Large transformer noise monitoring analysis and fault early warning method and system

Abstract

The invention discloses a noise monitoring analysis and fault early warning method and system for a large transformer, and relates to the technical field of noise monitoring, and the method comprises the following steps: collecting and preprocessing a coverage broadband noise signal through a sensor; a CNN-LSTM model is established for the collected noise signals through a deep learning module; and setting an early warning threshold value, performing judgment through the CNN-LSTM model, and triggering an early warning signal. According to the invention, key features can be automatically extracted from transformer noise, real-time analysis is carried out through the deep learning model, multiple fault types can be rapidly identified without manual intervention, early warning can be timely given out, power failure accidents caused by faults are reduced, the stability of power supply is improved, intelligent upgrading of a power equipment monitoring technology is promoted, and the power supply efficiency is improved. And a key technical support is provided for digital and intelligent transformation of the power industry.

Description

Technical Field

[0001] The present invention relates to the technical field of noise monitoring, and in particular to a method and system for noise monitoring, analysis and fault early warning of a large transformer. Background Art

[0002] Large transformers are important equipment in hydropower stations, and their operating status directly affects the stability and safety of power station operation. The noise generated by the transformer during operation contains a wealth of equipment status information, such as partial discharge, mechanical looseness and other fault signs. Traditional noise monitoring methods rely on manual monitoring and simple threshold (decibel size) judgment, which cannot capture the tiny noise changes caused by abnormal conditions or defects inside the transformer in real time. It is difficult to adapt to the complex and changeable noise environment and the need to identify early weak fault signals, resulting in the early stage of the fault being difficult to be timely and accurately identified and warned. Therefore, it is crucial to develop a method that can intelligently monitor and deeply learn transformer noise, and predict and alarm transformer abnormal conditions and potential faults as early as possible to improve the efficiency and safety of hydropower station operation and maintenance. Summary of the invention

[0003] In view of the problems existing in the existing large-scale transformer noise monitoring and analysis and fault warning and system, the present invention is proposed.

[0004] Therefore, the problem to be solved by the present invention is that it is impossible to capture in real time the tiny noise changes caused by abnormal conditions or defects inside the transformer, and it is difficult to adapt to the complex and changeable noise environment and the recognition requirements of early weak fault signals.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] In a first aspect, an embodiment of the present invention provides a large transformer noise monitoring and analysis and fault warning method, which comprises the following steps:

[0007] The wide-band noise signal is collected by sensors and pre-processed;

[0008] A CNN-LSTM model is established for the collected noise signal through the deep learning module;

[0009] Set the warning threshold, make judgments through the CNN-LSTM model, and trigger the warning signal.

[0010] As a preferred solution of the large transformer noise monitoring, analysis and fault warning method of the present invention, the step of collecting noise signals includes:

[0011] Multi-dimensional acoustic sensors are arranged around the transformer to form a dense monitoring network, covering wide-band noise signal collection to ensure the comprehensiveness and high fidelity of noise data;

[0012] Each group of sensors is equipped with an anti-interference hardware filter to reduce the impact of environmental noise;

[0013] The acquisition frequency range is set to 20Hz to 200kHz, covering the audible range of the human ear and part of the ultrasonic region, and is used to collect ultrasonic noise generated when a transformer has an internal fault;

[0014] The collected noise signals include the time series sound pressure level, sound intensity and spectrum components.

[0015] As a preferred solution of the large transformer noise monitoring and analysis and fault warning method of the present invention, the step of preprocessing the noise signal includes original signal denoising, signal framing and windowing, and feature analysis and feature value extraction of the noise signal;

[0016] The steps of denoising the original signal include:

[0017] Subtract the estimated ambient noise component from the original signal, retaining the noise signal unique to transformer operation;

[0018] When the data processing center receives the main transformer noise, it first pre-processes the noise data and denoises the main transformer noise data using the wavelet threshold denoising method. The specific steps are as follows:

[0019] First, perform wavelet transform on the main transformer noise signal x(t) to obtain a set of wavelet decomposition coefficients W j,k , by decomposing the wavelet coefficients W j,k Perform threshold processing to obtain denoised wavelet coefficients Finally, using the estimated wavelet coefficients Perform wavelet reconstruction to obtain the estimated signal This is the signal after denoising;

[0020] The steps of signal framing and windowing processing include:

[0021] A continuous main transformer noise signal is divided into multiple frames with a length of L. Then the nth frame can be expressed as x n (t), each frame is then processed individually;

[0022] According to the time domain distribution of transformer noise characteristics and fault features, the main transformer noise signal is divided into 50ms frames, and each frame signal is shifted 25ms relative to the previous frame on the time axis;

[0023] Apply a window function ω(t) to each frame signal, then the windowed signal y n (t) = x n (t)·ω(t-nL).

