characteristics of the normal state frequency response curve, and calculate the total number N of characteristic frequencies by extreme point detection and inflection point identification; extreme points include peak values ​​and valley values;

[0021] Step S22: Based on the total number of characteristic frequencies N, the mutual inductance effect in the high frequency band is compensated according to a preset ratio, and the number of equivalent network units n of the winding is dynamically determined; the calculation formula is: ,in, To round down;

[0022] Step S23: constructing a winding equivalent network of a chain series structure using the RLC units with the determined number of units;

[0023] Step S24: Based on a pre-stored least squares optimization algorithm and taking the measured frequency response curve as a benchmark, the resistance, inductance, and capacitance parameters of each unit are iteratively adjusted so that the frequency response residual between the frequency response curve of the winding equivalent network and the measured frequency response curve converges to a preset second residual threshold. The objective function of the pre-stored least squares optimization algorithm is: ;in, The frequency response function measured by the frequency sweep test is used to characterize the frequency response characteristics of the winding under normal conditions; The frequency of the transformer primary winding end in the equivalent network is The voltage amplitude at , in volts; The frequency of the transformer primary winding head in the equivalent network is The current amplitude at , in amperes; Is the phase factor, reflecting the phase difference between voltage and current in the equivalent network .

[0024] A further improvement of this technical solution is that step S3 includes:

[0025] Step S31: Based on the reconstructed equivalent network, the voltage and current values ​​of the RLC elements in all branches are solved by circuit analysis;

[0026] Step S32: For each RLC element, calculate the sensitivity of the frequency response function to the parameter based on its voltage or current value, where the sensitivity of the inductance parameter is characterized by the product of the square of its branch current value and the angular frequency, the sensitivity of the capacitance parameter is characterized by the product of the square of its branch voltage value and the angular frequency, the sensitivity of the resistance parameter is directly characterized by the square of its branch current value, and the sensitivity of the mutual inductance parameter is characterized by the product of the current product of the associated inductance branch and the angular frequency;

[0027] Step S33: Fill all sensitivity values ​​into the matrix according to the pre-stored test frequency points and parameter indexes to form a Jacobian matrix A;

[0028] Step S34: compare the measured sensitivity with the Jacobian matrix elements, and when the deviation between the measured sensitivity and the matrix elements exceeds a preset deviation threshold, re-execute steps S31 to S33 until the deviation converges.

[0029] A further improvement of this technical solution is that step S4 includes:

[0030] Step S41: Calculate the frequency response residuals of the normal frequency response curve and the abnormal frequency response curve, and decompose the real and imaginary parts of the frequency response residuals. Then, vertically concatenate the real and imaginary parts of the frequency response residuals to obtain a real residual vector ;

[0031] Step S42: Decompose the Jacobian matrix A into real and imaginary parts, and vertically concatenate the real and imaginary parts of the Jacobian matrix A to obtain a reconstructed real Jacobian matrix ;

[0032] Step S43: reconstruct the real Jacobian matrix and the real residual vector Make the association and construct its real number relationship equation: ,in, is the network parameter correction value vector, which is used to represent the parameter change caused by the fault.

[0033] A further improvement of this technical solution is that step S5 includes:

[0034] Step S51: Calculate the real Jacobian matrix according to the Pearson correlation coefficient PCC formula Each column vector of and real frequency response residual The correlation, ; The Pearson correlation coefficient formula is: ,in, is the test frequency number; is a column vector The value of the i-th element in ; is the real frequency response residual The value of the i-th element in ;

[0035] Step S52: Real Jacobian matrix All column vectors of traverse the operation, and substitute each column vector into the Pearson correlation coefficient formula to calculate its difference with the real frequency response residual During the calculation process, record the correlation coefficient value corresponding to each column vector and the column vector in the real Jacobian matrix The column index g in the transformer winding equivalent network is mapped to the transformer winding equivalent network parameter number, that is, each column index g corresponds to one or a group of component parameters in the transformer winding equivalent network.

[0036] Step S53: According to the calculated correlation coefficient value, the real Jacobian matrix is ​​sorted in descending order. All column vectors of are sorted; while sorting, the transformer winding equivalent network parameter number information corresponding to each column vector is output in sequence according to the recorded column index and the mapping relationship between the column index and the transformer winding equivalent network parameter number to form a parameter number sequence.

[0037] A further improvement of this technical solution is that step S6 includes:

[0038] Step S61: Based on the sorting results of the Pearson correlation coefficient, select The top 5% real Jacobian matrix of the correlation Column vector, records the corresponding network element number and forms the fault element number set S;

[0039] Step S62: According to the real Jacobian matrix Extract the column vector from the matrix and construct the Jacobian matrix subset ;

[0040] Step S63: Correct the network parameter value vector Extract the elements corresponding to the fault component number set S to form the network parameter correction value sub-vector ;

[0041] Step S64: Based on the Jacobian matrix subset and the network parameter correction value subvector Construct the least squares equation: ;

[0042] Step S65: Solve the constructed least squares equation , get the corrected values ​​of network component parameters , and its calculation formula is: ,in, is a subset of the Jacobian matrix The transposed matrix of

[0043] Step S66: Correction values ​​of network element parameters obtained based on the solution The parameters of the winding equivalent network are updated; and the frequency response curve is recalculated based on the updated network parameters.

