A transformer fault detection method based on sound spectrum

Abstract

The present application relates to the technical field of transformer fault detection, and particularly relates to a transformer fault detection method based on sound spectrum, comprising the following steps: a spectrum analysis module analyzes the frequency amplitude difference between multiple sound detection signals, and adjusts the installation positions of multiple sound sensors according to the comparison result; the spectrum analysis module compares and analyzes each group of sound detection noise reduction signals with a preset normal working condition sound spectrum, controls a warning sending module to send different levels of warning signals to a background display terminal according to the comparison result; the spectrum analysis module compares and analyzes multiple characteristic sound spectrums in a preset fault working condition matrix with multiple groups of sound detection noise reduction signals, and controls the warning sending module to send corresponding fault working conditions and their probability value signals to the background display terminal according to the comparison result. The present application can determine the severity level of transformer faults, and can also prompt the type and probability value of the faults.

Description

Technical Field

[0001] This invention relates to the field of transformer testing technology, and in particular to a transformer fault detection method based on sound spectrum. Background Technology

[0002] The power system currently contains countless transformers, which are widely distributed. During long-term operation, faults such as resonance and insulation aging discharge are inevitable. On-site inspections by personnel are insufficient to provide comprehensive coverage of transformer operation, inevitably leading to delayed fault detection and power outages. While many transformers are equipped with online monitoring devices for heavy gas, light gas, over-temperature, and overcurrent protection, faults such as insulation aging and insulation breakdown discharge are primarily detected and confirmed through on-site inspections. The limited online monitoring of transformer insulation aging and breakdown discharge faults prevents real-time monitoring of abnormal transformer operation, causing minor incidents to escalate. Existing technology utilizes acoustic sensors to provide real-time monitoring of discharge or breakdown faults caused by insulation aging or contamination in transformers, enabling timely detection and handling.

[0003] There are multiple types of transformer malfunctions, and the sounds emitted vary depending on the specific fault condition. For example, (1) A transformer emits a uniform "humming" sound during normal operation, but if there is a high and muffled "humming" sound, it may be caused by overload operation, high current, or increased core vibration. The transformer load change should be monitored. (2) A short-term "waaa" sound inside the transformer may be caused by the start-up of a large power equipment, a sudden increase in load, and the generation of high-order harmonics; it may also be caused by overvoltage, lightning surge intrusion, or through-circuit short circuit in the power grid. (3) A gapping "humming" sound inside the transformer, sometimes coarse and sometimes fine, sometimes high and sometimes low, may be caused by ferroresonance in the system, or by a broken wire or grounding fault. (4) A "crackling" and "hissing" discharge sound from the transformer may be caused by partial discharge if the sound is muffled, or by a cracked, damaged, or heavily contaminated bushing, or by poor contact of the equipment wiring. (5) A "gurgling" or "boiling" sound from the transformer may be caused by a short circuit in the winding or poor contact of the joints, resulting in severe overheating. (6) If the transformer makes a popping sound, it may be due to severe discharge caused by insulation breakdown inside the transformer. (7) If there is a "clanging" hammering sound and a "whooshing" wind sound inside the transformer, it may be due to loose iron core clamps or individual loose parts.

[0004] Existing transformer detection methods collect and reduce the noise in the transformer's sound signals, then analyze the reduced signals to determine if a fault has occurred. However, these methods cannot indicate the severity of the fault. Maintenance personnel must then randomly repair multiple faulty transformers one by one. In situations involving multiple transformers, this lack of information for quickly repairing severely faulty transformers can lead to serious consequences. Furthermore, existing methods cannot accurately predict the types of faults that may occur, hindering maintenance personnel from preparing for potential problems. They are forced to rely on experience or on-site assessments, resulting in extended repair cycles and increased inconvenience.

[0005] Therefore, in view of the problems existing in the prior art, how to provide a transformer fault detection method based on sound spectrum that can determine the severity level of transformer faults and display what kind of faults have occurred and their probability is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a transformer fault detection method based on sound spectrum, which can determine the severity level of transformer faults and indicate the type of fault and its probability value.

[0007] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0008] A transformer fault detection method based on sound spectrum includes the following steps:

[0009] Step a: Pre-install multiple sound sensors symmetrically around the transformer, with each pair of symmetrically arranged sound sensors forming a sound sensor group; the acquisition module acquires the sound detection signals from the multiple sound sensors and transmits them to the spectrum analysis module, which analyzes and compares the frequency amplitude differences between the multiple sound detection signals and adjusts the installation positions of the multiple sound sensors based on the comparison results;

[0010] Step b: The spectrum analysis module is set with the preset normal operating condition sound spectrum of the transformer. The spectrum analysis module samples and denoises the sound detection signals of each group of sound sensors to obtain a set of sound detection denoised signals. The module compares and analyzes each set of sound detection denoised signals with the preset normal operating condition sound spectrum. Based on the comparison results, the module controls the early warning sending module to send different levels of early warning signals to the background display terminal.

[0011] Step c: The spectrum analysis module extracts features from the sound spectra of multiple original fault conditions of the transformer to generate a preset fault condition sound spectrum matrix. The preset fault condition sound spectrum matrix includes feature sound spectra of multiple corresponding transformer faults.

[0012] Step d: The spectrum analysis module compares and analyzes the spectra of multiple characteristic sounds in the preset fault condition matrix with multiple sets of sound detection and noise reduction signals, and controls the early warning sending module to send the corresponding fault condition and its probability value signal to the background display terminal based on the comparison results.

[0013] In step a, the spectrum analysis module is set with a preset standard spectrum difference range. Based on the comparison results, the warning sending module is controlled to send an adjustment signal to the on-site display terminal. The installation positions of multiple sound sensors are adjusted according to the adjustment signal until the frequency amplitude difference between multiple sound detection signals is within the preset standard spectrum difference range, thus completing the adjustment and installation of multiple sound sensors.

[0014] In step b, the spectrum analysis module is configured with a preset standard frequency amplitude difference and a preset frequency amplitude difference matrix B. The spectrum analysis module analyzes and compares the single-group frequency amplitude difference between each group of sound detection noise reduction signals and the preset normal operating condition sound spectrum. Based on the number L of multiple single-group frequency amplitude differences greater than the preset standard spectrum difference, and the correspondence between the sum of L single-group frequency amplitude differences APL and the preset frequency amplitude difference matrix B, the warning sending module sends different levels of warning signals to the background display terminal.

[0015] In step d, the spectrum analysis module is set with a preset standard probability value; the spectrum analysis module analyzes and compares to obtain a fault probability value matrix corresponding to different fault conditions, and filters out fault conditions in the probability value matrix that are greater than the preset standard probability value. The warning sending module sends the corresponding fault condition and its probability value signal to the background display terminal.