[0024] As a preferred solution of the large transformer noise monitoring and analysis and fault warning method of the present invention, the steps of characteristic analysis and characteristic value extraction of the noise signal include:

[0025] For the windowed signal y n (t) Perform discrete Fourier transform (DFT) to convert the signal from the time domain to the frequency domain and obtain the spectrum:

[0026]

[0027] Then calculate the power spectral density (PSD) and take the square of the absolute value as the energy estimate: n (k)=|Y n (k)| 2 ;

[0028] Apply a filter to the power spectrum to obtain the energy of each band

[0029] In the formula, H m (k) represents the mth Mel filter;

[0030] Taking the logarithm of the energy of each band, we get

[0031] Perform discrete cosine transform on the logarithmic energy to obtain the Mel frequency cepstrum coefficients:

[0032]

[0033] Finally, the features are combined to obtain a feature vector

[0034]

[0035] The feature vector Fn is used as the vector input.

[0036] As a preferred solution of the large transformer noise monitoring, analysis and fault warning method of the present invention, the construction method of the CNN-LSTM model is as follows:

[0037] First, the eigenvector F n Each feature in is normalized to eliminate the scale differences between different features;

[0038] Serialize the continuous feature vectors to form a time series data set. Each sample consists of T continuous feature vectors. The sample is represented by {F n-T+1 ,F n-T+2 ,…,F n}

[0039] Divide the serialized data set into training set, validation set and test set;

[0040] The CNN-LSTM model includes an input layer, a convolution layer, a flattening layer, an LSTM layer, a fully connected layer, and an output layer.

[0041] As a preferred solution of the large transformer noise monitoring and analysis and fault warning method of the present invention, the construction method of the CNN-LSTM model also includes fault diagnosis, and the specific steps include:

[0042] Use the fault data and normal noise data set in the historical data to train the CNN-LSTM model;

[0043] After training, the CNN-LSTM model can transform the input feature vector F n Map it to the corresponding fault category or predict its health status, and use the Softmax function in the output layer to obtain the probability distribution of the fault type, where the category corresponding to the highest probability is the diagnosis result.

[0044] As a preferred solution of the large transformer noise monitoring and analysis and fault early warning method of the present invention, the early warning signal sending method includes:

[0045] Set the warning threshold θ and make a judgment:

[0046] When the CNN-LSTM model predicts the probability of failure P alert Exceeding this threshold, that is, P alert >θ, then the warning is triggered;

[0047] Input new data into the CNN-LSTM model in real time and calculate P at each moment alert , when P is found alert When the preset threshold is exceeded, a warning signal is immediately issued and measures are taken in advance to prevent failures.

[0048] In a second aspect, an embodiment of the present invention provides a large transformer noise monitoring analysis and fault warning system, which includes a signal acquisition module, a preprocessing module, a threshold setting module, and a warning trigger module;

[0049] The signal acquisition module is responsible for receiving the original noise signal from the sensor and performing preliminary digital processing;

[0050] The preprocessing module preprocesses the collected noise signal, which may include filtering, denoising, normalization, feature extraction and other steps to prepare data for subsequent deep learning model use;

[0051] The threshold setting module allows the user or the system to set a warning threshold as needed, and this threshold is used to determine whether the noise signal is abnormal;

[0052] The warning trigger module is used to trigger the warning signal when the CNN-LSTM model determines that the noise signal exceeds the warning threshold, and notify relevant personnel through sound, light, text message, and email.

[0053] In a third aspect, an embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the processor executes the computer program, any step of the above-mentioned large transformer noise monitoring, analysis and fault warning method is implemented.

[0054] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the above-mentioned large transformer noise monitoring, analysis and fault warning method is implemented.

[0055] The beneficial effects of the present invention are as follows: it can automatically extract key features from transformer noise and perform real-time analysis through a deep learning model, quickly identify multiple fault types without human intervention, issue early warnings in a timely manner, reduce power outages caused by faults, and improve the stability of power supply; and the CNN-LSTM hybrid model can efficiently capture the spatiotemporal features in noise signals, especially the LSTM unit is good at processing long-term dependencies in time series, so that the system can still maintain high diagnostic accuracy and sensitivity when facing complex and subtle fault signs, which is significantly improved compared with traditional methods; finally, it promotes the intelligent upgrading of power equipment monitoring technology and provides key technical support for the digital and intelligent transformation of the power industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0057] Figure 1 This is the structural diagram of the large transformer noise monitoring analysis and fault warning method. DETAILED DESCRIPTION

[0058] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0059] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0060] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0061] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0062] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0063] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0064] Example 1

[0065] Reference Figure 1, which is the first embodiment of the present invention, provides a large transformer noise monitoring and analysis and fault warning method, comprising the following steps:

[0066] S1. The wide-band noise signal is collected through the sensor and pre-processed.

