Transformer fault detection model training method, fault diagnosis method and related device

Abstract

The disclosure provides a transformer fault detection model training method, a fault diagnosis method, and a related device. The method comprises: obtaining an initial voiceprint signal of a transformer, and a fault type corresponding to the initial voiceprint signal; preprocessing the initial voiceprint signal by using a wavelet packet analysis method, so as to obtain an input signal, and according to the input signal and the fault type, establishing an input signal data set; according to a preset feature extraction algorithm, performing feature extraction on a first input signal in a training data set to obtain a first voiceprint feature; using the first voiceprint feature and a first fault type corresponding to the first input signal to train an initial detection model, so as to obtain a first training result; according to the first training result and the fault type, determining a loss function; on the basis of the loss function, using a back propagation algorithm to iteratively adjust weight values of the initial detection model, until the loss function converges, so as to obtain a fault detection model. In the present disclosure, accurate transformer fault detection is realized by utilizing the obtained fault detection model.

Description

Transformer fault detection model training method, fault diagnosis method and related equipment Technical Field

[0001] The present disclosure relates to the field of transformer fault detection, and in particular to a transformer fault detection model training method, a fault diagnosis method, and related equipment. Background Art

[0002] Transformers are crucial components of power systems, and their proper operation is crucial for ensuring their stability and reliability. However, due to long-term operation, aging, overload, and other factors, transformers are prone to various faults, such as winding deformation, poor contact, and short circuits. These faults not only affect the normal operation of the power system but can even lead to serious safety incidents. Therefore, timely diagnosis and location of transformer faults are crucial.

[0003] Traditional transformer fault diagnosis methods mainly include electrical testing and oil sample analysis, but these methods often require a lot of time and manpower, and in some cases it is difficult to accurately diagnose the fault type and location.

[0004] In view of this, how to achieve accurate detection of transformer faults has become an important problem to be solved.

[0005] In view of this, the purpose of the present disclosure is to propose a transformer fault detection model training method, a fault diagnosis method and related equipment to solve or partially solve the above problems.

[0006] Based on the above objectives, a first aspect of the present disclosure provides a transformer fault detection model training method, the method comprising:

[0007] Obtaining an initial voiceprint signal of the transformer and a fault type corresponding to the initial voiceprint signal;

[0008] Preprocessing the initial voiceprint signal using a wavelet packet analysis method to obtain an input signal, and establishing a signal data set based on the input signal and the fault type, wherein the input signal data set includes a training data set;

[0009] Extracting features from the first input signal in the training data set according to a preset feature extraction algorithm to obtain a first voiceprint feature corresponding to the first input signal;

[0010] Training an initial detection model using the first voiceprint feature and a first fault type corresponding to the first input signal to obtain a first training result;

[0011] Determine a loss function according to the first training result and the fault type;

[0012] Based on the loss function, a back-propagation algorithm is used to iteratively adjust the weight value of the initial detection model until the loss function converges, thereby obtaining a fault detection model, so as to regulate the transformer to be detected according to the fault diagnosis result obtained by performing fault detection on the transformer to be detected using the fault detection model.

[0013] Based on the same inventive concept, the second aspect of the present disclosure proposes a transformer fault diagnosis method, comprising:

[0014] Acquire an initial target voiceprint signal of a target transformer, and preprocess the initial target voiceprint signal to obtain a target voiceprint signal;

[0015] The target voiceprint signal is input into a fault detection model obtained by a training method based on a transformer fault detection model, and the fault diagnosis result corresponding to the target transformer is output through processing by the fault detection model.

[0016] Based on the same inventive concept, the third aspect of the present disclosure proposes a transformer fault detection model training device, comprising:

[0017] a signal acquisition module configured to acquire an initial voiceprint signal of the transformer and a fault type corresponding to the initial voiceprint signal;

[0018] a preprocessing module configured to preprocess the initial voiceprint signal using a wavelet packet analysis method to obtain an input signal, and establish a signal data set according to the input signal and the fault type, wherein the input signal data set includes a training data set;

[0019] a feature extraction module configured to extract features of the first input signal in the training data set according to a preset feature extraction algorithm to obtain a first voiceprint feature corresponding to the first input signal;

[0020] a model training module configured to train an initial detection model using the first voiceprint feature and a first fault type corresponding to the first input signal to obtain a first training result;

[0021] a loss function determination module, configured to determine a loss function according to the first training result and the fault type;

[0022] The weight adjustment module is configured to iteratively adjust the weight value of the initial detection model based on the loss function using a back propagation algorithm until the loss function converges to obtain a fault detection model.

[0023] Based on the same inventive concept, the fourth aspect of the present disclosure provides a transformer fault diagnosis device, comprising:

[0024] a preprocessing module configured to obtain an initial target voiceprint signal of a target transformer, and preprocess the initial target voiceprint signal to obtain a target voiceprint signal;

[0025] The diagnosis result output module is configured to input the target voiceprint signal into the fault detection model obtained by the training method of the transformer fault detection model, and output the fault diagnosis result corresponding to the target transformer through processing by the fault detection model.