[0044] A further improvement of this technical solution is that step S7 includes:

[0045] After the iteration is terminated, the final corrected network component number and parameter changes are combined with a pre-stored fault signature library to determine the location, type, and severity of the fault. The fault signature library contains the fault type and location corresponding to different component numbers, as well as the corresponding relationship between parameter changes and fault severity. The fault type and severity are determined based on the parameter changes of the network components. Fault types include short circuit, open circuit, and parameter drift, and severity includes mild fault, moderate fault, and severe fault.

[0046] In a second aspect, the present invention provides a terminal, comprising:

[0047] processor, memory, wherein

[0048] The memory is used to store computer programs,

[0049] The processor is used to call and run the computer program from the memory, so that the terminal executes the above-mentioned terminal method.

[0050] In a third aspect, the present invention provides a computer storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the methods described in the above aspects.

[0051] The beneficial effects of the present invention are:

[0052] Accurately locate fault location and type, breaking through the limitations of traditional methods: By constructing a high-precision equivalent network and performing frequency response residual analysis, combined with the Pearson correlation coefficient to screen highly correlated components, this method can accurately locate the specific location of winding deformation or fault and distinguish the fault type, completely overcoming the limitation of traditional frequency response methods that can only determine the overall degree of deformation.

[0053] Dynamically optimize the equivalent network to improve matching accuracy: Dynamically determine the number of units based on the characteristic frequency to compensate for high-frequency mutual inductance effects, avoid overfitting or underfitting, and ensure that the equivalent network covers the electromagnetic characteristics of the windings across the entire frequency range. Iteratively optimize parameters using a least-squares algorithm to converge the residual error between the network frequency response and the measured curve to a threshold, improving the equivalent network matching accuracy and laying a solid foundation for subsequent fault analysis.

[0054] This method calculates parameter sensitivity based on branch voltage and current, accurately quantifying the sensitivity of the frequency response to these parameters and avoiding errors based on human experience. By statistically ranking components based on their relevance (e.g., ranking the top 5%), it eliminates interference from irrelevant parameters, narrows fault location to key components, and improves computational efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0056] Figure 1 A schematic flow chart of a method according to an embodiment of the present invention.

[0057] Figure 2 Schematic diagram of a winding equivalent network according to an embodiment of the present invention.

[0058] Figure 3 A schematic diagram of the structure of a terminal provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0059] In order to make the purpose, features, and advantages of the present invention more obvious and easy to understand, the technical solutions of the present invention will be clearly and completely described below in conjunction with the drawings in the specific embodiments. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0061] Figure 1 The following is a schematic flow chart of a transformer winding fault diagnosis method provided by the present invention. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted.

[0062] like Figure 1 As shown, the method includes:

[0063] Step S1: Obtain frequency response curves of the transformer winding under normal and abnormal conditions through a frequency sweep test, wherein during the test, the secondary winding is short-circuited and grounded, the first end of the primary winding is connected to a frequency sweep signal and the end is left floating, and the test frequency range is 1 kHz to 2 MHz;

[0064] Step S2: determining the number of units of the equivalent network according to the number of characteristic frequencies of the frequency response curve of the normal state winding, and reconstructing the normal winding equivalent network composed of multiple RLC units;

[0065] Step S3: Calculate the voltage and current parameters of all branches based on the reconstructed equivalent network, and construct a Jacobian matrix for characterizing the sensitivity of the frequency response function to changes in network parameters;

[0066] Step S4: Calculate the frequency response residuals of the normal frequency response curve and the abnormal frequency response curve, decompose the Jacobian matrix and the frequency response residual respectively, and reconstruct them respectively to obtain the corresponding reconstructed real frequency response residual and real Jacobian matrix, and construct a real relationship equation between the reconstructed real Jacobian matrix and the real frequency response residual;

[0067] Step S5: Analyze the Pearson correlation coefficient between each column vector of the reconstructed real Jacobian matrix and the real frequency response residual, sort the column vectors from high to low according to the correlation, and record the corresponding network element numbers;

[0068] Step S6: Select the network components corresponding to the column vectors with the highest correlation, construct a least squares equation to solve the parameter correction value of the faulty component, update the equivalent network parameters and recalculate the frequency response curve;

[0069] Step S7, iteratively execute steps S3 to S6 until the residual between the updated frequency response curve and the measured abnormal curve is less than a preset first residual threshold, and determine the fault location, type and severity based on the final corrected network component number and parameter changes.

