Power equipment defect identification method and system based on sounding mechanism and data driving

Abstract

The invention belongs to the technical field of electrical equipment defect identification, and provides an electrical equipment defect identification method and system based on a sounding mechanism and data driving, and the technical scheme is as follows: firstly, the identification and diagnosis of a mechanism model are realized through the steps of signal processing, feature extraction and the like according to a diagnosis rule; wherein the feature acquisition in the spectrogram adopts a low-frequency band selection strategy to improve the accuracy of data features and reduce the redundancy of the data, then real-time collected data is transmitted to an optimized deep learning model, and the two results adopt a fusion driving thought to realize the accurate recognition of abnormal voiceprints. The robustness of the model for equipment difference and environmental noise is improved, and the complex environment and fault condition of the transformer substation are met.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric power equipment defect identification, and in particular relates to an electric power equipment defect identification method and system based on a sound generation mechanism and data drive. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] At present, computer technology represented by artificial intelligence is accelerating the construction of new power systems. Intelligent operation and inspection methods conduct real-time monitoring, status assessment and fault diagnosis of power grid equipment, realize intelligent management and optimization of power grid operation, reduce the operating costs of power companies and reduce fault handling time. Although voiceprint defect recognition technology is also included in the overall architecture of intelligent inspection, there is currently insufficient attention and development on voiceprint defect recognition technology for internal operation of transformers, and the corresponding defect recognition and analysis technology has been slowly applied. Therefore, it is urgent to study and apply key intelligent inspection technologies such as voiceprint defect recognition of power equipment such as transformers.

[0004] The existing power equipment defect diagnosis technology based on voiceprint features can effectively identify the operating status of power equipment to a certain extent, but it still has problems in model construction and field application, such as few identifiable defect categories, limited sample data, and low model accuracy. Summary of the invention

[0005] In order to solve at least one technical problem existing in the above-mentioned background technology, the present invention provides a method and system for identifying defects in power equipment based on sound mechanism and data-driven, which performs time-frequency feature analysis on multi-channel voiceprint signals, establishes a feature diagnosis model, and runs a fusion drive method to identify defects in the mechanism model and the deep learning model, thereby realizing accurate diagnosis of fault signals of power equipment and improving the engineering implementation effect of artificial intelligence technology.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] The first aspect of the present invention provides a method for identifying defects in electric power equipment based on a sound mechanism and data drive, comprising the following steps:

[0008] Obtain voiceprint signal data of power equipment;

[0009] Analyze the voiceprint signal data of the power equipment from the sound generation mechanism, extract the numerical features of multiple power equipment voiceprint signals, establish fault diagnosis rules according to the multiple numerical features of the power equipment voiceprint signals and the characteristic performance under different faults, and obtain the first power equipment defect discrimination result according to the established diagnosis rules;

[0010] Based on the electric power equipment voiceprint signal data and the trained electric power equipment defect recognition model, a second electric power equipment defect recognition result is obtained;

[0011] Combining the first power equipment defect discrimination result and the second power equipment defect identification result, a fusion driving strategy is adopted to obtain the final power equipment defect identification result.

[0012] Furthermore, the fault diagnosis rules are specifically as follows:

[0013] When the change rate of the transformer body soundprint signal-to-noise ratio is greater than the set change rate threshold, it is determined that there is an interference signal. If the interference signal persists, it is considered that there is a probability that the sensor is damaged.

[0014] When the sound pressure level is greater than the sound pressure level threshold, it is determined that there is an overload or heavy load fault;

[0015] When the odd-even harmonic ratio is greater than the set odd-even harmonic ratio threshold, it is determined that a DC bias magnetic defect exists;

[0016] When the proportion of high-frequency energy in the transformer is greater than the set threshold, it is judged as a resonance or loose fault.

[0017] When the transformer spectral entropy is greater than the set transformer spectral entropy threshold, it is determined to be a cooler abnormal noise fault;

[0018] When the difference between one or more features of the transformer voiceprint data and the history is greater than the set value, or the difference between the voiceprint features of a transformer and other transformers of the same model and working conditions is greater than the set threshold, it is other faults.

