A partial discharge identification method and device based on ultrasonic wave

Abstract

The present invention provides a method and device for identifying partial discharge based on ultrasound, which relates to the field of ultrasonic discharge identification. The method comprises: calibrating an ultrasonic signal of a first preset duration, determining the state information corresponding to each identification feature at each moment; constructing an initial state transfer matrix, calculating the transition probability of each state information transferring to any state information, filling all state transfer probabilities into the initial state transfer matrix, obtaining a state transfer probability matrix, traversing all identification features, obtaining a state transfer probability matrix; inputting all state transfer probability matrices into a preset partial discharge identification model, and obtaining partial discharge prediction results. The present invention realizes accurate identification and classification of partial discharge phenomena in GIS equipment through feature analysis and the construction of a state transfer probability matrix, combined with a preset partial discharge identification model, and quickly screens out data that does not require in-depth analysis, thereby reducing unnecessary processing of such data and improving overall identification efficiency.

Description

Technical Field

[0001] The present invention relates to the field of ultrasonic discharge recognition, and in particular to an ultrasonic-based local discharge recognition method and device. Background Art

[0002] Existing technologies often lack the flexibility to adapt to discharge detection requirements in different application scenarios. In application scenarios with high real-time requirements, it is often difficult to meet real-time requirements while maintaining high recognition accuracy. The Chinese invention patent with application publication number CN116503710 A discloses a GIS partial discharge type recognition method based on an adaptive convolutional neural network. The method is to input the PRPD image and the GAF image set into the CNN model based on the PSO algorithm, find the optimal hyperparameters of the CNN model, obtain the adaptive CNN models corresponding to the PRPD image and the GAF image, respectively, and perform feature extraction and information fusion on the PRPD image and the GAF image through the adaptive CNN model to realize GIS partial discharge type recognition; the authorization announcement number is CN 103267932 B's Chinese invention patent discloses a detection method, which has the following specific steps: Step 1: The GIS under test is tested and the test results are uploaded to the computer; Step 2: The computer processes and identifies the results of the ultrasonic detection method; According to the test results, the basic probability assignment function value of the evidence to the target, i.e., the assignment probability, is given; Step 3: The evidence space of each GIS fault is calculated; Step 4: The diagnostic result, i.e., the diagnostic type, is determined according to the diagnostic decision. These application scenarios usually focus on the specific type of discharge rather than the severity of the discharge.

[0003] However, in some application scenarios, it may be necessary to pay attention not only to the specific type of discharge, but also to the severity of the discharge. Existing solutions often directly identify the discharge type based on the collected ultrasonic information. Regardless of the severity of the discharge, complex calculations and analysis are required, resulting in a large consumption of computing resources, especially when processing large amounts of data. For samples with no discharge or slight discharge, existing solutions still require a complete discharge type identification process, which not only increases processing time, but also reduces overall identification efficiency.

[0004] However, there is currently no technical solution that can solve the above technical problems, and there is no partial discharge identification method and device based on ultrasound. Summary of the invention

[0005] The present invention provides a partial discharge recognition method and device based on ultrasonic wave. By presetting a partial discharge recognition model and first judging the severity of the discharge, data that does not require in-depth analysis can be quickly screened out, thereby reducing unnecessary processing of such data and improving overall recognition efficiency.

[0006] In a first aspect, the present invention provides a partial discharge identification method based on ultrasound, comprising:

[0007] Calibrate an ultrasonic signal of a first preset duration, determine state information corresponding to each identification feature at each moment in the first preset duration, the identification feature comprising at least one of an amplitude feature, a frequency feature, and a phase feature, and the ultrasonic signal of the first preset duration is obtained from any ultrasonic sensor on the GIS device in a current preset time window;

[0008] For each identification feature, an initial state transfer matrix is ​​constructed, and the transition probability of each state information transferring to any state information is calculated in each state information corresponding to different times, and all state transfer probabilities are obtained. All state transfer probabilities are filled into the initial state transfer matrix to obtain a state transfer probability matrix, and all identification features are traversed to obtain a state transfer probability matrix corresponding to each identification feature;

[0009] Inputting all state transition probability matrices into a preset partial discharge identification model to obtain a partial discharge prediction result output by the preset partial discharge identification model;

[0010] The preset partial discharge recognition model is determined after training based on all state transition probability matrices corresponding to all ultrasonic signal samples and the partial discharge prediction results corresponding to each ultrasonic signal sample, and the partial discharge prediction results include no discharge, slight discharge and severe discharge.