[0016] The beneficial effects of this invention are as follows: During on-site installation, the distance between multiple sound detection signals and the transformer sound source is adjusted. By detecting the corresponding frequency amplitude information between the multiple sound detection signals, the amplitudes of the multiple sound detection signals relative to the transformer sound source are adjusted to be consistent. At this time, the phase values ​​of the multiple sound detection signals are also basically consistent. The comparison results are displayed on the on-site display terminal. On-site installers adjust the corresponding sound sensors according to the display results to ensure the consistency of the amplitude and phase of the detection signals of the multiple sound sensors relative to the transformer sound source. Each sound sensor group includes two symmetrically arranged sound sensors with consistent phase for the transformer sound source detection signal, but inconsistent phase for the peripheral sound source signal. Thus, the two detection signals of each group of sound sensors can be... The system further performs noise reduction processing on external signals; and the spectrum analysis module can compare multiple sets of sound detection and noise reduction signals with a preset normal operating condition sound spectrum as a benchmark. Based on the comparison results, it sends different levels of warning information to the backend display terminal, allowing maintenance personnel to prioritize the handling of transformers with severe faults according to the different levels of warning signals. In addition, the spectrum analysis module pre-extracts the characteristic sound spectra of different faults of the corresponding transformers. When a fault is detected, it compares whether multiple sets of sound detection and noise reduction signals contain the sound spectra of preset different fault conditions and analyzes their probabilities. Based on the comparison results, it sends the corresponding fault conditions and probability values ​​to the backend display terminal, so that maintenance personnel can prepare the corresponding maintenance tools in advance according to the probability of the fault, which can improve maintenance efficiency.

[0017] Based on the above technical solution, the present invention can be further improved as follows.

[0018] Further, in step a, the spectrum analysis module analyzes and compares the frequency amplitude difference between multiple sound detection signals, including: the spectrum analysis module has a preset transformer sound source signal frequency band set within it; the spectrum analysis module performs filtering processing before comparing multiple sound detection signals, filtering out low-frequency components in the sound detection signals that are less than the minimum value of the preset transformer sound source signal frequency band and high-frequency components that are greater than the maximum value of the preset transformer sound source signal frequency band; the spectrum analysis module performs Fourier transform on the multiple sound detection signals to obtain the frequency-amplitude information nω-An for each sound detection signal, where nω is the frequency information in the sound detection signal, and An is the amplitude information corresponding to the nω isopleth, n = 1, 2, ..., N; one sound detection signal is selected as the calibration signal, and the frequency-amplitude difference A between the remaining sound detection signals and the calibration signal is analyzed. m The calculation formula is as follows:

[0019]

[0020] Where A 标n To calibrate the amplitude of the signal corresponding to an nω harmonic, A mnLet N be the amplitude of the nω harmonic corresponding to the m-th sound detection signal, where N is the total number of harmonics, and m = 1, 2, ..., M.

[0021] Furthermore, in step a, the control of the early warning sending module to send an adjustment signal to the on-site display terminal based on the comparison result includes: the spectrum analysis module is configured with a preset standard amplitude difference range (A). 01 A 02 ), where A 01 <0<A 02 And according to A m The installation position of the m-th sound sensor is adjusted based on the correspondence with the preset standard amplitude difference range.

[0022] When A 01 ≤A m ≤A 02 At that time, the early warning sending module sends a signal to the on-site display terminal that the installation of the m-th sound sensor is complete;

[0023] When A m <A 01 At that time, the early warning sending module sends a signal to the on-site display terminal to indicate that the m-th sound sensor is adjusting towards the transformer;

[0024] When A m >A 02 At that time, the early warning sending module sends a signal to the on-site display terminal that the m-th sound sensor is adjusting in the direction away from the transformer.

[0025] Further, in step b, the spectrum analysis module samples and denoises the sound detection signals of each group of sound sensors to obtain a set of sound detection denoising signals. This includes: each group of sound sensors comprises two symmetrically arranged sound sensors, whose sound detection signals are a first detection signal and a second detection signal, respectively. The spectrum analysis module samples the first and second detection signals at a Nyquist sampling period. Then, an addition circuit adds the sampled signals of the first and second detection signals, and an attenuation circuit attenuates the added detection signal to half its original value to obtain a set of sound detection denoising signals. The calculation formula is as follows:

[0026] S (n) =(S1) (n) +S2 (n) ) / 2

[0027] S1 (n) S2 is the first detection signal. (n) S is the second detection signal. (n) This is a set of noise-reduced signals for detecting noise reduction.

[0028] The beneficial effects of adopting the above-mentioned further scheme are: the amplitude and phase values ​​of the first detection signal and the second detection signal are equal for the transformer sound source; while for the external noise signal, the two have a phase difference, and after superposition, they cancel the external noise signal; after the first detection signal and the second detection signal are added and multiplied by 1 / 2, they will not cancel the transformer signal but will cancel the external noise signal after superposition, and finally a set of sound detection noise reduction signals for noise reduction processing of the external noise signal is obtained.

[0029] Further, in step b, the spectrum analysis module analyzes and compares the single-group frequency amplitude difference between each group of sound detection and noise reduction signals and the preset normal operating condition sound spectrum, including: the spectrum analysis module performs a Fourier transform on the preset normal operating condition sound spectrum to obtain the frequency-amplitude information n1ω-A0n1, n1=1,2,…,N1 of the transformer normal operating condition sound spectrum; the frequency-amplitude information obtained by performing a Fourier transform on the i-th group of sound detection and noise reduction signals is jω-Aij, j=1,2,…,J; then the formula for calculating the single-group frequency amplitude difference APi between the i-th group of sound detection and noise reduction signals and the preset normal operating condition sound spectrum is as follows:

[0030]

[0031] Where X is the number of frequencies that are the same as the preset normal operating condition sound spectrum in the i-th group of sound detection and noise reduction signals, i = 1, 2, ..., I; k1 is the frequency ratio weighting coefficient, k2 is the amplitude difference weighting coefficient, k1 > k2; A ix Let A be the amplitude of the x-th frequency corresponding to the i-th group of sound detection and noise reduction signals. 0x The amplitude of the xth frequency corresponding to the sound spectrum under normal operating conditions is preset; x = 1, 2, ..., X.

[0032] Furthermore, the spectrum analysis module is configured with a preset standard frequency amplitude difference (AP). 标 The spectrum analysis module counts multiple single-group frequency amplitude differences (AP1, AP2, ..., API) that are greater than AP. 标 The number of L;

[0033] When L=0, the warning sending module sends a normal detection signal to the background display terminal;

[0034] When L≥1, calculate the sum of the frequency amplitude differences of the L single groups, APL, using the following formula:

[0035] Where APl is the l-th element greater than AP 标 The single-group frequency amplitude difference;

[0036] The spectrum analysis module has a preset frequency amplitude difference matrix B, set as B = (B1, B2, B3), where B1 is the first preset frequency amplitude difference, B2 is the second preset frequency amplitude difference, B3 is the third preset frequency amplitude difference, and B1 < B2 < B3.

[0037] When APL < B1, the warning sending module sends a first-level warning signal to the background display terminal.

[0038] When B1≤APL<B2, the warning sending module sends a second-level warning signal to the background display terminal.

[0039] When B2≤APL<B3, the warning sending module sends a third-level warning signal to the background display terminal.