[0019] Furthermore, the specific fusion drive strategies include:

[0020] If the first power equipment defect determination result is that the voiceprint signal is an interference signal, directly output the result;

[0021] If the first power equipment defect identification result is other types, the confidence probability of the second power equipment defect identification result and the confidence probability of the first power equipment defect identification result are combined with the reliability weights of the second power equipment defect identification result and the first power equipment defect identification result to obtain a comprehensive result of the fault category confidence probability, and the output fault category with the highest probability is used as the final power equipment defect identification result.

[0022] Furthermore, the reliability weight of the first electric equipment defect identification result is greater than the reliability weight of the second electric equipment defect identification result.

[0023] Further, the second electric equipment defect recognition result is obtained based on the electric equipment voiceprint signal data and the trained electric equipment defect recognition model, including:

[0024] The normal samples obtained are fused with the sounds of DC bias magnetism, heavy overload, partial discharge, and abnormal noise of the cooler, and common interference noise is mixed in to construct a training data set;

[0025] The training data set is used to train the power equipment defect recognition model to obtain the trained power equipment defect recognition model; wherein the construction process of the power equipment defect recognition model includes:

[0026] The original audio signal is processed and input into the backbone network for convolution, normalization, pooling, and full connection to obtain the multi-scale features of the spectrogram;

[0027] The multi-scale features of the spectrogram output by the backbone network are fused with the Transformer encoder to form a new encoder, which encodes the input sequence into an internal semantic vector to obtain global and local feature information;

[0028] The global and local feature information is input into the feedforward network to obtain the width and height of the input image and the label of the category to which it belongs;

[0029] The trained electric power equipment defect recognition model is used to perform fault recognition on the collected audio signal to obtain a second electric power equipment defect recognition result.

[0030] Furthermore, when the multi-scale features of the spectrogram output by the backbone network are fused with the Transformer encoder to form a new encoder, the first feature layer and the last feature layer of the backbone network are selected and fused with Transformer as F0 and F_L respectively. According to the target characteristics in different scenarios, a weighted fusion strategy is adopted. When detecting small targets in large scenes, the F0 weight fusion ratio is increased; when detecting large targets, the F_L weight fusion ratio is increased.

[0031] Furthermore, the numerical features include transformer body soundprint signal-to-noise ratio, sound pressure level, odd-even harmonic ratio, high-frequency energy proportion and transformer spectral entropy.

[0032] The second aspect of the present invention provides a power equipment defect identification system based on a sound mechanism and data drive, comprising:

[0033] A data acquisition module, which is used to acquire voiceprint signal data of power equipment;

[0034] A first electric equipment defect identification module, which is used to analyze the electric equipment voiceprint signal data from the sound generation mechanism, extract the numerical features of multiple electric equipment voiceprint signals, establish a fault diagnosis rule according to the multiple numerical features of the electric equipment voiceprint signals and the characteristic performance under different faults, and obtain the first electric equipment defect identification result according to the established diagnosis rule;

[0035] A second electric equipment defect recognition module, which is used to obtain a second electric equipment defect recognition result based on the electric equipment voiceprint signal data and the trained electric equipment defect recognition model;

[0036] The data fusion driving module combines the first power equipment defect discrimination result and the second power equipment defect identification result, adopts a fusion driving strategy, and obtains the final power equipment defect identification result.

[0037] A third aspect of the present invention provides a computer-readable storage medium.

[0038] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in the above-mentioned method for identifying defects in electric power equipment based on a sound mechanism and data drive.

[0039] A fourth aspect of the present invention provides a computer device.

[0040] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the above-mentioned method for identifying defects in electric power equipment based on sound generation mechanism and data drive are implemented.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. The present invention proposes a defect identification method driven by the sound mechanism and data fusion of power transformers. Fault diagnosis rules are established according to multiple numerical features of the voiceprint signal of power equipment and the characteristic performance under different faults, and the recognition and diagnosis of the mechanism model is realized. The real-time collected data is transmitted to the tuned deep learning model, and the fusion-driven idea is adopted for the two results to realize the accurate identification of abnormal voiceprints, thereby avoiding misjudgment of the diagnosis results, improving the robustness of the model in the face of equipment differences and environmental noise, and meeting the complex environment and fault conditions of substations.