[0011] According to the ultrasonic-based partial discharge identification method provided by the present invention, when the identification feature is an amplitude feature, the ultrasonic signal of the first preset time length is calibrated to determine the state information corresponding to each identification feature at each moment in the first preset time length, including:

[0012] For each moment, determining the amplitude value at said moment;

[0013] When the amplitude value is less than the first preset amplitude value, determining that the amplitude characteristic at the moment is in a low amplitude state;

[0014] When the amplitude value is greater than or equal to the first preset amplitude value and less than or equal to the second preset amplitude value, determining that the amplitude characteristic at the moment is in a medium amplitude state;

[0015] When the amplitude value is greater than the second preset amplitude value, it is determined that the amplitude characteristic at the moment is in a high amplitude state.

[0016] According to the ultrasonic-based partial discharge identification method provided by the present invention, when the identification feature is a frequency feature, the ultrasonic signal of the first preset time length is calibrated to determine the state information corresponding to each identification feature at each moment in the first preset time length, including:

[0017] Determine the frequency value corresponding to each moment;

[0018] When the frequency difference at any adjacent time is less than or equal to the preset frequency, determining that the frequency characteristic at the time is a frequency stable state;

[0019] When the frequency difference at any adjacent time moments is greater than the preset frequency, it is determined that the frequency characteristic at the time moment is in a frequency fluctuation state.

[0020] According to the ultrasonic-based partial discharge identification method provided by the present invention, when the identification feature is a phase feature, the ultrasonic signal of the first preset time length is calibrated to determine the state information corresponding to each identification feature at each moment in the first preset time length, including:

[0021] When the phase at any moment remains unchanged from that at the previous moment, determining that the phase characteristic at the moment is in a phase-unchanged state;

[0022] When the phase at any moment changes from that at the previous moment, the phase characteristic at the moment is determined to be a phase change state.

[0023] According to the ultrasonic-based partial discharge identification method provided by the present invention, the construction of the initial state transfer matrix includes:

[0024] When the identification feature is an amplitude feature, constructing a 3×3 initial state transfer matrix according to the amount of state information of the amplitude feature;

[0025] In the case where the identification feature is a frequency feature or a phase feature, a 2×2 initial state transfer matrix is ​​constructed according to the amount of state information of the frequency feature or the phase feature.

[0026] According to the ultrasonic-based partial discharge identification method provided by the present invention, the transfer probability of each state information to any state information is calculated in each state information corresponding to different time points to obtain all state transfer probabilities, including:

[0027] For each state information, obtaining the number of transitions from the state information to any state information, and determining the state transition probability according to the number of transitions and the total number of transitions of the state information;

[0028] Traverse all state information and obtain all state transition probabilities.

[0029] According to the ultrasonic-based partial discharge identification method provided by the present invention, the state transition probability is determined according to the number of transitions and the total number of times the state information is transferred, including:

[0030]

[0031] Wherein, P(j|i) is the state transition probability from state information i to state information j, M is the number of transitions from state information i to state information j, and N is the total number of transitions from state information i to any state.

[0032] According to the ultrasonic-based partial discharge identification method provided by the present invention, after obtaining the partial discharge prediction result output by the preset partial discharge identification model, the method further includes:

[0033] When it is determined that the partial discharge prediction result is a slight discharge, obtaining a next ultrasonic signal corresponding to a next time window;

[0034] When it is determined that the local discharge prediction result corresponding to the next ultrasonic signal is no discharge, updating the local discharge prediction result to no discharge;

[0035] When it is determined that the local discharge prediction result corresponding to the next ultrasonic signal is slight discharge or severe discharge, the local discharge prediction result is updated to severe discharge.