[0040] Furthermore, in step c, the spectrum analysis module is configured with a preset fault condition sound spectrum matrix U, defined as U(U1, U2, U3, U4, U5, U6), wherein:

[0041] U1 = The sound spectrum of the first preset fault condition, which corresponds to the characteristic sound spectrum of the transformer insulation breakdown discharge fault.

[0042] U2 = The sound spectrum of the second preset fault condition, which corresponds to the characteristic sound spectrum of the overheating fault caused by transformer short circuit or poor contact.

[0043] U3 = The sound spectrum of the third preset fault condition, which corresponds to the characteristic sound spectrum of a transformer open-circuit grounding fault;

[0044] U4 = The sound spectrum of the fourth preset fault condition, which corresponds to the characteristic sound spectrum of transformer overvoltage or lightning wave intrusion fault.

[0045] U5 = The sound spectrum of the fifth preset fault condition, which corresponds to the characteristic sound spectrum of faults such as transformer bushing cracks or poor contact of equipment line cards.

[0046] U6 = The sound spectrum of the sixth preset fault condition, corresponding to the characteristic sound spectrum of a transformer core loosening fault.

[0047] Further, in step d, the spectrum analysis module analyzes and compares to obtain a fault probability matrix corresponding to different fault conditions, including: the spectrum analysis module performs a Fourier transform on the sound spectrum of the r-th preset fault condition in Ur to obtain the corresponding spectrum-amplitude information n2ω-Arn2, n2=1,2,…,N2; r=1,2,3,4,5,6; the spectrum analysis module analyzes the frequency-amplitude information jω-Aij of the i-th group of sound detection and noise reduction signals to find that the frequency spectrum n2ω-Arn2 of the r-th preset fault condition is included. ir The calculation formula is as follows:

[0048]

[0049] Where Y is the number of frequencies in the sound spectrum of the i-th group of sound detection and noise reduction signal and the r-th preset fault condition sound; σ0 is the difference benchmark coefficient; K1 is the frequency ratio weighting coefficient, K2 is the amplitude difference weighting coefficient, K1>K2, and K1+K2=1; A iy Let A be the amplitude of the y-th frequency corresponding to the i-th group of sound detection and noise reduction signal. ry The amplitude of the y-th frequency corresponding to the preset fault condition sound spectrum is given by y = 1, 2, ..., Y;

[0050] The spectrum analysis module calculates the average probability value W of multiple sets of sound detection and noise reduction signals corresponding to the sound spectrum of the r-th preset fault condition. 平r The calculation formula is as follows:

[0051]

[0052] Where I is the total number of groups of sound detection and noise reduction signals, i = 1, 2, ..., I;

[0053] This leads to the fault probability matrix W corresponding to different fault conditions. 平 Set W 平 (W 平1 W 平2 W 平3 W 平4 W 平5 W 平6 ), where W 平1 W represents the average probability value corresponding to the first fault condition. 平2 W represents the average probability value corresponding to the second fault condition. 平3 W represents the average probability value corresponding to the third fault condition. 平4 W represents the average probability value corresponding to the fourth fault condition. 平5 W represents the average probability value corresponding to the fifth fault condition. 平6 This represents the average probability value corresponding to the sixth fault condition.

[0054] Furthermore, the fault conditions within the probability value matrix that exceed a preset standard probability value are filtered out. The early warning sending module then sends the corresponding fault condition and its probability value signal to the backend display terminal. This includes: the spectrum analysis module has a preset standard probability value W0 set within it, and the fault probability value matrix W0 is filtered out. 平 For values ​​greater than W0, send the corresponding fault condition and its probability value signal;

[0055] When W 平1 When W ≥ 0, the warning sending module sends the first fault condition and the probability value W of the first fault condition to the backend display terminal. 平1 Signal;

[0056] When W 平2 When W ≥ 0, the warning sending module sends the second fault condition and its probability value W to the backend display terminal. 平2 Signal;

[0057] When W 平3 When W ≥ 0, the warning sending module sends the third fault and the probability value W of the third fault condition to the backend display terminal. 平3 Signal;

[0058] When W 平4 When W ≥ 0, the warning sending module sends the fourth fault condition and its probability value W to the backend display terminal. 平4 Signal;

[0059] When W 平5 When W ≥ 0, the warning sending module sends the fifth fault condition and its probability value W to the backend display terminal. 平5 Signal;

[0060] When W 平6 When W ≥ 0, the warning sending module sends the sixth fault condition and its probability value W to the backend display terminal. 平6 Signal.

[0061] Furthermore, the spectrum analysis module extracts features from the sound spectra of multiple original fault conditions of the transformer to generate a preset fault condition sound spectrum matrix U. The preset fault condition sound spectrum matrix U includes multiple feature sound spectra corresponding to different faults of the transformer. The method for generating the r-th preset fault condition sound spectrum Ur includes:

[0062] The acquisition module pre-extracts the original fault sound spectrum Gr for the r-th fault condition;

[0063] The spectrum analysis module samples the original fault sound spectrum Gr and performs a Fourier transform on the sampled signal to obtain frequency-amplitude-phase information n3ω-Agrn3-Φgrn3, where n3=1,2,…,N3; Agrn3 is the amplitude value in Gr corresponding to the frequency n3ω; and Φgrn3 is the phase value in Gr corresponding to the frequency n3ω.

[0064] The spectrum analysis module analyzes and compares the original fault sound spectrum Gr of the r-th fault condition with the preset normal condition sound spectrum, obtaining the characteristic frequencies in the original fault sound spectrum Gr of the r-th fault condition that differ from the preset normal condition sound spectrum, along with their corresponding amplitude and phase information, and forming a characteristic frequency matrix H, defined as H(H1, H2, ..., Hq); H1 = h1ω - Agrh1 - Φgrh1; H2 = h2ω - Agrh2 - Φgrh2; Hq = h q ω-Agrhq -Φgrh q Where q is the total number of characteristic frequencies; h1ω is the first characteristic frequency; h2ω is the second characteristic frequency; h q ω is the qth characteristic frequency;

[0065] The spectrum analysis module linearly superimposes the characteristic frequencies within the characteristic frequency matrix H to synthesize a new frequency domain signal Hr; the spectrum analysis module performs an inverse Fourier transform on the new frequency domain signal Hr to obtain the corresponding time domain signal, which generates the r-th preset fault condition sound spectrum Ur.

[0066] Compared with the prior art, the beneficial effects of the present invention, based on the above technical solution, are as follows:

[0067] This invention uses multiple sound sensors to monitor the sound of a transformer in real time. When installing the multiple sound sensors, the distance between the sensors and the transformer is adjusted to ensure that the frequency, amplitude, and phase of the multiple sound detection signals are the same for the transformer's sound source. Thus, the two sound detection signals from each pair of symmetrical sound sensors are in phase with the transformer's sound source, but have different phases for external sound sources. By superimposing the two sound detection signals and then reducing them by half, noise reduction can be achieved for external sound sources without affecting the transformer's sound source.