[0043] 2. The present invention proposes a method for accurate identification of power transformers, designs a multi-scale network with two parallel branches of CNN and Transformer network, and uses it as the encoder part of the CNN-Transformer network model. The local detail features and global semantic features extracted layer by layer are compensated for their respective feature expressions through information fusion, expanding the feature field of the same level and improving the ability to judge and locate the target. In addition, the residual module is used to connect multiple convolution calculations in CNN to maintain the model performance and network depth, and improve the accuracy and practicality of defect identification.

[0044] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0046] Figure 1 It is a flow chart of a method for identifying defects in electric power equipment based on sound generation mechanism and data drive provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0048] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0049] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0050] The existing transformer defect diagnosis technology based on voiceprint features mentioned in the background technology can effectively identify the operating status of the transformer to a certain extent, but it still has problems in model construction and field application, such as few identifiable defect categories, limited sample data, and low model accuracy.

[0051] For example, most existing studies focus on a certain type of transformer defect, and mainly focus on the diagnosis and identification of mechanical defects or insulation defects, such as winding looseness and deformation, DC bias magnetism, or insulation discharge. The identification results of the constructed diagnostic model do not cover all potential operating conditions of the transformer, which to some extent limits the field application of voiceprint diagnostic technology.

[0052] For example, most transformer voiceprint sample data come from test results. The number of samples is insufficient and the interference and influence of substation site environmental noise and other power transmission and transformation equipment are not taken into account. The limitations of voiceprint samples will increase the misjudgment rate of diagnostic results to a certain extent.

[0053] The present invention refines labels based on sample data, introduces mechanism and data fusion driven algorithms, conducts time-frequency feature analysis of the data, sets weights respectively, and constitutes the final output result; considering that the early deep learning model learning cases are limited, the mechanism model can be appropriately given more weight, and as the data set is enriched and expanded, the weight of the mechanism model can be gradually reduced or cancelled, and the deep learning model can be improved, ultimately improving the model recognition accuracy.

[0054] Embodiment 1

[0055] This embodiment provides a method for identifying defects in electric power equipment based on a sound mechanism and data drive, comprising the following steps:

[0056] Step 1: Obtain the voiceprint signal data of the power equipment;

[0057] Step 2: Analyze the sound generation mechanism based on the voiceprint signal data of the power equipment and obtain the first power equipment defect identification result in combination with the established diagnostic rules; specifically, the following steps are included:

[0058] Step 201: pre-processing the acquired voiceprint signal data of the electric power equipment, transplanting the noise of the voiceprint signal and performing data frequency domain conversion;

[0059] For interference signals with a fixed frequency band and little overlap with the transformer main frequency and harmonics, IIR can be used to form a narrowband filter group for noise reduction;

[0060] For interference where there is frequency band overlap between noise and transformer operation soundprint, and the frequency distribution and amplitude characteristics of the interference signal are relatively stable, the spectral subtraction denoising method can be used to suppress noise.

[0061] The soundprint characteristics of the transformer are mainly distributed within 50-2000Hz, and are mainly composed of 100, 200, 300... and other components, and the main frequency is mainly in the low frequency band.

[0062] After that, the time domain signal is converted into a frequency domain signal through FFT to facilitate frequency domain feature extraction. Since the transformer voiceprint base frequency is 100Hz and the overall signal is a stable signal, the data length of the FFT calculation is selected between 1-10s to ensure a reasonable frequency domain resolution and avoid redundant calculations, and the spectrum of the power equipment voiceprint signal is obtained:

[0063]

[0064] Among them, a n is the amplitude of the signal at the nth sampling point (or at time n), X(k) is the complex spectrum at frequency k, and N is the total number of sampling points in the time domain (the number of sampling points greater than or equal to 1s).