[0070] To facilitate understanding of the present invention, the transformer winding fault diagnosis method provided by the present invention is further described below based on the principle of the transformer winding fault diagnosis method of the present invention and the process of diagnosing transformer winding faults based on iterative calculation in the embodiment.

[0071] Wherein, step S1 specifically includes:

[0072] Step S11: The test frequency range is 1kHz to 2MHz. The sweep frequency signal is generated by a signal generator and input through the first end of the primary winding, with the end remaining suspended.

[0073] Step S12: The secondary winding of the transformer under test is grounded via a short-circuit wire to eliminate interference of the secondary side distributed parameters on the frequency response curve;

[0074] Step S13: Calculate the frequency response function by measuring the spectrum data of the primary side head end current and terminal voltage , the calculation formula is:

[0075] ;

[0076] in, is the frequency response function, which characterizes the impedance characteristics of the transformer winding under high-frequency signals; is the voltage amplitude at the end of the primary winding of the transformer at frequency f, in volts; is the current amplitude at the beginning of the primary winding of the transformer at frequency f, in amperes; Is the phase factor, reflecting the phase difference between voltage and current ; is an imaginary unit, that is, , characterizing the phase shift characteristics of inductance and capacitance.

[0077] In the winding test of a power transformer model SFZ-10000 / 110, the frequency response characteristic test operation is performed. The specific steps are as follows:

[0078] Step S11: Setting and inputting the sweep frequency signal.

[0079] A Keysight 33500B signal generator was used, configured to output a swept frequency signal with a frequency range of 1 kHz to 2 MHz, in 10 kHz increments, and an output level of 10 Vpp (peak-to-peak). The swept frequency signal was connected via a coaxial cable to the beginning of the transformer's primary (high-voltage) winding (terminal H1), leaving the end of the primary winding (terminal H2) floating. For example, when testing a 110 kV / 10 kV distribution transformer with a star-connected primary winding, the signal was connected to terminal H1, while terminal H2 was left floating to ensure unidirectional signal transmission.

[0080] Step S12: Secondary side winding grounding processing.

[0081] Use a 4mm² copper conductor to short-circuit all terminals (X1, X2, and X3) on the secondary (low-voltage) side and connect them to a grounding stake, with a grounding resistance of 0.1Ω or less. Comparing the frequency response curves before and after grounding reveals an abnormal resonant peak (amplitude deviation +5dB) at 1.5MHz when not grounded. After grounding, the peak disappears, and the curve becomes smoother.

[0082] Step S13: Calculate frequency response function.

[0083] Use the Hioki 9016-01 high-precision current probe to measure the primary current. , accuracy ±0.5%. Use Tektronix P5200A high voltage differential probe to measure the primary side terminal voltage. , accuracy ±0.2%. Use Keysight MXR series oscilloscope to measure the voltage and current phase difference For example, at 1MHz, the phase difference between voltage and current is 80 degrees. Calculate the frequency response function corresponding to this frequency point.

[0084] The advantages of using a test frequency range of 1kHz to 2MHz for this invention are: the low frequency range (1kHz to 100kHz) is used to capture the overall inductance characteristics of the winding (such as interturn looseness); the high frequency range (greater than 500kHz) is used to stimulate interturn capacitance and distributed parameters, locating partial discharge or insulation degradation. This invention uses a head-end input with the terminal suspended to test, which avoids signal reflection interference, reduces the measured signal attenuation rate, and improves the signal-to-noise ratio of the head-end input compared to the center-tap input method. Grounding the secondary winding eliminates interference from the secondary distributed parameters on the frequency response curve, improving the curve stability.

[0085] Furthermore, step S2 includes:

[0086] Step S21: jointly analyze the amplitude-frequency characteristics and phase-frequency characteristics of the normal state frequency response curve, and calculate the total number N of characteristic frequencies by extreme point detection and inflection point identification; extreme points include peak values ​​and valley values;

[0087] Step S22: Based on the total number of characteristic frequencies N, the mutual inductance effect in the high frequency band is compensated according to a preset ratio, and the number of equivalent network units n of the winding is dynamically determined; the calculation formula is: ,in, To round down;

[0088] Step S23: constructing a winding equivalent network of a chain series structure using the RLC units with the determined number of units;

[0089] Step S24: Based on a pre-stored least squares optimization algorithm and taking the measured frequency response curve as a benchmark, the resistance, inductance, and capacitance parameters of each unit are iteratively adjusted so that the frequency response residual between the frequency response curve of the winding equivalent network and the measured frequency response curve converges to a preset second residual threshold. The objective function of the pre-stored least squares optimization algorithm is: ;in, The frequency response function measured by the frequency sweep test is used to characterize the frequency response characteristics of the winding under normal conditions; The frequency of the transformer primary winding end in the equivalent network is The voltage amplitude at , in volts; The frequency of the transformer primary winding head in the equivalent network is The current amplitude at , in amperes; Is the phase factor, reflecting the phase difference between voltage and current in the equivalent network .