[0067] The steps for collecting noise signals include:

[0068] Multi-dimensional acoustic sensors are arranged around the transformer to form a dense monitoring network, covering wide-band noise signal collection to ensure the comprehensiveness and high fidelity of noise data;

[0069] Each group of sensors is equipped with an anti-interference hardware filter to reduce the impact of environmental noise;

[0070] The acquisition frequency range is set to 20Hz to 200kHz, covering the audible range of the human ear and part of the ultrasonic region, and is used to collect ultrasonic noise generated when a transformer has an internal fault;

[0071] The collected noise signals include the time series sound pressure level, sound intensity and spectrum components.

[0072] The steps of preprocessing the noise signal include original signal denoising, signal framing and windowing, and feature analysis and feature value extraction of the noise signal;

[0073] The steps of denoising the original signal include:

[0074] Subtract the estimated ambient noise component from the original signal, retaining the noise signal unique to transformer operation;

[0075] When the data processing center receives the main transformer noise, it first pre-processes the noise data and denoises the main transformer noise data using the wavelet threshold denoising method. The specific steps are as follows:

[0076] First, perform wavelet transform on the main transformer noise signal x(t) to obtain a set of wavelet decomposition coefficients W j,k , by decomposing the wavelet coefficients W j,k Perform threshold processing to obtain denoised wavelet coefficients Finally, using the estimated wavelet coefficients Perform wavelet reconstruction to obtain the estimated signal This is the signal after denoising;

[0077] The steps of signal framing and windowing processing include:

[0078] A continuous main transformer noise signal is divided into multiple frames with a length of L. Then the nth frame can be expressed as x n (t), each frame is then processed individually;

[0079] According to the time domain distribution of transformer noise characteristics and fault features, the main transformer noise signal is divided into 50ms frames, and each frame signal is shifted 25ms relative to the previous frame on the time axis;

[0080] To reduce spectrum leakage caused by sudden signal interruption and maintain continuity and smooth transition between adjacent frames, a window function ω(t) is applied to each frame signal, then the windowed signal y n (t) = x n (t)·ω(t-nL).

[0081] The steps of feature analysis and feature value extraction of the noise signal include:

[0082] For the windowed signal y n (t) Perform discrete Fourier transform (DFT) to convert the signal from the time domain to the frequency domain and obtain the spectrum:

[0083]

[0084] Then calculate the power spectral density (PSD) and take the square of the absolute value as the energy estimate: n (k)=|Y n (k)| 2 ;

[0085] Apply a filter to the power spectrum to obtain the energy of each band

[0086] In the formula, H m (k) represents the mth Mel filter;

[0087] Taking the logarithm of the energy of each band, we get

[0088] Perform discrete cosine transform on the logarithmic energy to obtain the Mel frequency cepstrum coefficients:

[0089]

[0090] Finally, the features are combined to obtain a feature vector

[0091]

[0092] The feature vector Fn is used as the vector input.

[0093] S2. Establish a CNN-LSTM model for the collected noise signal through the deep learning module.

[0094] The construction method of the CNN-LSTM model is:

[0095] First, each feature in the feature vector is normalized to eliminate the scale differences between different features;

[0096] Serialize the continuous feature vectors to form a time series data set. Each sample consists of continuous feature vectors. The sample is represented as

[0097] Divide the serialized data set into training set, validation set and test set;

[0098] The CNN-LSTM model includes an input layer, a convolution layer, a flattening layer, an LSTM layer, a fully connected layer, and an output layer.

[0099] Input layer: Receives a sequence of feature vectors with a shape of , where is the batch size, is the sequence length (i.e., the number of feature vectors), and is the dimension of the feature vector.

[0100] Convolutional layer: Use multiple convolution kernels to perform convolution operations on the input to extract local spatiotemporal patterns in features. Each layer is followed by an activation function (ReLU) and a maximum pooling layer to reduce the dimension and extract more abstract features.

[0101] Flattening layer: Flatten the CNN output and feed it into the LSTM layer.

[0102] LSTM layer: receives flattened features and is used to learn temporal dependencies in the sequence.