[0036] According to the ultrasonic-based partial discharge identification method provided by the present invention, when it is determined that the partial discharge prediction result is a severe discharge, the method further includes:

[0037] After acquiring the ultrasonic signal of the first preset time length, acquiring the target ultrasonic signal of the third preset time length after an interval of the second preset time length;

[0038] Processing the target ultrasonic signal according to Gram angular field conversion to obtain a GAF image;

[0039] Inputting the GAF image into a partial discharge classification model to obtain a partial discharge classification result output by the partial discharge classification model;

[0040] The partial discharge classification model is obtained after training based on all GAF sample images and the partial discharge classification results corresponding to each GAF sample image, wherein the partial discharge classification results include corona discharge, particle discharge, suspension discharge and cavitation discharge;

[0041] The first preset duration is the same as or different from the second preset duration, and the second preset duration is the same as or different from the third preset duration.

[0042] In a second aspect, a partial discharge identification device based on ultrasound is provided, comprising:

[0043] A determination unit, the determination unit is used to calibrate an ultrasonic signal of a first preset duration, determine state information corresponding to each identification feature at each moment in the first preset duration, the identification feature comprising at least one of an amplitude feature, a frequency feature, and a phase feature, and the ultrasonic signal of the first preset duration is obtained from any ultrasonic sensor on the GIS device in a current preset time window;

[0044] A construction unit, wherein the construction unit is used to construct an initial state transfer matrix for each identification feature, calculate the transition probability of each state information transferring to any state information in each state information corresponding to different time periods, obtain all state transfer probabilities, fill all state transfer probabilities into the initial state transfer matrix, obtain a state transfer probability matrix, traverse all identification features, and obtain a state transfer probability matrix corresponding to each identification feature;

[0045] An input unit, the input unit is used to input all state transition probability matrices into a preset partial discharge identification model to obtain a partial discharge prediction result output by the preset partial discharge identification model;

[0046] The preset partial discharge recognition model is determined after training based on all state transition probability matrices corresponding to all ultrasonic signal samples and the partial discharge prediction results corresponding to each ultrasonic signal sample, and the partial discharge prediction results include no discharge, slight discharge and severe discharge.

[0047] The present invention realizes accurate identification and classification of partial discharge phenomena in GIS equipment through detailed feature analysis and construction of state transition probability matrix, combined with a preset partial discharge identification model. Compared with the prior art, the present invention can quickly screen out data that does not require in-depth analysis by first judging the severity of the discharge, thereby reducing unnecessary processing of such data and improving overall identification efficiency; after confirming the existence of severe discharge, detailed discharge type identification can ensure that limited computing resources and time are invested in the most needed areas, optimizing resource allocation; for severe discharge, since its signal characteristics are usually more obvious and complex, the present invention can obtain higher accuracy when identifying the discharge type, and the early determination of no discharge or slight discharge can avoid unnecessary complex analysis of them, thereby reducing misjudgment; the judgment criteria and subsequent processing procedures of severe discharge can be adjusted according to actual needs to better adapt to different application scenarios;

[0048] The present invention reduces false alarms or missed alarms caused by a single misjudgment through continuous monitoring and re-prediction, improves the accuracy of partial discharge identification, and can more sensitively capture changes in discharge trends in the transition stage from slight discharge to severe discharge, thereby issuing early warnings to avoid more serious damage to equipment due to untimely maintenance; when it is confirmed that the slight discharge has been eliminated, unnecessary maintenance or repair work can be reduced, thereby optimizing the allocation of human and material resources; the prediction result of identifying severe discharge is used as a pre-step to identify the type of partial discharge, and by combining Gram angle field conversion and a partial discharge classification model, the type of partial discharge can be more accurately identified, and the possibility of identification failure or inaccurate identification is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0050] Figure 1 It is a flow chart of the ultrasonic-based partial discharge identification method provided by the present invention;