[0090] In the winding test of a power transformer model SFZ-10000 / 110 (rated voltage 110kV / 10 kV, high-voltage winding adopts a chain structure), 12 significant extreme points (6 peaks and 6 valleys) were detected in the amplitude-frequency curve of the normal winding; 8 inflection points (phase change rate ≥ 5° / Hz) were identified in the phase-frequency curve; the total number of characteristic frequencies N = 12 + 8 = 20. For example, the mutual inductance element M of the high-voltage winding 1(n+1) An implicit resonance peak is generated at 1.8MHz, which requires additional compensation units to cover. The compensation rules are: , that is, 5 new winding equivalent network units are added to compensate for the mutual inductance effect in the high frequency band (1MHz to 2MHz) (the mutual inductance element M of the high voltage winding 1(n+1) After compensation, we get Figure 2 The winding equivalent network diagram shown in the figure is as follows. The compensated winding equivalent network consists of 25 winding equivalent network units (i.e. RLC units), each of which contains a resistor R w 、Inductor L w , capacitance to ground C gw and inter-turn capacitance C sw , the component layout is shown in Figure 2. Mutual inductance coefficient M is set between adjacent inductors. wk (like Figure 2 The red dotted box area in the figure shows the magnetic field interaction between the winding turns; the chain series structure (refer to Figure 2 The first end is connected to the sweep frequency signal and the other end is left floating.

[0091] Resistor R w :Calculate the DC resistance based on the copper conductor cross-sectional area (120 mm²) and add skin effect correction; inductance L w :Initialize according to the number of winding turns (for example, 500 turns on the high-voltage side), geometric dimensions and mutual inductance distribution in the image; ground capacitance C gw , inter-turn capacitance C sw : Calculated based on the dielectric constant of the inter-turn insulation material (e.g., epoxy resin).

[0092] The present invention uses a dynamic compensation mechanism to additionally compensate for 5 winding equivalent network units to accurately characterize the mutual inductance M in the image. wk The high-frequency resonance points (such as the amplitude dip at 1.8 MHz) caused by this are detected, thus avoiding the frequency band omission problem of traditional testing. Through dynamic unit number compensation, chain topology construction, and closed-loop parameter optimization, the present invention overcomes the shortcomings of traditional testing methods in insufficient modeling of high-frequency resonance points in step S2, achieving high-precision matching across the entire frequency band (1 kHz to 2 MHz).

[0093] In addition, step S3 includes:

[0094] Step S31: Based on the reconstructed equivalent network, the voltage and current values ​​of the RLC elements in all branches are solved by circuit analysis;

[0095] Step S32: For each RLC element, calculate the sensitivity of the frequency response function to the parameter based on its voltage or current value, where the sensitivity of the inductance parameter is characterized by the product of the square of its branch current value and the angular frequency, the sensitivity of the capacitance parameter is characterized by the product of the square of its branch voltage value and the angular frequency, the sensitivity of the resistance parameter is directly characterized by the square of its branch current value, and the sensitivity of the mutual inductance parameter is characterized by the product of the current product of the associated inductance branch and the angular frequency;

[0096] Step S33: Fill all sensitivity values ​​into the matrix according to the pre-stored test frequency points and parameter indexes to form a Jacobian matrix A;

[0097] Step S34: compare the measured sensitivity with the Jacobian matrix elements, and when the deviation between the measured sensitivity and the matrix elements exceeds a preset deviation threshold, re-execute steps S31 to S33 until the deviation converges.

[0098] Based on the winding equivalent network constructed in step S2, the voltage of each branch is calculated using Kirchhoff's law and current value ; According to the first-order derivative rule of the frequency response function of the network parameters defined above, calculate respectively:

[0099] Inductor parameter sensitivity: ,in, is the effective value of the current in the w-th inductor branch;

[0100] Resistance parameter sensitivity: ,in, is the effective value of the current in the wth resistance branch;

[0101] Capacitance parameter sensitivity: ,in, is the effective value of the voltage of the w-th capacitor branch;

[0102] Mutual inductance parameter sensitivity: ,in, 、 is the effective value of the current in the mutual inductance associated branch.

[0103] Fill the calculated sensitivity values ​​of each parameter into the Jacobian matrix A according to the corresponding relationship between frequency point i and parameter w: , where the matrix elements are defined as: , Represents the wth network parameter (including ).

[0104] Matrix consistency is verified through frequency response sensitivity simulation: If the calculated parameter sensitivity value at any frequency point i deviates from the simulated value by more than 10%, the branch voltage and current calculation process is recalibrated; the complete Jacobian matrix A is output for subsequent frequency response residual correlation analysis and fault location.