[0103] Fully connected layer: A fully connected layer is added after LSTM to further integrate information, convert sequence features into category predictions, summarize and classify the information of the entire sequence, and the last layer usually outputs the probability distribution of the fault category.

[0104] Output layer: Use the softmax function to output fault classification.

[0105] The method for constructing the CNN-LSTM model also includes fault diagnosis, and the specific steps include:

[0106] Use the fault data and normal noise data set in the historical data to train the CNN-LSTM model;

[0107] After training, the CNN-LSTM model can map the input feature vector to the corresponding fault category or predict its health status. The Softmax function is used in the output layer to obtain the probability distribution of the fault type, where the category corresponding to the highest probability is the diagnosis result.

[0108] S3. Set the warning threshold, make judgments through the CNN-LSTM model, and trigger the warning signal.

[0109] Methods for sending early warning signals include:

[0110] Set the warning threshold θ and make a judgment:

[0111] When the CNN-LSTM model predicts the probability of failure P alert Exceeding this threshold, that is, P alert >θ, then the warning is triggered;

[0112] Input new data into the CNN-LSTM model in real time and calculate P at each moment alert , when P is found alert When the preset threshold is exceeded, a warning signal is immediately issued and measures are taken in advance to prevent failures.

[0113] In this embodiment, if the CNN-LSTM model predicts that the probability of a device failing in the future is greater than 70%, the device is considered to be in a high-risk state and the early warning mechanism is triggered. After receiving the early warning, the maintenance team can quickly check the relevant equipment and perform necessary maintenance work to avoid potential major accidents. At the same time, the early warning threshold can also be adjusted according to the actual situation to adapt to the ever-changing operating environment.

[0114] In summary, it is possible to automatically extract key features from transformer noise and perform real-time analysis through a deep learning model. It is possible to quickly identify multiple fault types without human intervention, issue early warnings in a timely manner, reduce power outages caused by faults, and improve the stability of power supply. The CNN-LSTM hybrid model can efficiently capture the spatiotemporal features in noise signals, especially the LSTM unit is good at processing long-term dependencies in time series, so that the system can maintain high diagnostic accuracy and sensitivity when faced with complex and subtle fault signs, which is significantly improved compared to traditional methods. Finally, it promotes the intelligent upgrade of power equipment monitoring technology and provides key technical support for the digital and intelligent transformation of the power industry.

[0115] Example 2

[0116] On the basis of the first embodiment, this embodiment further provides a large transformer noise monitoring analysis and fault warning system, including a signal acquisition module, a preprocessing module, a threshold setting module, and a warning trigger module;

[0117] The signal acquisition module is responsible for receiving the original noise signal from the sensor and performing preliminary digital processing;

[0118] The preprocessing module preprocesses the collected noise signal, which may include filtering, denoising, normalization, feature extraction and other steps to prepare data for subsequent deep learning model use;

[0119] The threshold setting module allows the user or the system to set a warning threshold as needed, and this threshold is used to determine whether the noise signal is abnormal;

[0120] The warning trigger module is used to trigger the warning signal when the CNN-LSTM model determines that the noise signal exceeds the warning threshold, and notify relevant personnel through sound, light, text message, and email.

[0121] This embodiment also provides a computer device, which is suitable for the large transformer noise monitoring, analysis and fault early warning method, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the large transformer noise monitoring, analysis and fault early warning method proposed in the above embodiment.

[0122] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0123] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for noise monitoring, analysis and fault early warning of a large transformer as proposed in the above embodiment is implemented.

[0124] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0125] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for noise monitoring, analysis and fault early warning of large transformers, characterized in that: The following steps are included: The wide-band noise signal is collected by sensors and pre-processed; A CNN-LSTM model is established for the collected noise signal through the deep learning module; Set the warning threshold, make judgments through the CNN-LSTM model, and trigger the warning signal.

2. The large transformer noise monitoring, analysis and fault early warning method according to claim 1, characterized in that: The steps for collecting noise signals include: Multi-dimensional acoustic sensors are arranged around the transformer to form a dense monitoring network, covering wide-band noise signal collection to ensure the comprehensiveness and high fidelity of noise data; Each group of sensors is equipped with an anti-interference hardware filter to reduce the impact of environmental noise; The acquisition frequency range is set to 20Hz to 200kHz, covering the audible range of the human ear and part of the ultrasonic region, and is used to collect ultrasonic noise generated when a transformer has an internal fault; The collected noise signals include the time series of sound pressure level, sound intensity and spectral components.