[0068] The spectrum analysis module compares the preset normal sound spectrum with each set of sound detection and noise reduction signals to obtain the difference. When the difference of one set is greater than the preset standard frequency amplitude difference, it is determined that the transformer has a fault. The different levels of transformer faults are determined according to the sum of the differences of multiple sets. Maintenance personnel determine the priority level based on the fault level information displayed on the back-end terminal, which facilitates timely repair of transformers with serious faults.

[0069] The spectrum analysis module extracts the characteristic frequencies of different fault conditions of the transformer in advance to generate the corresponding sound spectrum of different fault conditions. When the transformer is determined to have a fault, the spectrum analysis module analyzes the probability value of the sound spectrum of different fault conditions contained in each group of sound detection and noise reduction signals. Based on the transformer fault conditions and their probability values ​​displayed on the back-end display terminal, maintenance personnel can prepare the corresponding maintenance tools in advance to improve maintenance efficiency. Attached Figure Description

[0070] Figure 1 This is a flowchart of the transformer fault detection method based on sound spectrum according to the present invention;

[0071] Figure 2 This is a functional diagram of the transformer fault detection system based on sound spectrum according to the present invention;

[0072] Figure 3This is a schematic diagram of the sound sensor installation in a specific embodiment of the transformer fault detection system based on sound spectrum of the present invention.

[0073] The attached diagram lists the components represented by each number as follows:

[0074] 1 - First sound sensor; 2 - Second sound sensor; 3 - Third sound sensor; 4 - Fourth sound sensor; 5 - Fifth sound sensor; 6 - Sixth sound sensor; 10 - Transformer sound source; 20 - Peripheral sound source; 100 - Transformer; 110 - Controller; 120 - On-site display terminal; 130 - Back-end display terminal. Detailed Implementation

[0075] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0076] like Figure 2 and Figure 3 As shown, a specific embodiment of the present invention discloses a transformer fault detection system based on sound spectrum, including a transformer 100, and a first sound sensor 1, a second sound sensor 2, a third sound sensor 3, a fourth sound sensor 4, a fifth sound sensor 5 and a sixth sound sensor 6 installed around the transformer 100; it also includes a controller 110, a field display terminal 120 and a background display terminal 130, the controller including a data acquisition module, a spectrum analysis module and an alarm transmission module;

[0077] The data acquisition module is electrically or communicatively connected to the first sound sensor 1, the second sound sensor 2, the third sound sensor 3, the fourth sound sensor 4, the fifth sound sensor 5, and the sixth sound sensor 6; the early warning sending module is electrically or communicatively connected to the on-site display terminal 120; and the early warning sending module is communicatively connected to the back-end display terminal 130.

[0078] like Figure 3 As shown, the first sound sensor 1 and the second sound sensor 2 form the first sound sensor group; the third sound sensor 3 and the fourth sound sensor 4 form the second sound sensor group; the fifth sound sensor 5 and the sixth sound sensor 6 form the third sound sensor group; and the first sound sensor 1, the second sound sensor 2, the third sound sensor 3, the fourth sound sensor 4, the fifth sound sensor 5 and the sixth sound sensor 6 are all at the same distance from the transformer sound source 10; there is also an external sound source 20 around the transformer 100;

[0079] The transformer fault detection system based on sound spectrum of this invention collects sound detection signals from a first sound sensor 1, a second sound sensor 2, a third sound sensor 3, a fourth sound sensor 4, a fifth sound sensor 5, and a sixth sound sensor 6 through an acquisition module, and transmits them to a spectrum analysis module. The spectrum analysis module analyzes the sound detection signals and sends instructions to an alarm sending module based on the analysis results. The alarm sending module sends adjustment and installation information for each sound sensor to a field display terminal 120 according to the instructions. Field installation personnel adjust the installation position of each sound sensor according to the information on the field display terminal 120.

[0080] like Figure 3 As shown, the first sound sensor 1 and the second sound sensor 2 of the first sound sensor group are equidistant from the transformer sound source 10, but at different distances from the peripheral sound source 20. Therefore, the first detection signal of the first sound sensor 1 and the second detection signal of the second sensor 2 are in phase with respect to the transformer sound source 10, but at different phases with respect to the peripheral sound source 20. By adding the first detection signal and the second detection signal, the signals are superimposed with the same phase with respect to the transformer sound source 10, and canceled out with the different phases with respect to the peripheral sound source 20, thus obtaining a set of sound detection and noise reduction signals that have a canceling effect on the peripheral sound source 20.

[0081] The spectrum analysis module analyzes the noise-reduced sound signals from the first, second, and third sound sensor groups and compares them with the preset normal operating condition sound spectrum. Based on the comparison results, it sends a command to the warning sending module. The warning sending module then sends a first-level warning signal, a second-level warning signal, a third-level warning signal, or a normal detection signal to the background display terminal 130 according to the command.

[0082] The spectrum analysis module analyzes the noise-reduced sound signals from the first, second, and third sound sensor groups and compares them with the sound spectra of multiple preset fault conditions. Based on the comparison results, it sends a command to the warning sending module. The warning sending module then sends the first, second, third, fourth, fifth, or sixth preset fault and its probability signal to the background display terminal 130 according to the command.

[0083] like Figure 1 As shown, this invention discloses a transformer fault detection method based on sound spectrum, implemented using the aforementioned transformer fault detection system based on sound spectrum, and includes the following steps:

[0084] Step a: Pre-install multiple sound sensors symmetrically around the transformer, with each pair of symmetrically arranged sound sensors forming a sound sensor group; the acquisition module acquires the sound detection signals from the multiple sound sensors and transmits them to the spectrum analysis module, which analyzes and compares the frequency amplitude differences between the multiple sound detection signals and adjusts the installation positions of the multiple sound sensors based on the comparison results;

[0085] Step b: The spectrum analysis module is set with the preset normal operating condition sound spectrum of the transformer. The spectrum analysis module samples and denoises the sound detection signals of each group of sound sensors to obtain a set of sound detection denoised signals. The module compares and analyzes each set of sound detection denoised signals with the preset normal operating condition sound spectrum. Based on the comparison results, the module controls the early warning sending module to send different levels of early warning signals to the background display terminal.

[0086] Step c: The spectrum analysis module extracts features from the sound spectra of multiple original fault conditions of the transformer to generate a preset fault condition sound spectrum matrix. The preset fault condition sound spectrum matrix includes feature sound spectra of multiple corresponding transformer faults.

[0087] Step d: The spectrum analysis module compares and analyzes the spectra of multiple characteristic sounds in the preset fault condition matrix with multiple sets of sound detection and noise reduction signals, and controls the early warning sending module to send the corresponding fault condition and its probability value signal to the background display terminal based on the comparison results.

[0088] Specifically, in step a, the spectrum analysis module is set with a preset standard spectrum difference range. Based on the comparison results, the warning sending module is controlled to send an adjustment signal to the on-site display terminal. Based on the adjustment signal, the installation positions of multiple sound sensors are adjusted until the frequency amplitude difference between multiple sound detection signals is within the preset standard spectrum difference range, thus completing the adjustment and installation of multiple sound sensors.