[0065] Step 202: extracting features based on the obtained frequency spectrum of the voiceprint signal of the electric power equipment to obtain multiple numerical features of the voiceprint signal of the electric power equipment;

[0066] In this embodiment, only 50, 100, 150, and 200 Hz in the spectrum are extracted, which not only ensures effective features, but also achieves dimensionality reduction and reduces data redundancy. Specifically, it includes the extraction of transformer body voiceprint signal-to-noise ratio, sound pressure level, odd-even harmonic ratio, high-frequency energy ratio, transformer spectral entropy and other features.

[0067] The specific feature extraction calculation formula is:

[0068] Transformer body soundprint signal-to-noise ratio:

[0069]

[0070] Sound pressure level:

[0071] SPL = 20log 10 RMS / base,

[0072] Even-odd harmonic ratio:

[0073]

[0074] High frequency energy ratio:

[0075]

[0076] Transformer spectral entropy:

[0077]

[0078] Among them, RMS is the effective value of the voiceprint signal (frequency band above 20Hz), and base is the reference standard of the sound pressure level, specifically 20uPa. i*50 is the spectrum of the voiceprint signal, and the corresponding index is the i-th 50Hz frequency. is the energy proportion at the corresponding frequency.

[0079] Step 203: establishing fault diagnosis rules according to multiple numerical features of the voiceprint signal of the power equipment and characteristic manifestations under different faults;

[0080] When establishing fault diagnosis rules, multiple numerical features and characteristic performances under different faults are used, as shown in Table 1.

[0081] Table 1 Main features, meanings and uses

[0082]

[0083] The specific fault diagnosis rules are as follows:

[0084] When the change rate of the transformer body soundprint signal-to-noise ratio is greater than the set change rate threshold, there is a sensor damage and the environment is too noisy for a short time (thunder, car horns, firecracker noise). At this time, the signal-to-noise ratio is low and the reliability of the diagnosis result is low. It can be determined that there is an interference signal. If the interference signal persists, it can be considered that there is a probability that the sensor is damaged;

[0085] When the sound pressure level is greater than the sound pressure level threshold, due to the positive correlation between the transformer operating noise and the load, specifically, the transformer operating noise and the square of the load current change linearly. When the sound pressure level is greater than the sound pressure level threshold, it is determined that there is an overload or heavy load fault. If it works in an overload state for a long time, it is very easy to cause irreversible damage to the transformer.

[0086] When the odd-even harmonic ratio is greater than the set odd-even harmonic ratio threshold, it can be directly determined that a DC bias magnetic defect exists, which may be accompanied by an increase in noise. In this case, a DC bias magnetic fault can be directly output.

[0087] When the proportion of energy in the high frequency band of 800Hz-2000Hz of the transformer is greater than the set threshold, it is judged as a resonance or loose fault. At this time, the fault may be outside the transformer.

[0088] When the transformer spectral entropy is greater than the set transformer spectral entropy threshold, it is determined to be a cooler abnormal noise fault.

[0089] When the difference between one or more features of the transformer voiceprint data and the historical data is greater than the set value, that is, it is significantly increased or decreased, or the difference in the voiceprint features of a transformer compared with other transformers of the same model and working conditions is greater than the set threshold, it often indicates that there are some abnormalities, and the existing faults may not be fully covered, but it should still be treated as an abnormal alarm.

[0090] It should be noted that the thresholds mentioned above can be set according to actual working conditions and are not specifically limited in this embodiment.

[0091] Step 204: Determine the fault type of the power equipment based on the established fault diagnosis rules.

[0092] Step 3: Based on the power equipment voiceprint signal data and the trained power equipment defect recognition model, a second power equipment defect recognition result is obtained;

[0093] In this embodiment, taking the power equipment as a transformer as an example, the following steps are specifically included:

[0094] Step 301: The acquired normal samples are merged with the sounds of DC bias magnetism, heavy overload, partial discharge, abnormal noise of cooler, etc., and common interference noise (speech, environmental noise, etc.) is mixed in to construct a training data set;

[0095] Step 302: Using the training data set to train the power equipment defect recognition model to obtain a trained power equipment defect recognition model;

[0096] In this embodiment, the power equipment defect recognition model includes a backbone network, an encoder and a feedforward network;

[0097] Backbone network: The original audio signal is processed by signal processing such as framing, windowing, translation, spectrum calculation, etc. to obtain a time-frequency spectrogram. The two-dimensional image information is input into the backbone network for convolution, normalization, pooling, full connection and other steps to extract features and calculate the multi-scale features of the spectrogram;

[0098] In order to maintain the model performance while increasing the network depth and avoid the problem of gradient disappearance or explosion, multiple convolution calculations are connected by residual modules.