[0026] Based on the same inventive concept, the fifth aspect of the present disclosure proposes an electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable by the processor. When executing the computer program, the processor implements the transformer fault detection model training method or transformer fault diagnosis method as described above.

[0027] Based on the same inventive concept, the sixth aspect of the present disclosure proposes a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute the transformer fault detection model training method described above, such as the transformer fault diagnosis method described above.

[0028] As can be seen from the foregoing, the present disclosure proposes a transformer fault detection model training method, fault diagnosis method, and related equipment. The method obtains an initial voiceprint signal of a transformer and the fault type corresponding to the initial voiceprint signal. The initial voiceprint signal is preprocessed using wavelet packet analysis to obtain an input signal, reducing interference from ambient noise. A preset feature extraction algorithm is used to extract features from a first input signal in a training dataset to obtain a first voiceprint feature corresponding to the first input signal. This first voiceprint feature is then used to train a model. The initial detection model is then trained using the first voiceprint feature and the first fault type corresponding to the first input signal. After obtaining a first training result, a loss function is determined based on the first training result and the fault type. This loss function is used to determine when model training is complete. Based on this loss function, a backpropagation algorithm is used to iteratively adjust the weights of the initial detection model until the loss function converges. This results in a fault detection model. The trained fault detection model is then used to identify the voiceprint signal of the input model, thereby more accurately determining whether the transformer corresponding to the voiceprint signal is faulty. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the present disclosure or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0030] FIG1 is a flow chart of a transformer fault detection model training method according to an embodiment of the present disclosure;

[0031] FIG2 is a flow chart of a transformer fault diagnosis method according to an embodiment of the present disclosure;

[0032] FIG3 is a structural block diagram of a transformer fault detection model training device according to an embodiment of the present disclosure;

[0033] FIG4 is a structural block diagram of a transformer fault diagnosis device according to an embodiment of the present disclosure;

[0034] FIG5 is a schematic structural diagram of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0035] In order to make the objectives, technical solutions and advantages of the present disclosure more clearly understood, the present disclosure is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings.

[0036] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present disclosure should have the usual meanings understood by people with ordinary skills in the field to which the present disclosure belongs. The "first", "second" and similar words used in the embodiments of the present disclosure do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative position relationships. When the absolute position of the described object changes, the relative position relationship may also change accordingly.

[0037] The terms used in this disclosure are explained as follows:

[0038] DSP: Digital Signal Processor (DSP) is a microprocessor with a special structure that uses digital signals to process large amounts of information.

[0039] wav: Waveform sound file (WAV) is a standard digital audio file developed by Microsoft specifically for Windows.

[0040] PCM encoding: Pulse Code Modulation (PCM) is a coding method used in digital communications. It samples analog signals such as voice and images at regular intervals to discretize them. The sampled values ​​are rounded off and quantized by layer unit. The sampled values ​​are then represented by a set of binary codes to represent the amplitude of the sampled pulses.

[0041] ADC: Analog-to-digital converter (ADC) is a type of device used to convert continuous signals in analog form into discrete signals in digital form.

[0042] Based on the above description, this embodiment proposes a transformer fault detection model training method, as shown in FIG1 , the method comprising:

[0043] Step 101: Acquire an initial voiceprint signal of a transformer and a fault type corresponding to the initial voiceprint signal.

[0044] In a specific implementation, at least one voiceprint sensor is installed around the transformer to collect voiceprint signals generated during transformer operation. The voiceprint sensor collects the transformer's voiceprint signals at a preset sampling frequency, and the collected voiceprint signals belong to the same frequency range. In this embodiment, the preset sampling frequency is 96 kHz, and the frequency range is 0-40 kHz.

[0045] The data format of the voiceprint signal is a single audio duration of 10s, an audio sampling rate of 48kHz, a sampling accuracy of 16bit, a single channel, an encoding method of PCM encoding, and a wav format.

[0046] Obtain an initial voiceprint signal of the transformer and a fault type corresponding to the initial voiceprint signal. The fault type is a pre-marked transformer fault type, and the transformer fault type includes at least one of the following: winding fault, bushing fault, core fault, gas protection fault, transformer fire, and tap changer fault.

[0047] Step 102: pre-process the initial voiceprint signal using wavelet packet analysis to obtain an input signal, and establish an input signal data set according to the input signal and the fault type, wherein the input signal data set includes a training data set.

[0048] In specific implementations, the acquired initial voiceprint signal is preprocessed using wavelet packet analysis to obtain an input signal. The preprocessing method includes at least one of the following: denoising or data enhancement. Denoising includes at least one of the following: segmentation, framing, windowing, and adaptive filtering. Data enhancement includes at least one of the following: segmentation, noise addition, and voice tuning.

[0049] An input signal data set is established based on the input signal obtained through preprocessing and the fault type corresponding to the device. The data in the input signal data set is randomly divided according to a preset ratio to obtain a training data set and a test data set.