[0051] Figure 2 It is a structural schematic diagram of a partial discharge identification device based on ultrasound provided by the present invention;

[0052] Figure 3 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] Figure 1 : is a flow chart of a partial discharge identification method based on ultrasound provided by the present invention, wherein the partial discharge identification method based on ultrasound comprises:

[0055] Step 101, calibrate the ultrasonic signal of a first preset duration, determine the state information corresponding to each identification feature at each moment in the first preset duration, the identification feature includes at least one of an amplitude feature, a frequency feature, and a phase feature, and the ultrasonic signal of the first preset duration is obtained from any ultrasonic sensor on the GIS device in the current preset time window;

[0056] Step 102: for each identification feature, construct an initial state transfer matrix, calculate the transition probability of each state information transferring to any state information in each state information corresponding to different time points, obtain all state transfer probabilities, fill all state transfer probabilities into the initial state transfer matrix, obtain a state transfer probability matrix, traverse all identification features, and obtain a state transfer probability matrix corresponding to each identification feature;

[0057] Step 103, inputting all state transition probability matrices into a preset partial discharge identification model to obtain a partial discharge prediction result output by the preset partial discharge identification model;

[0058] The preset partial discharge recognition model is determined after training based on all state transition probability matrices corresponding to all ultrasonic signal samples and the partial discharge prediction results corresponding to each ultrasonic signal sample, and the partial discharge prediction results include no discharge, slight discharge and severe discharge.

[0059] In step 101, the present invention first predefines a first preset time length, which should be sufficient to capture the ultrasonic signal characteristics generated by partial discharge in the GIS, while avoiding being too long to cause excessive data processing. For example, 10 milliseconds can be selected as the first preset time length; then, the ultrasonic sensor on the GIS device collects ultrasonic signals within the current preset time window.

[0060] The identification features include at least one of amplitude features, frequency features, and phase features. For each moment within these 10 milliseconds, the identification features such as amplitude, frequency, and phase are extracted, and the state information corresponding to these features is determined. The state information can be discrete, such as the amplitude is divided into three levels: low, medium, and high, or continuous, such as the specific value of the amplitude. In an optional embodiment, assuming that the first preset time length is 10 milliseconds, sampling is performed once per millisecond, and a total of 10 data points are obtained. By accurately calibrating the ultrasonic signal characteristics within the time period, a detailed data foundation is provided for the subsequent construction of the state transfer matrix.

[0061] Optionally, when the identification feature is an amplitude feature, the calibrating the ultrasonic signal of the first preset duration to determine the state information corresponding to each identification feature at each moment in the first preset duration includes:

[0062] For each moment, determining the amplitude value at said moment;

[0063] When the amplitude value is less than the first preset amplitude value, determining that the amplitude characteristic at the moment is in a low amplitude state;

[0064] When the amplitude value is greater than or equal to the first preset amplitude value and less than or equal to the second preset amplitude value, determining that the amplitude characteristic at the moment is in a medium amplitude state;

[0065] When the amplitude value is greater than the second preset amplitude value, it is determined that the amplitude characteristic at the moment is in a high amplitude state.

[0066] Optionally, the amplitude feature is extracted at each moment, the first preset amplitude value is 10mV, the second preset amplitude value is 50mV, and the amplitude state information is divided into three levels: low amplitude state, medium amplitude state, and high amplitude state, wherein the judgment standard of the low amplitude state is less than 10mV, the judgment standard of the medium amplitude state is greater than or equal to 10mV and less than or equal to 50mV, and the judgment standard of the high amplitude state is greater than 50mV. Taking the amplitude feature as an example, assuming that there are three states (low, medium, and high), the initial state transfer matrix is ​​a 3x3 matrix. By counting the changes in the amplitude state within 10 milliseconds, the probability of all possible state transitions from "low" to "low", "low" to "medium", etc. is calculated and filled in the matrix.