3. The large transformer noise monitoring, analysis and fault early warning method according to claim 2, characterized in that: The steps of preprocessing the noise signal include original signal denoising, signal framing and windowing, and feature analysis and feature value extraction of the noise signal; The steps of denoising the original signal include: Subtract the estimated ambient noise component from the original signal, retaining the noise signal unique to transformer operation; When the data processing center receives the main transformer noise, it first pre-processes the noise data and denoises the main transformer noise data using the wavelet threshold denoising method. The specific steps are as follows: First, perform wavelet transform on the main transformer noise signal x(t) to obtain a set of wavelet decomposition coefficients W j,k , by decomposing the wavelet coefficients W j,k Perform threshold processing to obtain denoised wavelet coefficients Finally, using the estimated wavelet coefficients Perform wavelet reconstruction to obtain the estimated signal This is the signal after denoising; The steps of signal framing and windowing processing include: A continuous main transformer noise signal is divided into multiple frames with a length of L. Then the nth frame can be expressed as x n (t), each frame is then processed individually; According to the time domain distribution of transformer noise characteristics and fault features, the main transformer noise signal is divided into 50ms frames, and each frame signal is shifted 25ms relative to the previous frame on the time axis; Apply a window function ω(t) to each frame signal, then the windowed signal y n (t) = x n (t)·ω(t-nL).

4. The large transformer noise monitoring, analysis and fault early warning method according to claim 3, characterized in that: The steps of feature analysis and feature value extraction of the noise signal include: For the windowed signal y n (t) Perform discrete Fourier transform (DFT) to convert the signal from time domain to frequency domain and obtain the spectrum: Then calculate the power spectral density (PSD) and take the square of the absolute value as the energy estimate: n (k)=|Y n (k)| 2 ; Apply a filter to the power spectrum to obtain the energy of each band In the formula, H m (k) represents the mth Mel filter; Taking the logarithm of the energy of each band, we get Perform discrete cosine transform on the logarithmic energy to obtain the Mel frequency cepstrum coefficients: Finally, the features are combined to obtain a feature vector The feature vector Fn is used as the vector input.

5. The large transformer noise monitoring, analysis and fault early warning method according to claim 4, characterized in that: The construction method of the CNN-LSTM model is: First, the eigenvector F n Each feature in is normalized to eliminate the scale differences between different features; Serialize the continuous feature vectors to form a time series data set. Each sample consists of T continuous feature vectors. The sample is represented by {F n-T+1 ,F n-T+2,…, F n } Divide the serialized data set into training set, validation set and test set; The CNN-LSTM model includes an input layer, a convolution layer, a flattening layer, an LSTM layer, a fully connected layer, and an output layer.

6. The large transformer noise monitoring, analysis and fault early warning method according to claim 5, characterized in that: The method for constructing the CNN-LSTM model also includes fault diagnosis, and the specific steps include: Use the fault data and normal noise data set in the historical data to train the CNN-LSTM model; After training, the CNN-LSTM model can transform the input feature vector F n Map it to the corresponding fault category or predict its health status, and use the Softmax function in the output layer to obtain the probability distribution of the fault type, where the category corresponding to the highest probability is the diagnosis result.

7. The large transformer noise monitoring, analysis and fault early warning method according to claim 6, characterized in that: Methods for sending early warning signals include: Set the warning threshold θ and make a judgment: When the CNN-LSTM model predicts the probability of failure P alert Exceeding this threshold, that is, P alert >θ, then the warning is triggered; Input new data into the CNN-LSTM model in real time and calculate P at each moment alert , when P is found alert When the preset threshold is exceeded, a warning signal is immediately issued and measures are taken in advance to prevent failures.

8. A large transformer noise monitoring, analysis and fault early warning system, based on the large transformer noise monitoring, analysis and fault early warning method according to any one of claims 1 to 7, characterized in that: It includes a signal acquisition module, a preprocessing module, a threshold setting module, and an early warning trigger module; The signal acquisition module is responsible for receiving the original noise signal from the sensor and performing preliminary digital processing; The preprocessing module preprocesses the collected noise signal, which may include filtering, denoising, normalization, feature extraction and other steps to prepare data for subsequent deep learning model use; The threshold setting module allows the user or the system to set the warning threshold as needed, and this threshold is used to determine whether the noise signal is abnormal; The warning trigger module is used to trigger the warning signal when the CNN-LSTM model determines that the noise signal exceeds the warning threshold, and notify relevant personnel through sound, light, text message, and email.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the large transformer noise monitoring, analysis and fault warning method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the large transformer noise monitoring, analysis and fault warning method according to any one of claims 1 to 7 are implemented.

Images