[0050] Exemplarily, the number of data items in the input signal data set is 6000, the preset ratio is 5 to 1, the amount of data in the training data set is 5000, and the amount of data in the test data set is 1000.

[0051] Step 103: extract features of the first input signal in the training data set according to a preset feature extraction algorithm to obtain a first voiceprint feature corresponding to the first input signal.

[0052] In a specific implementation, a preset feature extraction algorithm is used to extract features from a first input signal in a training data set to obtain a first voiceprint feature corresponding to the first input signal. The first voiceprint feature is a characteristic parameter reflecting the operating state of the transformer, and the feature includes at least one of the following: a spectral feature, a cepstrum feature, and a linear predictive coding coefficient.

[0053] Step 104: Train an initial detection model using the first voiceprint feature and the first fault type corresponding to the first input signal to obtain a first training result.

[0054] In a specific implementation, the initial detection model is trained using the acquired first voiceprint feature and the first fault type corresponding to the first input signal, and a first training result is output through the model, wherein the first training result indicates whether the transformer corresponding to the first voiceprint feature is faulty.

[0055] Step 105: Determine a loss function according to the first training result and the fault type.

[0056] Step 106: Based on the loss function, a back propagation algorithm is used to iteratively adjust the weight value of the initial detection model until the loss function converges, thereby obtaining a fault detection model, so as to regulate the transformer to be detected according to the fault diagnosis result obtained by performing fault detection on the transformer to be detected using the fault detection model.

[0057] In specific implementation, a loss function is determined based on the output first training result and the fault type, and the weight values ​​of the initial detection model are updated using a back-propagation algorithm. Training is iterated until the loss function converges, and the training ends to obtain a fault detection model.

[0058] The backpropagation algorithm is a method for correcting the annotation model of each layer. It starts from the high-level annotation results, and according to the annotation features and the dependency relationship between feature functions, reversely corrects the weight of the feature function that the annotation results depend on and is less than a preset threshold. Through the hierarchical progressive relationship, the weight of the feature function with low credibility in each layer is reduced.

[0059] The back propagation algorithm is used to update the weight value of the initial detection model, which is expressed by the formula:

[0060] delta_weights=-alpha*(gradient*weights)+regularization*weights.

[0061] Among them, alpha is the learning rate, gradient is the gradient, and weights is the weight of the network.

[0062] Through machine learning and deep learning algorithms, different voiceprint features can be more accurately identified and distinguished. This allows for a higher degree of differentiation between voiceprint features associated with different faults, reducing the possibility of misidentification. Furthermore, AI-based voiceprint recognition technology is highly robust and can automatically adapt to various environmental noise and other interference factors, reducing their impact on recognition accuracy.

[0063] Through the above scheme, the transformer's initial voiceprint signal and the fault type corresponding to the initial voiceprint signal are obtained. The initial voiceprint signal is preprocessed using wavelet packet analysis to obtain an input signal, reducing interference from environmental noise. Feature extraction is performed on the first input signal in the training data set using a preset feature extraction algorithm to obtain a first voiceprint feature corresponding to the first input signal. This first voiceprint feature is then used for subsequent model training. The initial detection model is trained using the first voiceprint feature and the first fault type corresponding to the first input signal. After obtaining the first training result, a loss function is determined based on the first training result and the fault type. This loss function is used to determine when model training is complete. Based on the loss function, a backpropagation algorithm is used to iteratively adjust the weights of the initial detection model until the loss function converges. This results in a fault detection model. The trained fault detection model is then used to identify the voiceprint signal of the input model to determine whether the transformer corresponding to the voiceprint signal is faulty, resulting in a more accurate judgment.

[0064] Specifically, the fault detection model detects faults in the transformer under test, including DC bias, partial discharge, transmission jamming, internal looseness, overload, overexcitation, or winding deformation. Based on the fault diagnosis results obtained by the fault detection model, the transformer's output voltage and / or relay on / off control are adjusted to correct the fault and ensure normal operation of the transformer.

[0065] In some embodiments, step 102 specifically includes:

[0066] Step 1021: digitally process the initial voiceprint signal to obtain an initial digital signal.

[0067] In a specific implementation, after obtaining the initial voiceprint signal, the initial voiceprint signal is first digitized using a digital signal processor to convert the initial voiceprint signal into an initial digital signal, wherein the initial digital signal is signal data that can be recognized by a computer.

[0068] Step 1022: De-noise the initial digital signal using wavelet packet analysis to obtain a standard digital signal.

[0069] In specific implementation, the initial digital signal is subjected to denoising processing by wavelet packet analysis to obtain a clean digital signal, thereby removing noise interference in the initial digital signal.

[0070] In some embodiments, missing values ​​are filled in the clean digital signal to obtain a standard digital signal. The missing values ​​are filled using an interpolation method, wherein the interpolation method is a method of estimating the value of an unknown point by fitting a curve or polynomial through known data points, and then using the curve or polynomial to estimate the value of the unknown point.