[0067] Optionally, when the identification feature is a frequency feature, the calibrating the ultrasonic signal of the first preset duration to determine the state information corresponding to each identification feature at each moment in the first preset duration includes:

[0068] Determine the frequency value corresponding to each moment;

[0069] When the frequency difference at any adjacent time is less than or equal to the preset frequency, determining that the frequency characteristic at the time is a frequency stable state;

[0070] When the frequency difference at any adjacent time moments is greater than the preset frequency, it is determined that the frequency characteristic at the time moment is in a frequency fluctuation state.

[0071] Optionally, on the ultrasonic sensor of the GIS device, pay attention to the data within the last 10 milliseconds, obtain the data points corresponding to each millisecond, a total of 10 data points, each data point represents the frequency value at a moment, and for these 10 data points, determine the frequency value at each moment. For example, the following frequency sequence (unit: kHz) may be obtained: [20.1, 20.2, 20.1, 20.3, 20.0, 20.2, 20.1, 20.4, 19.8, 20.0]. Next, it is necessary to judge the state of the frequency characteristics based on the frequency difference between adjacent moments. Assuming that the preset frequency is 0.3kHz, it is considered that when the frequency change between adjacent moments exceeds 0.3kHz, the frequency is in a fluctuating state; otherwise, the frequency is in a stable state. For the first pair of adjacent moments (20.1kHz and 20.2kHz), the frequency difference is 0.1kHz, which is less than the preset frequency of 0.3kHz, so the frequency characteristics of these two moments are in a stable frequency state.

[0072] Similarly, we check the frequency difference between each pair of adjacent moments and determine their states; assuming that in the above frequency sequence, only the frequency difference between the 7th moment and the 8th moment (20.1kHz to 20.4kHz) is greater than the preset frequency (0.3kHz), so the frequency characteristics of this pair of moments are determined to be in a frequency fluctuation state, and the rest of the moments are in a frequency stable state. By distinguishing between the stable state and the fluctuating state of the frequency, the ultrasonic frequency change characteristics caused by partial discharge can be captured more accurately, thereby improving the accuracy of partial discharge identification. Refining the frequency characteristics into two states, stable and fluctuating, helps to distinguish the real frequency changes caused by partial discharge from the random fluctuations caused by other factors (such as environmental noise), thereby reducing misjudgment.

[0073] Optionally, when the identification feature is a phase feature, the calibrating the ultrasonic signal of the first preset time length to determine the state information corresponding to each identification feature at each moment in the first preset time length includes:

[0074] When the phase at any moment remains unchanged from that at the previous moment, determining that the phase characteristic at the moment is in a phase-unchanged state;

[0075] When the phase at any moment changes from that at the previous moment, the phase characteristic at the moment is determined to be a phase change state.

[0076] Similarly, the first preset duration is set to 10 milliseconds, and data is collected from the ultrasonic sensor of the GIS equipment once per second. Pay attention to the data within 10 milliseconds. The data within 10 milliseconds contains the phase information of each moment. For these 10 data points, compare the phase value of each moment with the previous moment one by one. If the phase value at a certain moment is the same as or very close to the phase value at the previous moment (taking into account the measurement error, a very small threshold can be set, such as 0.1 degrees or less), the phase feature at that moment is considered to be "phase unchanged state". If the phase value at a certain moment is significantly different from the phase value at the previous moment (that is, greater than the set threshold), the phase feature at that moment is considered to be "phase change state".

[0077] Optionally, changes in phase characteristics can often sensitively reflect subtle changes in ultrasonic signals, thereby helping to detect signs of partial discharge earlier. By distinguishing between the unchanged and changed states of the phase, false alarms caused by environmental noise or other non-local discharge factors can be reduced. In partial discharge identification, combining multiple features such as amplitude, frequency and phase can build a more comprehensive and accurate identification model, thereby improving the reliability and accuracy of identification.