[0089] Specifically, in step b, the spectrum analysis module is set with a preset standard frequency amplitude difference and a preset frequency amplitude difference matrix B; the spectrum analysis module analyzes and compares the single-group frequency amplitude difference between each group of sound detection noise reduction signals and the preset normal working condition sound spectrum, and according to the number L of multiple single-group frequency amplitude differences greater than the preset standard spectrum difference, and the correspondence between the sum of L single-group frequency amplitude differences APL and the preset frequency amplitude difference matrix B, the warning sending module sends warning signals of different levels to the background display terminal;

[0090] Specifically, in step d, the spectrum analysis module is set with a preset standard probability value; the spectrum analysis module analyzes and compares to obtain a fault probability value matrix corresponding to different fault conditions, and filters out fault conditions in the probability value matrix that are greater than the preset standard probability value. The warning sending module sends the corresponding fault condition and its probability value signal to the background display terminal.

[0091] Specifically, in step a, the spectrum analysis module analyzes and compares the frequency amplitude differences between multiple sound detection signals, including:

[0092] The spectrum analysis module is set with a preset transformer sound source signal frequency band. Before comparing multiple sound detection signals, the spectrum analysis module performs filtering processing to filter out low-frequency components that are less than the minimum value of the preset transformer sound source signal frequency band and high-frequency components that are greater than the maximum value of the preset transformer sound source signal frequency band.

[0093] The spectrum analysis module performs Fourier transform on multiple sound detection signals to obtain the frequency-amplitude information nω-An for each sound detection signal, where the nω harmonic is the frequency information in the sound detection signal, and An is the amplitude information corresponding to the nω harmonic, n = 1, 2, ..., N;

[0094] Select one sound detection signal as the calibration signal, and analyze the frequency-amplitude difference A between the remaining sound detection signals and the calibration signal. m The calculation formula is as follows:

[0095] Where A 标n To calibrate the amplitude of the signal corresponding to an nω harmonic, A mn Let N be the amplitude of the nω harmonic corresponding to the m-th sound detection signal, N be the total number of harmonics, m = 1, 2, ..., M, and M be the total number of sound detection signals.

[0096] Specifically, in step a, the warning sending module sends an adjustment signal to the on-site display terminal based on the comparison result, including: the spectrum analysis module has a preset standard amplitude difference range (A). 01 A 02 ), where A 01 <0<A 02 And according to A m The installation position of the m-th sound sensor is adjusted based on the correspondence with the preset standard amplitude difference range.

[0097] When A 01 ≤A m ≤A 02 At that time, the early warning sending module sends a signal to the on-site display terminal that the installation of the m-th sound sensor is complete;

[0098] When A m <A 01 At that time, the early warning sending module sends a signal to the on-site display terminal to indicate that the m-th sound sensor is adjusting towards the transformer;

[0099] When A m >A 02At that time, the early warning sending module sends a signal to the on-site display terminal that the m-th sound sensor is adjusting in the direction away from the transformer.

[0100] Specifically, in step b, the spectrum analysis module samples and performs noise reduction processing on the sound detection signals of each group of sound sensors to obtain a set of sound detection noise-reduced signals, including:

[0101] Each sound sensor group includes two symmetrically arranged sound sensors, whose sound detection signals are a first detection signal and a second detection signal, respectively. The spectrum analysis module samples the first and second detection signals at a Nyquist sampling period. Then, the sampled signals of the first and second detection signals are added together by an addition circuit, and then the added detection signal is attenuated to half its original value by an attenuation circuit to obtain a set of sound detection noise-reduced signals. The calculation formula is as follows:

[0102] S (n) =(S1) (n) +S2 (n) ) / 2

[0103] S1 (n) S2 is the first detection signal. (n) S is the second detection signal. (n) This refers to a set of noise-reduced sound detection signals after noise reduction processing. Specifically, in step b above, the spectrum analysis module analyzes and compares the single-group frequency amplitude difference between each set of noise-reduced sound detection signals and the preset normal operating condition sound spectrum, including:

[0104] The spectrum analysis module performs a Fourier transform on the preset normal operating condition sound spectrum to obtain the frequency-amplitude information n1ω-A0n1, n1=1,2,…,N1 of the transformer normal operating condition sound spectrum;

[0105] The frequency-amplitude information obtained by performing a Fourier transform on the i-th group of sound detection and noise reduction signals is jω-Aij, j=1,2,…,J;

[0106] The formula for calculating the single-group frequency amplitude difference (APi) between the i-th group of sound detection and noise reduction signal and the preset normal operating condition sound spectrum is as follows:

[0107]

[0108] Where X is the number of frequencies that are the same as the preset normal operating condition sound spectrum in the i-th group of sound detection and noise reduction signals, i = 1, 2, ..., I; k1 is the frequency ratio weighting coefficient, k2 is the amplitude difference weighting coefficient, k1 > k2; A ix Let A be the amplitude of the x-th frequency corresponding to the i-th group of sound detection and noise reduction signals. 0xThe amplitude of the xth frequency corresponding to the sound spectrum under normal operating conditions is preset; x = 1, 2, ..., X.

[0109] Specifically, the spectrum analysis module has a preset standard frequency amplitude difference (AP) value. 标 The spectrum analysis module counts multiple single-group frequency amplitude differences (AP1, AP2, ..., API) that are greater than AP. 标 The number of L;

[0110] When L=0, the warning sending module sends a normal detection signal to the background display terminal;

[0111] When L≥1, calculate the sum of the frequency amplitude differences of the L single groups, APL, using the following formula:

[0112] Where APl is the l-th element greater than AP 标 The single-group frequency amplitude difference;

[0113] The spectrum analysis module has a preset frequency amplitude difference matrix B, set as B = (B1, B2, B3), where B1 is the first preset frequency amplitude difference, B2 is the second preset frequency amplitude difference, B3 is the third preset frequency amplitude difference, and B1 < B2 < B3.

[0114] When APL < B1, the warning sending module sends a first-level warning signal to the background display terminal.

[0115] When B1≤APL<B2, the warning sending module sends a second-level warning signal to the background display terminal.

[0116] When B2≤APL<B3, the warning sending module sends a third-level warning signal to the background display terminal.

[0117] Specifically, in step c, the spectrum analysis module is configured with a preset fault condition sound spectrum matrix U, defined as U(U1, U2, U3, U4, U5, U6), where:

[0118] U1 = The sound spectrum of the first preset fault condition, which corresponds to the characteristic sound spectrum of the transformer insulation breakdown discharge fault.

[0119] U2 = The sound spectrum of the second preset fault condition, which corresponds to the characteristic sound spectrum of the overheating fault caused by transformer short circuit or poor contact.

[0120] U3 = The sound spectrum of the third preset fault condition, which corresponds to the characteristic sound spectrum of a transformer open-circuit grounding fault;

[0121] U4 = The sound spectrum of the fourth preset fault condition, which corresponds to the characteristic sound spectrum of transformer overvoltage or lightning wave intrusion fault.

[0122] U5 = The sound spectrum of the fifth preset fault condition, which corresponds to the characteristic sound spectrum of faults such as transformer bushing cracks or poor contact of equipment line cards.