[0071] In this embodiment, linear interpolation is used to fill missing values, which is expressed by the formula: y=y1+(x-x1)*(y2-y1) / (x2-x1);

[0072] Among them, (x1, y1) and (x2, y2) are known adjacent data points, and (x, y) is the unknown point that needs to be estimated.

[0073] Step 1023: normalize the standard digital signal to obtain an input signal.

[0074] In a specific implementation, the obtained standard digital signal is normalized to obtain an input signal. In this embodiment, the normalization method adopts the maximum-minimum normalization method, which maps the data to the range of [0, 1]. The formula is expressed as: y = (x-min) / (max-min);

[0075] Where x is the original data, max and min are the maximum and minimum values ​​in the data, respectively. After min-max normalization, the data is mapped to the range [0, 1].

[0076] In some embodiments, step 1022 specifically includes:

[0077] Step 10221: Decompose the initial digital signal according to a preset scale to obtain a plurality of initial digital sub-signals.

[0078] In specific implementation, the initial digital signal is decomposed according to a preset scale. In this embodiment, the preset scale is preferably 3, wherein the wavelet packet analysis method is expressed by the formula: y(n)=x(n)+d(n);

[0079] Where y(n) is the initial digital signal. Let y(n) = a0(k), and take the mother wavelet function as the Daubechies wavelet. Perform multi-scale decomposition on a0(k) to obtain multiple initial digital sub-signals, where the initial digital sub-signals are signals at different frequencies. After wavelet transform, the initial digital signal a0(k) is decomposed into:

[0080] Where N = 3, d in the formula N (k) has the same meaning as d(n) above.

[0081] Step 10222: For each initial digital sub-signal, perform a spectrum subtraction operation on the initial digital sub-signal based on a preset window width to obtain a digital sub-signal.

[0082] In a specific implementation, each initial digital sub-signal is regarded as an independent signal for spectral subtraction operation. When performing spectral subtraction operation, a preset window width is used. In this embodiment, the window lengths of a1(k), a2(k), a3(k) and d3(k) are 64, 128, 256 and 512 respectively, and the window shape is a Hamming window.

[0083] Step 10223: perform summing processing on all the obtained digital sub-signals to obtain a standard digital signal.

[0084] In specific implementation, all the obtained digital sub-signals are combined and added to obtain a standard digital signal.

[0085] In some embodiments, step 103 specifically includes:

[0086] Step 1031 : For each first input signal in the training data set: perform feature extraction on the first input signal using a preset feature extraction algorithm to obtain time domain features and frequency domain features corresponding to the input signal.

[0087] In specific implementations, a preset feature extraction algorithm is used to extract time-domain and frequency-domain features for each first input signal in the training dataset. Time-domain feature extraction involves analyzing the first input signal using a preset algorithm to obtain voiceprint features reflecting the transformer's operating status. These voiceprint features are time-domain features. Frequency-domain feature extraction involves performing spectral analysis on the first input signal to extract spectral parameters reflecting the transformer's operating status.

[0088] Step 1032: Perform weighted calculation on the time domain features and the frequency domain features to obtain a first voiceprint feature corresponding to the first input signal.

[0089] In a specific implementation, the time domain features and frequency domain features are weighted and calculated using preset weight values ​​to obtain the first voiceprint features corresponding to the first input signal. For example, the weight value corresponding to the time domain features is 0.8, and the weight value corresponding to the frequency domain features is 0.2.

[0090] In some embodiments, step 1031 specifically includes:

[0091] Step 10311: Calculate the energy value of the first input signal using an energy detection algorithm.

[0092] Step 10312: Acquire the signal length and signal amplitude of the first input signal, determine the duration of the first input signal according to the signal length, and determine the amplitude of the first input signal according to the signal amplitude.

[0093] In a specific implementation, an energy detection algorithm is used to calculate the energy value of the first input signal. The code of the energy detection algorithm is: energy=sum(abs(signal)^2);

[0094] Wherein, energy is the energy value, and singal is the first input signal.

[0095] The signal amplitude and signal length of the first input signal are obtained, and the amplitude corresponding to the first input signal is calculated using the signal amplitude. The code of the amplitude calculation method is: amplitude_distribution=histogram(abs(signal));

[0096] Among them, amplitude_distribution is the amplitude.

[0097] The duration of the first input signal is calculated by obtaining the signal length of the first input signal. The code for the duration is: duration=length(signal);

[0098] Among them, duration is the duration.

[0099] Step 10313: Input the energy value of the first input signal, the duration of the first input signal, and the amplitude of the first input signal into a preset feature extraction algorithm to obtain the time domain features of the first input signal.

[0100] During specific implementation, the calculated energy value of the first input signal, the duration of the first input signal, and the amplitude of the first input signal are input into a preset feature extraction algorithm, and the feature extraction algorithm is used to output the time domain features corresponding to the first input signal, wherein the preset feature extraction algorithm can be a pre-trained algorithm that obtains the time domain features of the first input signal based on the energy value of the first input signal, the duration of the first input signal, and the amplitude of the first input signal.

[0101] Step 10314: Perform Fourier transform on the first input signal to obtain a frequency spectrum corresponding to the first input signal.