[0078] In step 102, for each identification feature (such as amplitude, frequency, phase) extracted in step 101, an initial state transfer matrix is ​​constructed respectively, and the element value in the initial state transfer matrix represents the probability of transitioning from one state to another. By counting the transitions between states within a first preset time length, the state transition probability is calculated and filled into the initial state transfer matrix.

[0079] Optionally, constructing an initial state transfer matrix includes:

[0080] When the identification feature is an amplitude feature, constructing a 3×3 initial state transfer matrix according to the amount of state information of the amplitude feature;

[0081] In the case where the identification feature is a frequency feature or a phase feature, a 2×2 initial state transfer matrix is ​​constructed according to the amount of state information of the frequency feature or the phase feature.

[0082] Optionally, for amplitude features, their states can usually be divided into three levels or categories, such as "low", "medium" and "high". Therefore, in the case of amplitude features, a 3×3 initial state transfer matrix is ​​constructed, each row and column of this matrix represents an amplitude state, and the elements in the matrix represent the probability of transitioning from one state to another; for frequency features and phase features, their states can usually be simplified to two types: stable or unchanged, and fluctuating or changing. Therefore, in the case of frequency or phase features, a 2×2 initial state transfer matrix is ​​constructed. Similarly, each row and column represents a state, and the elements in the matrix represent the transition probability between states.

[0083] Optionally, at the time of initialization, if an equal probability method is used, the probability of each state transferring to any other state is 1 / 3. The initial state transfer matrix A of the amplitude feature can be expressed as:

[0084]

[0085] The dimension of the frequency state transfer matrix F is 2x2, because the frequency has two states. Similarly, it is initialized with equal probability:

[0086]

[0087] The dimension of the phase state transfer matrix P is also 2x2, because the phase has two states, and the equal probability is initialized as follows:

[0088]

[0089] Optionally, the calculating of the transition probability of each state information to any state information in each state information corresponding to different time instants to obtain all state transition probabilities includes:

[0090] For each state information, obtaining the number of transitions from the state information to any state information, and determining the state transition probability according to the number of transitions and the total number of transitions of the state information;

[0091] Traverse all state information and obtain all state transition probabilities.

[0092] Optionally, for each state pair (i, j), the number of transitions from state i to state j in the labeled data is counted, for each state i, the total number of transitions from the state to any state is calculated, the number of transitions for each state pair (i, j) is divided by the total number of transitions from state i, to obtain a probability estimate of transitioning from state i to state j, and determining the state transition probability according to the number of transitions and the total number of transitions occurring in the state information includes:

[0093]

[0094] Wherein, P(j|i) is the state transition probability from state information i to state information j, M is the number of transitions from state information i to state information j, and N is the total number of transitions from state information i to any state.

[0095] The present invention extracts the amplitude, frequency or phase characteristics at each moment from the ultrasonic signal and determines the state information corresponding to them. For each state information, a transfer count matrix or a similar data structure is maintained to record the number of times the state is transferred to all other possible states. The dimension of the matrix should match the initial state transfer matrix, and the total number of times each state information is transferred is recorded. For each state information, the number of times it is transferred to other states is divided by the total number of transfers to the state to calculate each possible transfer probability. The calculated transfer probability is used to update the initial state transfer matrix to obtain the final state transfer probability matrix.

[0096] In an optional embodiment, it is assumed that some observation data are obtained, which record the number of transitions between states. Then, the total number of transitions for each state is calculated, and the number of transitions for each state is divided by the total number of transitions for the state to obtain the transition probability:

[0097] Low middle high Low 0 5 2 middle 3 0 4 high 1 1 0

[0098] Then, calculate the total number of transitions to each state:

[0099] The low state is 0+5+2, which is 7 times, the middle state is 3+0+4, which is 7 times, and the high state is 1+1+0, which is 2 times. Divide the number of transitions of each state by the total number of transitions of that state to obtain the transition probability matrix:

[0100] Low middle high Low 0 / 7 5 / 7 2 / 7 middle 3 / 7 0 / 7 4 / 7 high 1 / 2 1 / 2 0 / 2