[0123] U6 = The sound spectrum of the sixth preset fault condition, corresponding to the characteristic sound spectrum of a transformer core loosening fault.

[0124] Specifically, in step d, the spectrum analysis module analyzes and compares to obtain a fault probability matrix corresponding to different fault conditions, including:

[0125] The spectrum analysis module performs a Fourier transform on the sound spectrum of the r-th preset fault condition in Ur to obtain the corresponding spectrum-amplitude information n2ω-Arn2, n2=1,2,…,N2; r=1,2,3,4,5,6;

[0126] The spectrum analysis module analyzes the frequency-amplitude information jω-Aij of the i-th group of sound detection and noise reduction signals, which contains the probability value W of the r-th preset fault condition sound spectrum n2ω-Arn2. ir The calculation formula is as follows:

[0127]

[0128] Where Y is the number of frequencies in the sound spectrum of the i-th group of sound detection and noise reduction signal and the r-th preset fault condition sound; σ0 is the difference benchmark coefficient; K1 is the frequency ratio weighting coefficient, K2 is the amplitude difference weighting coefficient, K1>K2, and K1+K2=1; A iy Let A be the amplitude of the y-th frequency corresponding to the i-th group of sound detection and noise reduction signal. ry The amplitude of the y-th frequency corresponding to the preset fault condition sound spectrum is given by y = 1, 2, ..., Y;

[0129] The spectrum analysis module calculates the average probability value W of multiple sets of sound detection and noise reduction signals corresponding to the sound spectrum of the r-th preset fault condition. 平r The calculation formula is as follows:

[0130]

[0131] Where I is the total number of groups of sound detection and noise reduction signals, i = 1, 2, ..., I.

[0132] In another specific embodiment, the spectrum analysis module analyzes the l-th spectrum greater than AP. 标 The probability value W of the l-th group of sound detection and noise reduction signal corresponding to the single-group amplitude difference Apl contains the r-th preset fault condition sound spectrum n2ω-Arn. lr The calculation formula is as follows:

[0133]

[0134] Where Z is the number of frequencies in the sound spectrum of the l-th group of sound detection noise reduction signal and the r-th preset fault condition sound spectrum that are identical; σ0 is the difference benchmark coefficient; K1 is the frequency ratio weighting coefficient, K2 is the amplitude difference weighting coefficient, K1>K2, and K1+K2=1; A lz Let A be the amplitude of the z-th frequency corresponding to the l-th group of sound detection and noise reduction signals. rz The amplitude of the z-th frequency corresponding to the preset fault condition sound spectrum is given for the r-th fault condition, where z = 1, 2, ..., Z;

[0135] Furthermore, the spectrum analysis module calculates the average probability value W of the L groups of sound detection and noise reduction signals corresponding to the r-th preset fault condition sound spectrum. 平r The calculation formula is as follows:

[0136]

[0137] Where L is the value of the single-group frequency amplitude difference that is greater than AP. 标 The number of elements is L, where l = 1, 2, ..., L.

[0138] Specifically, this leads to the fault probability matrix W corresponding to different fault conditions. 平 ,

[0139] Setting W 平 (W 平1 W 平2 W 平3 W 平4 W 平5 W 平6 ), where W 平1 W represents the average probability value corresponding to the first fault condition. 平2 W represents the average probability value corresponding to the second fault condition. 平3 W represents the average probability value corresponding to the third fault condition. 平4 W represents the average probability value corresponding to the fourth fault condition. 平5 W represents the average probability value corresponding to the fifth fault condition. 平6 This represents the average probability value corresponding to the sixth fault condition.

[0140] Specifically, in step d, fault conditions within the probability value matrix that are greater than a preset standard probability value are selected, and the early warning sending module sends the corresponding fault condition and its probability value signal to the backend display terminal, including:

[0141] The spectrum analysis module has a preset standard probability value W0, and a fault probability value matrix W is selected. 平 For values ​​greater than W0, send the corresponding fault condition and its probability value signal;

[0142] When W 平1When W ≥ 0, the warning sending module sends the first fault condition and the probability value W of the first fault condition to the backend display terminal. 平1 Signal;

[0143] When W 平2 When W ≥ 0, the warning sending module sends the second fault condition and its probability value W to the backend display terminal. 平2 Signal;

[0144] When W 平3 When W ≥ 0, the warning sending module sends the third fault and the probability value W of the third fault condition to the backend display terminal. 平3 Signal;

[0145] When W 平4 When W ≥ 0, the warning sending module sends the fourth fault condition and its probability value W to the backend display terminal. 平4 Signal;

[0146] When W 平5 When W ≥ 0, the warning sending module sends the fifth fault condition and its probability value W to the backend display terminal. 平5 Signal;

[0147] When W 平6 When W ≥ 0, the warning sending module sends the sixth fault condition and its probability value W to the backend display terminal. 平6 Signal.

[0148] Specifically, in step c, the spectrum analysis module extracts features from the sound spectra of multiple original fault conditions of the transformer to generate a preset fault condition sound spectrum matrix U. The preset fault condition sound spectrum matrix U includes multiple feature sound spectra corresponding to different faults of the transformer. The method for generating the r-th preset fault condition sound spectrum Ur includes:

[0149] The acquisition module pre-extracts the original fault sound spectrum Gr for the r-th fault condition;

[0150] The spectrum analysis module samples the original fault sound spectrum Gr and performs a Fourier transform on the sampled signal to obtain frequency-amplitude-phase information n3ω-Agrn3-Φgrn3, where n3=1,2,…,N3; Agrn3 is the amplitude value in Gr corresponding to the frequency n3ω; and Φgrn3 is the phase value in Gr corresponding to the frequency n3ω.

[0151] The spectrum analysis module analyzes and compares the original fault sound spectrum Gr of the r-th fault condition with the preset normal condition sound spectrum, obtains the characteristic frequencies in the original fault sound spectrum Gr of the r-th fault condition that are different from the preset normal condition sound spectrum and their corresponding amplitude and phase information, and forms a characteristic frequency matrix H, which is set as H(H1, H2, ..., Hq).

[0152] H1=h1ω-Agrh1-Φgrh1; H2=h2ω-Agrh2-Φgrh2; Hq=h q ω-Agrh q -Φgrh q ;

[0153] Where q is the total number of characteristic frequencies; h1ω is the first characteristic frequency; h2ω is the second characteristic frequency; h q ω is the qth characteristic frequency;

[0154] The spectrum analysis module performs linear superposition of the characteristic frequencies within the characteristic frequency matrix H to synthesize a new frequency domain signal Hr;

[0155] The spectrum analysis module performs an inverse Fourier transform on the new frequency domain signal Hr to obtain the corresponding time domain signal, which generates the r-th preset fault condition sound spectrum Ur.

[0156] The transformer fault detection method based on sound spectrum of the present invention can send the fault level signal of transformer 100, the fault condition signal and its probability value signal obtained by analysis to the background display terminal 130 through the controller 110. This makes it convenient for maintenance personnel to repair transformer 100 according to the fault severity level based on the information received by the background display terminal 130, and to prepare maintenance tools corresponding to different fault conditions in advance, thereby improving maintenance efficiency.