[0102] In a specific implementation, the first input signal is a time domain signal, and the first input signal is converted, that is, from a time domain signal to a frequency domain signal. In this embodiment, the conversion method is to use Fourier transform to obtain the spectrum corresponding to the first input signal.

[0103] Step 10315: Obtain the maximum value and the minimum value in the frequency spectrum corresponding to the first input signal, input the maximum value and the minimum value into a preset feature extraction algorithm, and obtain the frequency domain features of the first input signal.

[0104] During specific implementation, peak extraction and valley extraction are performed on the spectrum corresponding to the first input signal, i.e., the maximum and minimum values ​​in the spectrum corresponding to the first input signal are obtained. The maximum and minimum values ​​are input into a preset feature extraction algorithm, and the feature extraction algorithm is used to output frequency domain features corresponding to the first input signal. The preset feature extraction algorithm may be a pre-trained algorithm that obtains frequency domain features of the first input signal based on the maximum and minimum values ​​of the first input signal.

[0105] In some embodiments, the input signal dataset further includes a test dataset, and the method further includes:

[0106] Step A: extracting features from the second input signal in the test data set to obtain a second voiceprint feature.

[0107] During specific implementation, second input data in the test data set is obtained, and feature extraction is performed on the second input data to obtain a second voiceprint feature.

[0108] The method for extracting the second voiceprint feature includes:

[0109] An energy value of the second input signal is calculated using an energy detection algorithm, a signal length and a signal amplitude of the second input signal are obtained, a duration of the second input signal is determined based on the signal length, and an amplitude of the second input signal is determined based on the signal amplitude. The energy value of the second input signal, the duration of the second input signal, and the amplitude of the second input signal are input into a preset feature extraction algorithm to obtain a time domain feature of the second input signal.

[0110] Performing a Fourier transform on the second input signal to obtain a frequency spectrum corresponding to the second input signal. Obtaining a maximum value and a minimum value in the frequency spectrum corresponding to the second input signal, and inputting the maximum value and the minimum value into a preset feature extraction algorithm to obtain a frequency domain feature of the second input signal.

[0111] A weighted calculation is performed on the time domain feature and the frequency domain feature to obtain a second voiceprint feature corresponding to the second hundredth input signal.

[0112] Step B: inputting the second voiceprint feature into the fault detection model to obtain a second training result.

[0113] Step C: Compare the second training result with the fault type corresponding to the second voiceprint feature, and output a comparison result, wherein the comparison result is used to indicate whether the second training result and the fault type corresponding to the second voiceprint feature are the same.

[0114] During specific implementation, the second voiceprint feature is input into a fault detection model, and the second training result is obtained through processing by the fault detection model.

[0115] If the transformer corresponding to the second voiceprint feature is faulty, obtain the fault type of the transformer corresponding to the second voiceprint feature, determine whether the second training result indicates that the transformer is faulty, and compare the fault type corresponding to the second training result with the fault type corresponding to the second voiceprint feature to determine whether the second training result is the same as the fault type to determine the accuracy of the model.

[0116] In some embodiments, the transformer corresponding to the second voiceprint feature may not be faulty. Therefore, after obtaining the second training result, it is determined whether the second training result indicates that the transformer is not faulty to determine the accuracy of the model.

[0117] The four indicators of accuracy, false detection rate, missed detection rate and average calculation time corresponding to the fault detection model are statistically calculated based on the test data set, and the performance of the fault detection model is judged based on the four indicators to determine whether it can be put into use in the future.

[0118] Another embodiment of the present disclosure provides a transformer fault diagnosis method, which applies the fault detection model obtained in the above embodiment. The method is shown in FIG2 and includes:

[0119] Step 201: Acquire an initial target voiceprint signal of a target transformer, and preprocess the initial target voiceprint signal to obtain a target voiceprint signal.

[0120] In a specific implementation, an initial target voiceprint signal of a target transformer is obtained, where the target transformer is a transformer to be detected, and the initial target voiceprint signal is preprocessed to obtain a target voiceprint signal.

[0121] The process of preprocessing the initial target voiceprint signal specifically includes:

[0122] The initial target voiceprint signal is digitally processed to obtain an initial target digital signal, and the initial target digital signal is decomposed according to a preset scale to obtain multiple initial target digital sub-signals.

[0123] For each initial target digital sub-signal, a spectral subtraction operation is performed on the initial target digital sub-signal based on a preset window width to obtain a target digital sub-signal. All the obtained target digital sub-signals are summed to obtain a standard target digital signal. The standard target digital signal is normalized to obtain a target voiceprint signal.

[0124] In step 202, the target voiceprint signal is input into a fault detection model obtained by a training method of a transformer fault detection model, and the fault diagnosis result corresponding to the target transformer is output through processing by the fault detection model.

[0125] In a specific implementation, a fault detection model trained based on the above embodiment is obtained, the target voiceprint signal is input into the fault detection model, and the fault detection model processes the signal to output a fault diagnosis result. The fault diagnosis result indicates whether the target transformer is faulty and, if so, the corresponding fault type.