[0105] This method constructs a highly reliable Jacobian matrix through precise branch parameter calculation, sensitivity formula mapping, and matrix calibration. This solves the problem of traditional methods' inadequate modeling of high-frequency mutual inductance effects. Combined with the chain topology in the image, it comprehensively covers the electromagnetic coupling characteristics of all components, providing a reliable data foundation for fault diagnosis.

[0106] Furthermore, step S4 includes:

[0107] Step S41: Calculate the frequency response residuals of the normal frequency response curve and the abnormal frequency response curve, and decompose the real and imaginary parts of the frequency response residuals. Then, vertically concatenate the real and imaginary parts of the frequency response residuals to obtain a real residual vector ;

[0108] Step S42: Decompose the Jacobian matrix A into real and imaginary parts, and vertically concatenate the real and imaginary parts of the Jacobian matrix A to obtain a reconstructed real Jacobian matrix ;

[0109] Step S43: reconstruct the real Jacobian matrix and the real residual vector Make the association and construct its real number relationship equation: ,in, is the network parameter correction value vector, which is used to represent the parameter change caused by the fault.

[0110] The real part of the decomposed frequency response residual represents amplitude differences, reflecting changes in resistance or mutual inductance; the imaginary part of the decomposed frequency response residual represents phase differences, indicating inductance or capacitance faults. By separating the real and imaginary parts, the fault contributions of resistance / mutual inductance and inductance / capacitance are independently quantified. The real and imaginary residuals are vertically concatenated into a vector of dimension 2n × 1.

[0111] The real part of the Jacobian matrix A corresponds to the sensitivity of the frequency response amplitude to the parameters; the imaginary part of the Jacobian matrix A corresponds to the sensitivity of the frequency response phase to the parameters; the real and imaginary matrices are vertically spliced ​​to form a matrix with a dimension of 2n×m.

[0112] For example, for the mutual inductance parameter M in the image 12 , whose sensitivity elements are: ; Solve by least squares, and get , indicating that the mutual inductance value drops by 15%.

[0113] This method uses complex residual decomposition and vertical splicing to analyze high-frequency resonance characteristics as independent amplitude and phase in the real domain, overcoming the technical bottleneck of traditional frequency response methods in distinguishing multidimensional fault sources. Combining the mutual inductance sign and chain topology in the image, it accurately quantifies the magnetic field coupling effect, supporting rapid fault component location and design optimization.

[0114] Then, step S5 includes:

[0115] Step S51: Calculate the real Jacobian matrix according to the Pearson correlation coefficient PCC formula Each column vector of and real frequency response residual The correlation, ; The Pearson correlation coefficient formula is: ,in, is the test frequency number; is a column vector The value of the i-th element in ; is the real frequency response residual The value of the i-th element in ;

[0116] Step S52: Real Jacobian matrix All column vectors of traverse the operation, and substitute each column vector into the Pearson correlation coefficient formula to calculate its difference with the real frequency response residual During the calculation process, record the correlation coefficient value corresponding to each column vector and the column vector in the real Jacobian matrix The column index g in the transformer winding equivalent network is mapped to the transformer winding equivalent network parameter number, that is, each column index g corresponds to one or a group of component parameters in the transformer winding equivalent network.

[0117] Step S53: According to the calculated correlation coefficient value, the real Jacobian matrix is ​​sorted in descending order. All column vectors of are sorted; while sorting, the transformer winding equivalent network parameter number information corresponding to each column vector is output in sequence according to the recorded column index and the mapping relationship between the column index and the transformer winding equivalent network parameter number to form a parameter number sequence.

[0118] For example, for the test of the power transformer model SFZ-10000 / 110, the high-voltage winding includes inductance, resistance, and capacitance to ground, and the low-voltage winding includes inductance, mutual inductance, and inter-winding capacitance; from the reconstructed real Jacobian matrix Extract the column vector of each network parameter and calculate the real Jacobian matrix according to the Pearson correlation coefficient PCC formula Each column vector of and real frequency response residual and for all network parameters ( ) are sorted by absolute value, for example, the generated sequence is: , mutual induction The correlation coefficient (corresponding to the mutual inductive coupling region between the high-voltage winding L1 and the low-voltage winding L11) is 0.92, indicating that its value is decreasing (e.g., the winding displacement leads to a weakening of the coupling); The correlation coefficient of the capacitive coupling region between the high-voltage winding L1 and the low-voltage winding L11 is 0.85, indicating an increase in its value (e.g., insulation degradation leading to an increase in capacitance).

[0119] The present invention accurately locates the mutual inductance and capacitive coupling fault areas marked in the image through Pearson correlation coefficient calculation and sorting, solving the diagnostic ambiguity problem of traditional methods in complex coupling scenarios.