[0099] In this embodiment, the backbone network adopts CNN;

[0100] Encoder: The multi-scale features of the spectrogram output by the backbone network are fused with the Transformer encoder to form a new encoder. Since the Transformer model was originally designed for sequence-to-sequence learning tasks, the encoder is responsible for encoding the input sequence into an internal semantic vector, while the decoder is responsible for decoding this internal semantic vector into an output sequence.

[0101] However, when the Transformer model is used for classification tasks, its structure is adjusted and the decoder part is no longer needed.

[0102] In the classification task, the Transformer model mainly uses the output of the encoder to make classification decisions through one or more fully connected layers. This adjustment allows the model to focus on extracting useful features from the input data for classification. The first layer of the CNN network output has more detailed local basic feature information, and the last layer has a larger receptive field and global feature information. The first feature layer and the last feature layer are selected and fused with the Transformfer network, respectively, and are denoted as F0 and F_L. According to the target characteristics in different scenes, a weighted strategy is adopted for fusion. When detecting small targets in large scenes, in order to enhance the ability to express the details of small targets, the proportion of F0 weight fusion is increased; when detecting large targets, in order to enhance the ability to express the global features of large targets, the proportion of F_L weight fusion is increased.

[0103] Feedforward network: The feedforward network consists of a 3-layer perceptron with a ReLU activation function and a linear mapping layer. The output prediction result is the normalized target center point coordinates, the width and height of the detection box relative to the input image, and the label of the category to which it belongs. Finally, data classification is achieved through the linear layer.

[0104] Step 303: Use the trained power equipment defect recognition model to perform fault recognition on the collected audio signal to obtain a fault recognition result;

[0105] Based on the above-mentioned CNN and Transformer network dual-branch parallel multi-scale network, as the encoder part of the CNN-Transformer network model, the local detail features and global semantic features extracted layer by layer are mutually integrated to compensate for their respective feature expressions, expand the field of view of features at the same level, and improve the ability to judge and locate targets. In addition, the residual module is used to connect multiple convolution calculations in CNN to maintain model performance and network depth, thereby improving the accuracy and practicality of defect recognition.

[0106] Step 4: Combining the first power equipment defect identification result and the second power equipment defect identification result, adopting a fusion drive strategy to obtain a final power equipment defect identification result; specifically comprising the following steps:

[0107] Step 401: if the first power equipment defect determination result is that the voiceprint signal is an interference signal, directly output the result;

[0108] The mechanism model is used to realize the discrimination of abnormal data of non-transformer faults. When the voiceprint data noise is large and the signal-to-noise ratio of the transformer voiceprint is low, it may be due to sensor failure, abnormal sensor installation, low data acquisition quality, strong interference, etc. At this time, it is not easy to enter the fault discrimination logic, and the voiceprint signal can be directly output as an interference signal;

[0109] Step 402: If the first power equipment defect identification result is the other result, it means that the voiceprint is likely to be the real transformer voiceprint data. The collected voiceprint samples are classified and predicted by the mechanism model and the deep learning model respectively. The identification result and confidence probability of the mechanism model and the classification result of the deep learning model with the maximum probability corresponding to the category and its probability are output, corresponding to the label 机理 , P 机理 、label 深度学习 , P 深度学习 ;

[0110] Step 403: Set the reliability weights w of the mechanism model and the deep learning model results respectively 机理 、w 深度学习 , the overall reliability or category reliability segmentation can be adopted. Taking the overall reliability as an example, it can be determined based on the performance of the mechanism model and the deep learning model on the test samples and combined with certain expert experience evaluation;

[0111] Step 404: Compare the model result reliability weight and the comprehensive result P = max(P 机理 *w机理 , P 深度学习 *w 深度学习 ), determine the output fault category with the highest probability.