[0126] Through the above solution, using the trained fault detection model, it is possible to use the collected voiceprint information of the target transformer to identify and analyze the voiceprint information to determine whether the target transformer is faulty. This makes the judgment more convenient and improves the accuracy of the judgment. At the same time, if the target transformer is determined to be faulty, the fault type of the target transformer will be output in the fault diagnosis results, allowing the user to gain a preliminary understanding of the transformer fault and take appropriate measures.

[0127] It should be noted that the method of the embodiments of the present disclosure can be performed by a single device, such as a computer or server. The method of the embodiments of the present disclosure can also be applied in a distributed scenario, where multiple devices cooperate to perform the method. In such a distributed scenario, one of the multiple devices may only perform one or more steps of the method of the embodiments of the present disclosure, and the multiple devices will interact with each other to complete the method.

[0128] It should be noted that the above description is limited to some embodiments of the present disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in an order different from that described in the above embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0129] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present disclosure also provides a transformer fault detection model training device.

[0130] Referring to FIG3 , FIG3 is a transformer fault detection model training device according to an embodiment, comprising:

[0131] The signal acquisition module 301 is configured to acquire an initial voiceprint signal of the transformer and a fault type corresponding to the initial voiceprint signal;

[0132] A preprocessing module 302 is configured to preprocess the initial voiceprint signal using a wavelet packet analysis method to obtain an input signal, and establish a signal data set based on the input signal and the fault type, wherein the input signal data set includes a training data set;

[0133] The feature extraction module 303 is configured to extract features of the first input signal in the training data set according to a preset feature extraction algorithm to obtain a first voiceprint feature corresponding to the first input signal;

[0134] A model training module 304 is configured to train an initial detection model using the first voiceprint feature and the first fault type corresponding to the first input signal to obtain a first training result;

[0135] A loss function determination module 305 is configured to determine a loss function according to the first training result and the fault type;

[0136] The weight adjustment module 306 is configured to iteratively adjust the weight values ​​of the initial detection model based on the loss function using a back propagation algorithm until the loss function converges to obtain a fault detection model.

[0137] In some embodiments, the pre-processing module 302 specifically includes:

[0138] a digital processing unit configured to digitally process the initial voiceprint signal to obtain an initial digital signal;

[0139] a denoising processing unit configured to perform denoising processing on the initial digital signal using a wavelet packet analysis method to obtain a standard digital signal;

[0140] The normalization processing unit is configured to perform normalization processing on the standard digital signal to obtain an input signal.

[0141] In some embodiments, the denoising processing unit specifically includes:

[0142] a signal decomposition subunit, configured to decompose the initial digital signal according to a preset scale to obtain a plurality of initial digital sub-signals;

[0143] a spectrum subtraction operation subunit, configured to perform a spectrum subtraction operation on each initial digital sub-signal based on a preset window width to obtain a digital sub-signal;

[0144] The summing processing subunit is configured to perform summing processing on all the obtained digital sub-signals to obtain a standard digital signal.

[0145] In some embodiments, the feature extraction module 303 specifically includes:

[0146] A feature extraction unit is configured to: for each first input signal in the training data set: perform feature extraction on the first input signal using a preset feature extraction algorithm to obtain a time domain feature and a frequency domain feature corresponding to the first input signal;

[0147] The weighted calculation unit is configured to perform weighted calculation on the time domain feature and the frequency domain feature to obtain a first voiceprint feature corresponding to the first input signal.

[0148] In some embodiments, the feature extraction unit specifically includes:

[0149] an energy value calculation subunit, configured to calculate an energy value of the first input signal using an energy detection algorithm;

[0150] an amplitude calculation subunit, configured to obtain a signal length and a signal amplitude of the first input signal, determine a duration of the first input signal according to the signal length, and determine an amplitude of the first input signal according to the signal amplitude;

[0151] a time domain feature extraction subunit, configured to input the energy value of the first input signal, the duration of the first input signal, and the amplitude of the first input signal into a preset feature extraction algorithm to obtain a time domain feature of the first input signal;

[0152] a spectrum determination subunit, configured to perform Fourier transform on the first input signal to obtain a spectrum corresponding to the first input signal;

[0153] The frequency domain feature extraction subunit is configured to obtain the maximum value and the minimum value in the frequency spectrum corresponding to the first input signal, input the maximum value and the minimum value into a preset feature extraction algorithm, and obtain the frequency domain feature of the first input signal.

[0154] In some embodiments, the apparatus further comprises a testing module, wherein the testing module comprises:

[0155] a feature extraction unit configured to extract features from a second input signal in the test data set to obtain a second voiceprint feature;

[0156] a training result generating unit, configured to input the second voiceprint feature into the fault detection model to obtain a second training result;

[0157] A result comparison unit is configured to compare the second training result with the fault type corresponding to the second voiceprint feature, and output a comparison result, wherein the comparison result is used to indicate whether the second training result and the fault type corresponding to the second voiceprint feature are the same.