[0120] In addition, step S6 includes:

[0121] Step S61: Based on the sorting results of the Pearson correlation coefficient, select The top 5% real Jacobian matrix of the correlation Column vector, records the corresponding network element number and forms the fault element number set S;

[0122] Step S62: According to the real Jacobian matrix Extract the column vector from the matrix and construct the Jacobian matrix subset ;

[0123] Step S63: Correct the network parameter value vector Extract the elements corresponding to the fault component number set S to form the network parameter correction value sub-vector ;

[0124] Step S64: Based on the Jacobian matrix subset and the network parameter correction value subvector Construct the least squares equation: ;

[0125] Step S65: Solve the constructed least squares equation , get the corrected values ​​of network component parameters , and its calculation formula is: ,in, is a subset of the Jacobian matrix The transposed matrix of

[0126] Step S66: Correction values ​​of network element parameters obtained based on the solution The parameters of the winding equivalent network are updated; and the frequency response curve is recalculated based on the updated network parameters.

[0127] For example, when testing the power transformer model SFZ-10000 / 110, the following results are obtained: Figure 2 For the equivalent network shown in the figure, if the constructed least squares equation is solved, the corrected values ​​of the network component parameters are obtained: 、 The corresponding parameters are updated as follows: modify the inductance value of L5 to L5+10μH to optimize the power frequency impedance characteristics; change M 1(n+1) Adjust from 0.5 mH to 0.3 mH to suppress high-frequency coupling crosstalk; after correction, the resonance peak deviation of the frequency response curve at 1.5 MHz is reduced from ±5dB to ±0.5dB.

[0128] The present invention directly and accurately corrects the mutual inductance coupling and capacitance fault areas in the image through high-correlation component screening, least squares solution and parameter update, thereby improving fault repair efficiency and system stability.

[0129] Finally, step S7 includes:

[0130] After the iteration is terminated, the final corrected network component number and parameter changes are combined with a pre-stored fault signature library to determine the location, type, and severity of the fault. The fault signature library contains the fault type and location corresponding to different component numbers, as well as the corresponding relationship between parameter changes and fault severity. The fault type and severity are determined based on the parameter changes of the network components. Fault types include short circuit, open circuit, and parameter drift, and severity includes mild fault, moderate fault, and severe fault.

[0131] For example, when testing the power transformer model SFZ-10000 / 110, the following results are obtained: Figure 2 The equivalent network shown in Figure 1 has the following final correction parameters: 、 and ; Obtain the corresponding fault location, type and severity by querying the pre-stored fault feature library; Among them, the severity judgment criteria are:

[0132] Minor fault: , ; Moderate failure: , ; Serious failure: , .

[0133] :

[0134] The matching fault type is winding displacement / deformation, corresponding to the high-voltage winding L1 and the low-voltage winding L in the image. 26 Mechanical looseness between the components; this is a moderate fault.

[0135] :

[0136] The matching fault type is insulation oil degradation, corresponding to the high-voltage winding L5 to ground capacitance C in the image. g5 Area; serious failure.

[0137] :

[0138] The matching fault type is inter-turn looseness, corresponding to the high-voltage section L in the image. 15 Axial displacement; it is a minor fault.

[0139] This invention uses parameter matching, classification rules, and a dynamic feature library to accurately associate component numbers in images with physical faults, achieving a closed-loop diagnosis from parameter correction to maintenance decision-making. Combining mutual inductance coupling zones with capacitor layout significantly improves fault location efficiency and repair accuracy. Based on severity classification, high-risk faults (such as severe insulation degradation) are prioritized, reducing downtime.

[0140] Figure 3 This is a structural diagram of a terminal 300 provided in an embodiment of the present invention. The terminal 300 can be used to execute the transformer winding fault diagnosis method provided in an embodiment of the present invention.

[0141] The terminal 300 may include a processor 310, a memory 320, and a communication module 330. These components communicate via one or more buses. Those skilled in the art will appreciate that the server structure shown in the figure does not limit the present invention. The server structure may be a bus structure or a star structure, and may include more or fewer components than shown, or may combine certain components or arrange the components differently.

[0142] Memory 320 can be used to store execution instructions of processor 310. Memory 320 can be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk. When the execution instructions in memory 320 are executed by processor 310, terminal 300 can perform some or all of the steps in the above-described method embodiments.

[0143] The processor 310 is the control center of the storage terminal. It uses various interfaces and lines to connect various parts of the entire electronic terminal. It executes various functions of the electronic terminal and / or processes data by running or executing software programs and / or modules stored in the memory 320, and calling data stored in the memory. The processor can be composed of an integrated circuit (IC), for example, it can be composed of a single packaged IC, or it can be composed of multiple packaged ICs with the same or different functions. For example, the processor 310 can only include a central processing unit (CPU). In the embodiment of the present invention, the CPU can be a single computing core or multiple computing cores.

[0144] The communication module 330 is used to establish a communication channel so that the storage terminal can communicate with other terminals, receive user data sent by other terminals, or send user data to other terminals.