[0157] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A transformer fault detection method based on sound spectrum, characterized in that, Includes the following steps: Step a: Pre-install multiple sound sensors symmetrically around the transformer, with each pair of symmetrically arranged sound sensors forming a sound sensor group; the acquisition module acquires the sound detection signals from the multiple sound sensors and transmits them to the spectrum analysis module, which analyzes and compares the frequency amplitude differences between the multiple sound detection signals and adjusts the installation positions of the multiple sound sensors based on the comparison results; Step b: The spectrum analysis module is set with the preset normal operating condition sound spectrum of the transformer. The spectrum analysis module samples and denoises the sound detection signals of each group of sound sensors to obtain a set of sound detection denoised signals. The module compares and analyzes each set of sound detection denoised signals with the preset normal operating condition sound spectrum. Based on the comparison results, the module controls the early warning sending module to send different levels of early warning signals to the background display terminal. Step c: The spectrum analysis module extracts features from the sound spectra of multiple original fault conditions of the transformer to generate a preset fault condition sound spectrum matrix. The preset fault condition sound spectrum matrix includes feature sound spectra of multiple corresponding transformer faults. Step d: The spectrum analysis module compares and analyzes the spectra of multiple characteristic sounds in the preset fault condition matrix with multiple sets of sound detection and noise reduction signals, and controls the early warning sending module to send the corresponding fault condition and its probability value signal to the background display terminal based on the comparison results. In step a, the spectrum analysis module is set with a preset standard spectrum difference range. Based on the comparison results, the warning sending module is controlled to send an adjustment signal to the on-site display terminal. The installation positions of multiple sound sensors are adjusted according to the adjustment signal until the frequency amplitude difference between multiple sound detection signals is within the preset standard spectrum difference range, thus completing the adjustment and installation of multiple sound sensors. In step b, the spectrum analysis module is configured with a preset standard frequency amplitude difference and a preset frequency amplitude difference matrix B. The spectrum analysis module analyzes and compares the single-group frequency amplitude difference between each group of sound detection noise reduction signals and the preset normal operating condition sound spectrum. Based on the number L of multiple single-group frequency amplitude differences greater than the preset standard spectrum difference, and the correspondence between the sum of L single-group frequency amplitude differences APL and the preset frequency amplitude difference matrix B, the warning sending module sends different levels of warning signals to the background display terminal. In step d, a preset standard probability value is set in the spectrum analysis module; The spectrum analysis module analyzes and compares to obtain the fault probability value matrix corresponding to different fault conditions, and filters out fault conditions in the probability value matrix that are greater than the preset standard probability value. The warning sending module sends the corresponding fault condition and its probability value signal to the background display terminal. In step a, the spectrum analysis module analyzes and compares the frequency amplitude differences between multiple sound detection signals, including: The spectrum analysis module is set with a preset transformer sound source signal frequency band. Before comparing multiple sound detection signals, the spectrum analysis module performs filtering processing to filter out low-frequency components that are less than the minimum value of the preset transformer sound source signal frequency band and high-frequency components that are greater than the maximum value of the preset transformer sound source signal frequency band. The spectrum analysis module performs Fourier transform on multiple sound detection signals to obtain the frequency-amplitude information nω-An for each sound detection signal, where nω is the frequency information in the sound detection signal and An is the amplitude information corresponding to the nω, n=1,2,…,N; Select one sound detection signal as the calibration signal, and analyze the frequency-amplitude difference A between the remaining sound detection signals and the calibration signal. m The calculation formula is as follows: A m = Where A 标n To calibrate the amplitude of the signal corresponding to an nω harmonic, A mn Let N be the amplitude of the nω harmonic corresponding to the m-th sound detection signal, where N is the total number of harmonics, and m = 1, 2, ..., M; In step a, the control of the early warning sending module to send an adjustment signal to the field display terminal based on the comparison result includes: The spectrum analysis module has a preset standard amplitude difference range (A). 01 A 02 ), where A 01 <0<A 02 And according to A m The installation position of the m-th sound sensor is adjusted based on the correspondence with the preset standard amplitude difference range. When A 01 ≤A m ≤A 02 At that time, the early warning sending module sends a signal to the on-site display terminal that the installation of the m-th sound sensor is complete; When A m <A 01 At that time, the early warning sending module sends a signal to the on-site display terminal to indicate that the m-th sound sensor is adjusting towards the transformer; When A m >A 02 At that time, the early warning sending module sends a signal to the on-site display terminal that the m-th sound sensor is adjusting in the direction away from the transformer.

2. The transformer fault detection method based on sound spectrum according to claim 1, characterized in that, In step b, the spectrum analysis module samples and performs noise reduction processing on the sound detection signals of each group of sound sensors to obtain a set of sound detection noise-reduced signals, including: Each sound sensor group includes two symmetrically arranged sound sensors, whose sound detection signals are a first detection signal and a second detection signal, respectively. The spectrum analysis module samples the first and second detection signals at a Nyquist sampling period. Then, the sampled signals of the first and second detection signals are added together by an addition circuit, and then the added detection signal is attenuated to half its original value by an attenuation circuit to obtain a set of sound detection noise-reduced signals. The calculation formula is as follows: S （n） =(S1 （n） +S2 （n） ) / 2 S1 （n） S2 is the first detection signal. （n） S is the second detection signal. （n） This is a set of noise-reduced signals for detecting noise reduction.

3. The transformer fault detection method based on sound spectrum according to claim 2, characterized in that, In step b, the spectrum analysis module analyzes and compares the single-group frequency amplitude difference between each group of sound detection and noise reduction signals and the preset normal operating condition sound spectrum, including: The spectrum analysis module performs a Fourier transform on the preset normal operating condition sound spectrum to obtain the frequency-amplitude information n1ω-A0n1, n1=1,2,…,N1 of the transformer's normal operating condition sound spectrum; The frequency-amplitude information obtained by performing a Fourier transform on the i-th group of sound detection and noise reduction signals is jω-Aij, where j=1,2,…,J; The formula for calculating the single-group frequency amplitude difference (APi) between the i-th group of sound detection and noise reduction signal and the preset normal operating condition sound spectrum is as follows: APi= Where X is the number of frequencies that are the same as the preset normal operating condition sound spectrum in the i-th group of sound detection and noise reduction signals, where i = 1, 2, ..., I; k1 is the frequency ratio weighting coefficient, k2 is the amplitude difference weighting coefficient, and k1 > k2; A ix Let A be the amplitude of the x-th frequency corresponding to the i-th group of sound detection and noise reduction signals. 0x Let x be the amplitude of the xth frequency corresponding to the sound spectrum under normal operating conditions; where x = 1, 2, ..., X.