[0158] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, the present disclosure also provides a transformer fault diagnosis device.

[0159] Referring to FIG4 , FIG4 is a transformer fault diagnosis device according to an embodiment, comprising:

[0160] The signal acquisition module 401 is configured to acquire an initial target voiceprint signal of a target transformer, and preprocess the initial target voiceprint signal to obtain a target voiceprint signal;

[0161] The result output module 402 is configured to input the target voiceprint signal into a fault detection model, and output a fault diagnosis result corresponding to the target transformer through processing by the fault detection model.

[0162] In another implementation example, the above-mentioned transformer fault detection model training device includes: a processor, wherein the processor is used to execute the above-mentioned program modules stored in the memory, including: a signal acquisition module 301, a preprocessing module 302, a feature extraction module 303, a model training module 304, a loss function determination module 305, a weight adjustment module 306, a signal acquisition module 401 and a result output module 402.

[0163] For the convenience of description, the above devices are described as being functionally divided into various modules. Of course, when implementing the present disclosure, the functions of each module can be implemented in the same or multiple software and / or hardware.

[0164] The device of the above embodiment is used to implement the corresponding transformer fault detection model training method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0165] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, the present disclosure also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and runnable on the processor, wherein when the processor executes the program, the transformer fault detection model training method described in any of the above embodiments is implemented.

[0166] FIG5 shows a more specific schematic diagram of the hardware structure of an electronic device provided in this embodiment. The device may include: a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040 are communicatively connected to each other within the device via the bus 1050.

[0167] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0168] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage devices, dynamic storage devices, etc. The memory 1020 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0169] The input / output interface 1030 is used to connect input / output modules to implement information input and output. The input / output modules can be configured as components within the device (not shown in the figure) or can be externally connected to the device to provide corresponding functions. Input devices may include a keyboard, mouse, touch screen, microphone, various sensors, etc., and output devices may include a display, speaker, vibrator, indicator light, etc.

[0170] The communication interface 1040 is used to connect to a communication module (not shown) to enable communication between the device and other devices. The communication module can communicate via a wired method (such as USB, network cable, etc.) or a wireless method (such as mobile network, WiFi, Bluetooth, etc.).

[0171] The bus 1050 comprises a pathway for transmitting information between the various components of the device (eg, the processor 1010 , the memory 1020 , the input / output interface 1030 , and the communication interface 1040 ).

[0172] It should be noted that although the above device only shows the processor 1010, the memory 1020, the input / output interface 1030, the communication interface 1040, and the bus 1050, in a specific implementation, the device may also include other components necessary for normal operation. In addition, it will be understood by those skilled in the art that the above device may only include the components necessary to implement the embodiments of this specification, and does not necessarily include all the components shown in the figure.

[0173] The electronic device of the above embodiment is used to implement the corresponding transformer fault detection model training method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0174] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present disclosure also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the transformer fault detection model training method described in any of the above embodiments.

[0175] The computer-readable media of this embodiment include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0176] The computer instructions stored in the storage medium of the above embodiment are used to enable the computer to execute the transformer fault detection model training method described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0177] It is understandable that before using the technical solutions of each embodiment of the present disclosure, the type, scope of use, usage scenarios, etc. of the personal information involved will be informed to the user in an appropriate manner, and the user's authorization will be obtained.

[0178] For example, in response to a user's active request, a prompt message is sent to the user to clearly inform the user that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the electronic device, application, server, storage medium, or other software or hardware that performs the operation of the disclosed technical solution based on the prompt message.

[0179] As an optional but non-limiting implementation, in response to a user's active request, the prompt information may be sent to the user in the form of a pop-up window, in which the prompt information may be presented in text form. Furthermore, the pop-up window may also contain a selection control for the user to select "agree" or "disagree" to provide personal information to the electronic device.

[0180] It is understandable that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of the present disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of the present disclosure.

[0181] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present disclosure (including the claims) is limited to these examples. Within the scope of the present disclosure, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present disclosure as described above, which are not provided in detail for the sake of simplicity.

[0182] In addition, to simplify the description and discussion, and so as not to obscure the embodiments of the present disclosure, known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided figures. In addition, devices may be shown in the form of block diagrams to avoid obscuring the embodiments of the present disclosure, and this also takes into account the fact that the details of the implementation of these block diagram devices are highly dependent on the platform on which the embodiments of the present disclosure are to be implemented (i.e., these details should be fully within the purview of those skilled in the art). Where specific details (e.g., circuits) are set forth to describe exemplary embodiments of the present disclosure, it will be apparent to those skilled in the art that the embodiments of the present disclosure may be implemented without these specific details or with variations in these specific details. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0183] Although the present disclosure has been described in conjunction with specific embodiments thereof, many alternatives, modifications, and variations of these embodiments will be apparent to those skilled in the art based on the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may use the embodiments discussed.

[0184] The embodiments of the present disclosure are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present disclosure should be included in the scope of protection of the present disclosure.