[0145] The present invention also provides a computer storage medium, wherein the computer storage medium may store a program that, when executed, may include some or all of the steps of each embodiment provided herein. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0146] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus a necessary general-purpose hardware platform. Based on this understanding, the technical solutions in the embodiments of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, among other media capable of storing program code, and includes instructions for causing a computer terminal (which can be a personal computer, a server, or a second terminal, a network terminal, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention.

[0147] In this specification, the same or similar parts between the various embodiments can be referred to each other. In particular, for the terminal embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description in the method embodiment.

[0148] Although the present invention has been described in detail with reference to the accompanying drawings and in conjunction with preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, persons of ordinary skill in the art may make various equivalent modifications or substitutions to the embodiments of the present invention, and such modifications or substitutions shall be within the scope of the present invention. Any changes or substitutions that can be easily conceived by persons skilled in the art within the technical scope disclosed in the present invention shall be within the scope of protection of the present invention.

Claims

1. A transformer winding fault diagnosis method, characterized in that: include: Step S1: Obtain frequency response curves of the transformer winding under normal and abnormal conditions through a frequency sweep test, wherein during the test, the secondary winding is short-circuited and grounded, the first end of the primary winding is connected to a frequency sweep signal and the end is left floating, and the test frequency range is 1 kHz to 2 MHz; Step S2: determining the number of units of the equivalent network according to the number of characteristic frequencies of the frequency response curve of the normal state winding, and reconstructing the normal winding equivalent network composed of multiple RLC units; Step S3: Calculate the voltage and current parameters of all branches based on the reconstructed equivalent network, and construct a Jacobian matrix for characterizing the sensitivity of the frequency response function to changes in network parameters; Step S4: Calculate the frequency response residuals of the normal frequency response curve and the abnormal frequency response curve, decompose the Jacobian matrix and the frequency response residual respectively, and reconstruct them respectively to obtain the corresponding reconstructed real frequency response residual and real Jacobian matrix, and construct a real relationship equation between the reconstructed real Jacobian matrix and the real frequency response residual; Step S5: Analyze the Pearson correlation coefficient between each column vector of the reconstructed real Jacobian matrix and the real frequency response residual, sort the column vectors from high to low according to the correlation, and record the corresponding network element numbers; Step S6: Select the network components corresponding to the column vectors with the highest correlation, construct a least squares equation to solve the parameter correction value of the faulty component, update the equivalent network parameters and recalculate the frequency response curve; Step S7, iteratively execute steps S3 to S6 until the residual between the updated frequency response curve and the measured abnormal curve is less than a preset first residual threshold, and determine the fault location, type and severity based on the final corrected network component number and parameter changes.

2. The transformer winding fault diagnosis method according to claim 1, characterized in that: Step S1 specifically includes: Step S11: The test frequency range is 1kHz to 2MHz. The sweep frequency signal is generated by a signal generator and input through the first end of the primary winding, with the end remaining suspended. Step S12: The secondary winding of the transformer under test is grounded via a short-circuit wire to eliminate interference of the secondary side distributed parameters on the frequency response curve; Step S13: Calculate the frequency response function by measuring the spectrum data of the primary side head end current and terminal voltage , the calculation formula is: ； in, is the frequency response function, which characterizes the impedance characteristics of the transformer winding under high-frequency signals; is the voltage amplitude at the end of the primary winding of the transformer at frequency f, in volts; is the current amplitude at the beginning of the primary winding of the transformer at frequency f, in amperes; Is the phase factor, reflecting the phase difference between voltage and current ; is an imaginary unit, that is, , characterizing the phase shift characteristics of inductance and capacitance.

3. The transformer winding fault diagnosis method according to claim 2, characterized in that: Step S2 includes: Step S21: jointly analyze the amplitude-frequency characteristics and phase-frequency characteristics of the normal state frequency response curve, and calculate the total number N of characteristic frequencies by extreme point detection and inflection point identification; extreme points include peak values ​​and valley values; Step S22: Based on the total number of characteristic frequencies N, the mutual inductance effect in the high frequency band is compensated according to a preset ratio, and the number of winding equivalent network units n is dynamically determined; the calculation formula is: ,in, To round down; Step S23: constructing a winding equivalent network of a chain series structure using the RLC units with the determined number of units; Step S24: Based on a pre-stored least squares optimization algorithm and taking the measured frequency response curve as a benchmark, the resistance, inductance, and capacitance parameters of each unit are iteratively adjusted so that the frequency response residual between the frequency response curve of the winding equivalent network and the measured frequency response curve converges to a preset second residual threshold. The objective function of the pre-stored least squares optimization algorithm is: ;in, The frequency response function measured by the frequency sweep test is used to characterize the frequency response characteristics of the winding under normal conditions; The frequency of the transformer primary winding end in the equivalent network is The voltage amplitude at , in volts; The frequency of the transformer primary winding head in the equivalent network is The current amplitude at , in amperes; Is the phase factor, reflecting the phase difference between voltage and current in the equivalent network .