[0112] Step 405: Update the reliability weights of the mechanism model and the deep learning model results;

[0113] When there is new progress in the research of a certain type of fault mechanism, the reliability weight of the corresponding category of faults in the mechanism model can be increased. When the number of accumulated samples reaches a certain requirement and the accuracy of the deep learning model is high, such as reaching more than 95%, the reliability weight of the deep learning model can be appropriately increased.

[0114] Given that the transformer fault samples contained in the early data sets are not rich enough and are mostly data collected from different angles of the same transformer, the trained machine learning model is difficult to be widely applied to other transformer defect identification. 深度学习 =0.3, while the mechanism model is easy to set appropriate parameters according to the horizontal and vertical comparison results because of its adjustable parameters. It has relatively high accuracy and is recommended to be based on higher weights, such as w 机理 =0.7.

[0115] By adopting a fusion-driven approach for the two results, accurate identification of abnormal voiceprints can be achieved, improving the robustness of the model in the face of equipment differences and environmental noise, and meeting the complex environment and fault conditions of substations.

[0116] Embodiment 2

[0117] This embodiment provides a power equipment defect identification system based on sound generation mechanism and data drive, including:

[0118] A data acquisition module, which is used to acquire voiceprint signal data of power equipment;

[0119] A first electric equipment defect identification module, which is used to analyze the electric equipment voiceprint signal data from the sound generation mechanism, extract the numerical features of multiple electric equipment voiceprint signals, establish a fault diagnosis rule according to the multiple numerical features of the electric equipment voiceprint signals and the characteristic performance under different faults, and obtain the first electric equipment defect identification result according to the established diagnosis rule;

[0120] A second electric equipment defect recognition module, which is used to obtain a second electric equipment defect recognition result based on the electric equipment voiceprint signal data and the trained electric equipment defect recognition model;

[0121] The data fusion driving module combines the first power equipment defect discrimination result and the second power equipment defect identification result, adopts a fusion driving strategy, and obtains the final power equipment defect identification result.

[0122] Embodiment 3

[0123] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the method for identifying defects in electric power equipment based on sound generation mechanism and data drive as described above are implemented.

[0124] Embodiment 4

[0125] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the method for identifying defects in electric power equipment based on sound generation mechanism and data drive as described above are implemented.

[0126] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0127] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0128] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0129] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0130] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0131] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for identifying defects in power equipment based on sound generation mechanism and data-driven, characterized in that: The steps include: Obtain voiceprint signal data of power equipment; Analyze the voiceprint signal data of the power equipment from the sound generation mechanism, extract the numerical features of multiple power equipment voiceprint signals, establish fault diagnosis rules according to the multiple numerical features of the power equipment voiceprint signals and the characteristic performance under different faults, and obtain the first power equipment defect discrimination result according to the established diagnosis rules; Based on the electric power equipment voiceprint signal data and the trained electric power equipment defect recognition model, a second electric power equipment defect recognition result is obtained; Combining the first power equipment defect discrimination result and the second power equipment defect identification result, a fusion driving strategy is adopted to obtain the final power equipment defect identification result.

2. The method for identifying defects in electric power equipment based on sound generation mechanism and data drive according to claim 1, characterized in that: The fault diagnosis rules are specifically as follows: When the change rate of the transformer body soundprint signal-to-noise ratio is greater than the set change rate threshold, it is determined that there is an interference signal. If the interference signal persists, it is considered that there is a probability that the sensor is damaged. When the sound pressure level is greater than the sound pressure level threshold, it is determined that there is an overload or heavy load fault; When the odd-even harmonic ratio is greater than the set odd-even harmonic ratio threshold, it is determined that a DC bias magnetic defect exists; When the proportion of high-frequency energy in the transformer is greater than the set threshold, it is judged as a resonance or loose fault. When the transformer spectral entropy is greater than the set transformer spectral entropy threshold, it is determined to be a cooler abnormal noise fault; When the difference between one or more features of the transformer voiceprint data and the history is greater than the set value, or the difference between the voiceprint features of a transformer and other transformers of the same model and working conditions is greater than the set threshold, it is other faults.