Claims

1. A transformer fault detection model training method, characterized in that: include: Obtaining an initial voiceprint signal of the transformer and a fault type corresponding to the initial voiceprint signal; Preprocessing the initial voiceprint signal by wavelet packet analysis to obtain an input signal, and establishing an input signal data set according to the input signal and the fault type, wherein the input signal data set includes a training data set; Extracting features of a first input signal in the training data set according to a preset feature extraction algorithm to obtain a first voiceprint feature corresponding to the first input signal; Training an initial detection model using the first voiceprint feature and a first fault type corresponding to the first input signal to obtain a first training result; Determine a loss function according to the first training result and the first fault type; Based on the loss function, a back propagation algorithm is used to iteratively adjust the weight value of the initial detection model until the loss function converges, thereby obtaining a fault detection model, so as to regulate the transformer to be detected according to the fault diagnosis result obtained by performing fault detection on the transformer to be detected according to the fault detection model.

2. The method according to claim 1, characterized in that The method of preprocessing the initial voiceprint signal by wavelet packet analysis to obtain an input signal includes: Digitally process the initial voiceprint signal to obtain an initial digital signal; Using wavelet packet analysis to perform denoising on the initial digital signal to obtain a standard digital signal; The standard digital signal is normalized to obtain an input signal.

3. The method according to claim 2, characterized in that The method of performing denoising on the initial digital signal by using the wavelet packet analysis method to obtain a standard digital signal includes: Decomposing the initial digital signal according to a preset scale to obtain a plurality of initial digital sub-signals; For each initial digital sub-signal, performing a spectral subtraction operation on the initial digital sub-signal based on a preset window width to obtain a digital sub-signal; All the obtained digital sub-signals are summed up to obtain a standard digital signal.

4. The method according to claim 1, characterized in that: The step of extracting features from the first input signal in the training data set according to a preset feature extraction algorithm to obtain a first voiceprint feature corresponding to the first input signal includes: For each first input signal in the training dataset: Using a preset feature extraction algorithm to extract features from the first input signal, to obtain time domain features and frequency domain features corresponding to the first input signal; The time domain feature and the frequency domain feature are weightedly calculated to obtain a first voiceprint feature corresponding to the first input signal.

5. The method according to claim 4, characterized in that The extracting features of the first input signal by using a preset feature extraction algorithm to obtain time domain features and frequency domain features corresponding to the first input signal includes: Calculating the energy value of the first input signal using an energy detection algorithm; Acquire a signal length and a signal amplitude of the first input signal, determine a duration of the first input signal according to the signal length, and determine an amplitude of the first input signal according to the signal amplitude; Inputting the energy value of the first input signal, the duration of the first input signal, and the amplitude of the first input signal into a preset feature extraction algorithm to obtain a time domain feature of the first input signal; Performing Fourier transform on the first input signal to obtain a frequency spectrum corresponding to the first input signal; A maximum value and a minimum value in the frequency spectrum corresponding to the first input signal are obtained, and the maximum value and the minimum value are input into a preset feature extraction algorithm to obtain a frequency domain feature of the first input signal.

6. The method according to claim 1, characterized in that The input signal data set also includes a test data set. The method further comprises: Performing feature extraction on a second input signal in the test data set to obtain a second voiceprint feature; Inputting the second voiceprint feature into the fault detection model to obtain a second training result; The second training result is compared with the fault type corresponding to the second voiceprint feature, and a comparison result is output, wherein the comparison result is used to indicate whether the second training result is the same as the fault type corresponding to the second voiceprint feature.

7. A transformer fault diagnosis method, characterized in that: include: Acquire an initial target voiceprint signal of a target transformer, and preprocess the initial target voiceprint signal to obtain a target voiceprint signal; The target voiceprint signal is input into a fault detection model obtained by the training method of the transformer fault detection model according to any one of claims 1 to 6, and the fault diagnosis result corresponding to the target transformer is output through processing by the fault detection model.

8. A transformer fault detection model training device, characterized in that: include: A signal acquisition module, configured to acquire an initial voiceprint signal of the transformer and a fault type corresponding to the initial voiceprint signal; A preprocessing module is configured to preprocess the initial voiceprint signal using a wavelet packet analysis method to obtain an input signal, and establish a signal data set according to the input signal and the fault type, wherein the input signal data set includes a training data set; a feature extraction module, configured to extract features of a first input signal in the training data set according to a preset feature extraction algorithm, to obtain a first voiceprint feature corresponding to the first input signal; a model training module, configured to train an initial detection model using the first voiceprint feature and a first fault type corresponding to the first input signal to obtain a first training result; a loss function determination module, configured to determine a loss function according to the first training result and the fault type; The weight adjustment module is configured to iteratively adjust the weight value of the initial detection model based on the loss function by using a back propagation algorithm until the loss function converges to obtain a fault detection model.

9. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the training method for a transformer fault detection model as described in any one of claims 1 to 6 or the transformer fault diagnosis method as described in claim 7 is implemented.

10. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the transformer fault detection model training method according to any one of claims 1 to 6 or the transformer fault diagnosis method according to claim 7.

Images