4. The transformer fault detection method based on sound spectrum according to claim 3, characterized in that, The spectrum analysis module has a preset standard frequency amplitude difference (AP) value. 标 The spectrum analysis module counts multiple single-group frequency amplitude differences (AP1, AP2, ..., API) that are greater than AP. 标 The number of L; When L=0, the warning sending module sends a normal detection signal to the background display terminal; When L≥1, calculate the sum of the frequency amplitude differences of the L single groups, APL, using the following formula: APL= Where AP1 is the l-th element greater than AP 标 The single-group frequency amplitude difference; The spectrum analysis module has a preset frequency amplitude difference matrix B, set as B = (B1, B2, B3), where B1 is the first preset frequency amplitude difference, B2 is the second preset frequency amplitude difference, and B3 is the third preset frequency amplitude difference, with B1 < B2 < B3. When APL < B1, the warning sending module sends a first-level warning signal to the background display terminal. When B1≤APL<B2, the warning sending module sends a second-level warning signal to the background display terminal. When B2≤APL<B3, the warning sending module sends a third-level warning signal to the background display terminal.

5. The transformer fault detection method based on sound spectrum according to claim 4, characterized in that, In step c, the spectrum analysis module is configured with a preset fault condition sound spectrum matrix U, where U (U1, U2, U3, U4, U5, U6) is defined, wherein: U1 = The sound spectrum of the first preset fault condition, which corresponds to the characteristic sound spectrum of the transformer insulation breakdown discharge fault. U2 = Second preset fault condition sound spectrum, corresponding to the characteristic sound spectrum of overheating faults caused by transformer short circuit or poor contact; U3 = The sound spectrum of the third preset fault condition, which corresponds to the characteristic sound spectrum of a transformer open-circuit grounding fault; U4 = The sound spectrum of the fourth preset fault condition, which corresponds to the characteristic sound spectrum of transformer overvoltage or lightning wave intrusion fault. U5 = Fifth preset fault condition sound spectrum, corresponding to the characteristic sound spectrum of transformer bushing crack damage or equipment line card poor contact fault; U6 = The sound spectrum of the sixth preset fault condition, which corresponds to the characteristic sound spectrum of a transformer core loosening fault.

6. The transformer fault detection method based on sound spectrum according to claim 5, characterized in that, In step d, the spectrum analysis module analyzes and compares to obtain a fault probability matrix corresponding to different fault conditions, including: The spectrum analysis module performs a Fourier transform on the sound spectrum of the r-th preset fault condition in Ur to obtain the corresponding spectrum-amplitude information n2ω-Arn2, n2=1,2,…,N2; r=1,2,3,4,5,6; The spectrum analysis module analyzes the frequency-amplitude information jω-Aij of the i-th group of sound detection and noise reduction signals, finding that it contains the probability value W of the r-th preset fault condition sound spectrum n2ω-Arn2. ir The calculation formula is as follows: W ir = ， Where Y is the number of frequencies in the sound spectrum of the i-th group of sound detection and noise reduction signal and the r-th preset fault condition sound; σ0 is the difference benchmark coefficient; K1 is the frequency ratio weighting coefficient, K2 is the amplitude difference weighting coefficient, K1>K2, and K1+K2=1; A iy Let A be the amplitude of the y-th frequency corresponding to the i-th group of sound detection and noise reduction signal. ry The amplitude of the y-th frequency corresponding to the preset fault condition sound spectrum is given by y=1,2,…,Y; The spectrum analysis module calculates the average probability value W of multiple sets of sound detection and noise reduction signals corresponding to the sound spectrum of the r-th preset fault condition. 平r The calculation formula is as follows: W 平r = Where I is the total number of groups of sound detection and noise reduction signals, and i = 1, 2, ..., I; This leads to the fault probability matrix W corresponding to different fault conditions. 平 , Setting W 平 (W) 平1 W 平2 W 平3 W 平4 W 平5 W 平6 ), where W 平1 W represents the average probability value corresponding to the first fault condition. 平2 W represents the average probability value corresponding to the second fault condition. 平3 W represents the average probability value corresponding to the third fault condition. 平4 W represents the average probability value corresponding to the fourth fault condition. 平5 W represents the average probability value corresponding to the fifth fault condition. 平6 This represents the average probability value corresponding to the sixth fault condition.

7. The transformer fault detection method based on sound spectrum according to claim 6, characterized in that, In step d, fault conditions with probabilities greater than a preset standard probability value are filtered out from the probability value matrix. The early warning sending module sends the corresponding fault condition and its probability value signal to the background display terminal, including: The spectrum analysis module has a preset standard probability value W0, and a fault probability value matrix W is selected. 平 For values ​​greater than W0, send the corresponding fault condition and its probability value signal; When W 平1 When W ≥ 0, the warning sending module sends the first fault condition and the probability value W of the first fault condition to the backend display terminal. 平1 Signal; When W 平2 When W ≥ 0, the warning sending module sends the second fault condition and its probability value W to the backend display terminal. 平2 Signal; When W 平3 When W ≥ 0, the warning sending module sends the third fault and the probability value W of the third fault condition to the backend display terminal. 平3 Signal; When W 平4 When W ≥ 0, the warning sending module sends the fourth fault condition and its probability value W to the backend display terminal. 平4 Signal; When W 平5 When W ≥ 0, the warning sending module sends the fifth fault condition and its probability value W to the backend display terminal. 平5 Signal; When W 平6 When W ≥ 0, the warning sending module sends the sixth fault condition and its probability value W to the backend display terminal. 平6 Signal.

8. The transformer fault detection method based on sound spectrum according to claim 6, characterized in that, In step c, the spectrum analysis module extracts features from the sound spectra of multiple original fault conditions of the transformer to generate a preset fault condition sound spectrum matrix U. The preset fault condition sound spectrum matrix U includes multiple feature sound spectra corresponding to different faults of the transformer. The method for generating the r-th preset fault condition sound spectrum Ur includes: The acquisition module pre-extracts the original fault sound spectrum Gr for the r-th fault condition; The spectrum analysis module samples the original fault sound spectrum Gr and performs a Fourier transform on the sampled signal to obtain frequency-amplitude-phase information n3ω-Agrn3-Φgrn3, where n3=1,2,…,N3; Agrn3 is the amplitude value in Gr corresponding to the frequency n3ω; and Φgrn3 is the phase value in Gr corresponding to the frequency n3ω. The spectrum analysis module analyzes and compares the original fault sound spectrum Gr of the r-th fault condition with the preset normal condition sound spectrum, obtains the characteristic frequencies in the original fault sound spectrum Gr of the r-th fault condition that are different from the preset normal condition sound spectrum and their corresponding amplitude and phase information, and forms a characteristic frequency matrix H, which is set as H(H1,H2,...,Hq). H1=h1ω-Agrh1-Φgrh1;H2=h2ω-Agrh2-Φgrh2;Hq=h q ω-Agrh q -Φgrh q ; Where q is the total number of characteristic frequencies; h1ω is the first characteristic frequency; h2ω is the second characteristic frequency; h q ω is the qth characteristic frequency; The spectrum analysis module linearly superimposes the characteristic frequencies within the characteristic frequency matrix H to synthesize a new frequency domain signal Hr. The spectrum analysis module performs an inverse Fourier transform on the new frequency domain signal Hr to obtain the corresponding time domain signal, which generates the r-th preset fault condition sound spectrum Ur.

Images