3. The method for identifying defects in electric power equipment based on sound generation mechanism and data drive according to claim 1, characterized in that: Specific fusion-driven strategies include: If the first power equipment defect determination result is that the voiceprint signal is an interference signal, directly output the result; If the first power equipment defect identification result is other types, the confidence probability of the second power equipment defect identification result and the confidence probability of the first power equipment defect identification result are combined with the reliability weights of the second power equipment defect identification result and the first power equipment defect identification result to obtain a comprehensive result of the fault category confidence probability, and the output fault category with the highest probability is used as the final power equipment defect identification result.

4. The method for identifying defects in electric power equipment based on sound generation mechanism and data drive according to claim 3, characterized in that: The reliability weight of the first electric equipment defect identification result is greater than the reliability weight of the second electric equipment defect identification result.

5. The method for identifying defects in electric power equipment based on sound generation mechanism and data drive according to claim 1, characterized in that: The obtaining of a second electric equipment defect recognition result based on the electric equipment voiceprint signal data and the trained electric equipment defect recognition model includes: The normal samples obtained are fused with the sounds of DC bias magnetism, heavy overload, partial discharge, and abnormal noise of the cooler, and common interference noise is mixed in to construct a training data set; The training data set is used to train the power equipment defect recognition model to obtain the trained power equipment defect recognition model; wherein the construction process of the power equipment defect recognition model includes: The original audio signal is processed and input into the backbone network for convolution, normalization, pooling, and full connection to obtain the multi-scale features of the spectrogram; The multi-scale features of the spectrogram output by the backbone network are fused with the Transformer encoder to form a new encoder, which encodes the input sequence into an internal semantic vector to obtain global and local feature information; The global and local feature information is input into the feedforward network to obtain the width and height of the input image and the label of the category to which it belongs; The trained electric power equipment defect recognition model is used to perform fault recognition on the collected audio signal to obtain a second electric power equipment defect recognition result.

6. The method for identifying defects in electric power equipment based on sound generation mechanism and data drive according to claim 5, characterized in that: When the multi-scale features of the spectrogram output by the backbone network are fused with the Transformer encoder to form a new encoder, the first feature layer and the last feature layer of the backbone network are selected and fused with Transformer as F0 and F_L respectively. According to the target characteristics in different scenarios, a weighted fusion strategy is adopted. When detecting small targets in large scenes, the F0 weight fusion ratio is increased; when detecting large targets, the F_L weight fusion ratio is increased.

7. The method for identifying defects in electric power equipment based on sound generation mechanism and data drive according to claim 1, characterized in that: The numerical features include the transformer body soundprint signal-to-noise ratio, sound pressure level, odd-even harmonic ratio, high-frequency energy proportion and transformer spectral entropy.

8. The power equipment defect identification system based on sound mechanism and data drive is characterized by: include: A data acquisition module, which is used to acquire voiceprint signal data of power equipment; A first electric equipment defect identification module, which is used to analyze the electric equipment voiceprint signal data from the sound generation mechanism, extract the numerical features of multiple electric equipment voiceprint signals, establish a fault diagnosis rule according to the multiple numerical features of the electric equipment voiceprint signals and the characteristic performance under different faults, and obtain the first electric equipment defect identification result according to the established diagnosis rule; A second electric equipment defect recognition module, which is used to obtain a second electric equipment defect recognition result based on the electric equipment voiceprint signal data and the trained electric equipment defect recognition model; The data fusion driving module combines the first power equipment defect discrimination result and the second power equipment defect identification result, adopts a fusion driving strategy, and obtains the final power equipment defect identification result.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the method for identifying defects in electric power equipment based on sound generation mechanism and data drive as described in any one of claims 1 to 7 are implemented.

10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the method for identifying defects in electric power equipment based on sound mechanism and data drive as described in any one of claims 1-7 are implemented.

Images