4. The transformer winding fault diagnosis method according to claim 3, characterized in that: Step S3 includes: Step S31: Based on the reconstructed equivalent network, the voltage and current values ​​of the RLC elements in all branches are solved by circuit analysis; Step S32: For each RLC element, calculate the sensitivity of the frequency response function to the parameter based on its voltage or current value, where the sensitivity of the inductance parameter is characterized by the product of the square of its branch current value and the angular frequency, the sensitivity of the capacitance parameter is characterized by the product of the square of its branch voltage value and the angular frequency, the sensitivity of the resistance parameter is directly characterized by the square of its branch current value, and the sensitivity of the mutual inductance parameter is characterized by the product of the current product of the associated inductance branch and the angular frequency; Step S33: Fill all sensitivity values ​​into the matrix according to the pre-stored test frequency points and parameter indexes to form a Jacobian matrix A; Step S34: compare the measured sensitivity with the Jacobian matrix elements, and when the deviation between the measured sensitivity and the matrix elements exceeds a preset deviation threshold, re-execute steps S31 to S33 until the deviation converges.

5. The transformer winding fault diagnosis method according to claim 4, characterized in that: Step S4 includes: Step S41: Calculate the frequency response residuals of the normal frequency response curve and the abnormal frequency response curve, and decompose the real and imaginary parts of the frequency response residuals. Then, vertically concatenate the real and imaginary parts of the frequency response residuals to obtain a real residual vector ; Step S42: Decompose the Jacobian matrix A into real and imaginary parts, and vertically concatenate the real and imaginary parts of the Jacobian matrix A to obtain a reconstructed real Jacobian matrix. ; Step S43: reconstruct the real Jacobian matrix and the real residual vector Make the association and construct its real number relationship equation: ,in, is the network parameter correction value vector, which is used to represent the parameter change caused by the fault.

6. The transformer winding fault diagnosis method according to claim 5, characterized in that: Step S5 includes: Step S51: Calculate the real Jacobian matrix according to the Pearson correlation coefficient PCC formula Each column vector of and real frequency response residual The correlation, ; The Pearson correlation coefficient formula is: ,in, is the test frequency number; is a column vector The value of the i-th element in ; is the real frequency response residual The value of the i-th element in ; Step S52: Real Jacobian matrix All column vectors of traverse the operation, and substitute each column vector into the Pearson correlation coefficient formula to calculate its difference with the real frequency response residual During the calculation process, record the correlation coefficient value corresponding to each column vector and the column vector in the real Jacobian matrix The column index g in the transformer winding equivalent network is mapped to the transformer winding equivalent network parameter number, that is, each column index g corresponds to one or a group of component parameters in the transformer winding equivalent network. Step S53: According to the calculated correlation coefficient value, the real Jacobian matrix is ​​sorted in descending order. All column vectors of are sorted; while sorting, the transformer winding equivalent network parameter number information corresponding to each column vector is output in sequence according to the recorded column index and the mapping relationship between the column index and the transformer winding equivalent network parameter number to form a parameter number sequence.

7. The transformer winding fault diagnosis method according to claim 6, characterized in that: Step S6 includes: Step S61: Based on the sorting results of the Pearson correlation coefficient, select The top 5% real Jacobian matrix of the correlation Column vector, records the corresponding network element number and forms the fault element number set S; Step S62: According to the real Jacobian matrix Extract the column vector from the matrix and construct the Jacobian matrix subset ; Step S63: Correct the network parameter value vector Extract the elements corresponding to the fault component number set S to form the network parameter correction value sub-vector ; Step S64: Based on the Jacobian matrix subset and the network parameter correction value subvector Construct the least squares equation: ; Step S65: Solve the constructed least squares equation , get the corrected values ​​of network component parameters , and its calculation formula is: ,in, is a subset of the Jacobian matrix The transposed matrix of Step S66: Correction values ​​of network element parameters obtained based on the solution The parameters of the winding equivalent network are updated; and the frequency response curve is recalculated based on the updated network parameters.

8. The transformer winding fault diagnosis method according to claim 7, characterized in that: Step S7 includes: After the iteration is terminated, the final corrected network component number and parameter changes are combined with a pre-stored fault signature library to determine the location, type, and severity of the fault. The fault signature library contains the fault type and location corresponding to different component numbers, as well as the corresponding relationship between parameter changes and fault severity. The fault type and severity are determined based on the parameter changes of the network components. Fault types include short circuit, open circuit, and parameter drift, and severity includes mild fault, moderate fault, and severe fault.

9. A terminal, characterized in that: include: processor; a memory for storing execution instructions of the processor; The processor is configured to execute the method according to any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images