[0101] To more intuitively represent probabilities (i.e. decimals or percentages), the above matrix can be transformed into:

[0102] Low middle high Low 0.00％ 71.43％ 28.57％ middle 42.86％ 0.00％ 57.14％ high 50.00％ 50.00％ 0.00％

[0103] In step 103, in the training stage, a large number of ultrasonic signal samples including partial discharge events and no-discharge events are used to construct a state transition probability matrix according to the methods of steps 101 and 102. Then, the preset partial discharge recognition model is trained using these matrices and the corresponding partial discharge prediction results (no discharge, slight discharge, severe discharge). In the test or application stage, the state transition probability matrix of the newly collected ultrasonic signal is input into the trained model to obtain the partial discharge prediction result, that is, "all state transition probability matrices are input into the preset partial discharge recognition model to obtain the partial discharge prediction result output by the preset partial discharge recognition model." The preset partial discharge recognition model is determined after training based on all state transition probability matrices corresponding to all ultrasonic signal samples and the partial discharge prediction result corresponding to each ultrasonic signal sample, and the partial discharge prediction result includes no discharge, slight discharge and severe discharge.

[0104] The present invention can adopt machine learning algorithms, such as random forests, neural networks, etc., as preset local discharge recognition models. During the training process, the model parameters are continuously adjusted to optimize the prediction performance. By constructing and training the local discharge recognition model, intelligent recognition of local discharge events in GIS equipment is achieved. The present invention uses matrices and corresponding local discharge prediction results to train the preset local discharge recognition model, which not only improves the accuracy and efficiency of recognition, but also reduces the possibility of manual intervention and misjudgment.

[0105] Optionally, after obtaining the partial discharge prediction result output by the preset partial discharge identification model, the method further includes:

[0106] When it is determined that the partial discharge prediction result is a slight discharge, obtaining a next ultrasonic signal corresponding to a next time window;

[0107] When it is determined that the local discharge prediction result corresponding to the next ultrasonic signal is no discharge, updating the local discharge prediction result to no discharge;

[0108] When it is determined that the local discharge prediction result corresponding to the next ultrasonic signal is slight discharge or severe discharge, the local discharge prediction result is updated to severe discharge.

[0109] Optionally, the collected ultrasonic signal is analyzed by a preset local discharge recognition model to output a prediction result of local discharge, wherein the local discharge prediction result includes no discharge, slight discharge, and severe discharge. When the model outputs a prediction result of slight discharge for the first time, the next ultrasonic signal corresponding to the next time window is obtained without directly making a final judgment. The local discharge recognition model is used to predict the new ultrasonic signal again. If the new prediction result is no discharge, it means that the previous slight discharge may be accidental or has been eliminated by itself. Therefore, the system updates the final local discharge prediction result to no discharge. If the new prediction result is still slight discharge or is upgraded to severe discharge, the system believes that the slight discharge has a tendency to continue or intensify. Therefore, the final local discharge prediction result is updated to severe discharge to facilitate the determination of subsequent local discharge classification results.

[0110] Optionally, the present invention reduces false alarms or missed alarms caused by a single misjudgment through continuous monitoring and re-prediction, improves the accuracy of local discharge identification, and can more sensitively capture changes in discharge trends in the transition stage from slight discharge to severe discharge, thereby issuing early warnings to avoid more serious damage to equipment due to untimely maintenance; when it is confirmed that the slight discharge has been eliminated, unnecessary maintenance or repair work can be reduced, thereby optimizing the allocation of human and material resources.

[0111] Next, when it is determined that the partial discharge prediction result is severe discharge, the method further includes:

[0112] After acquiring the ultrasonic signal of the first preset time length, acquiring the target ultrasonic signal of the third preset time length after an interval of the second preset time length;

[0113] Processing the target ultrasonic signal according to Gram angular field conversion to obtain a GAF image;

[0114] Inputting the GAF image into a partial discharge classification model to obtain a partial discharge classification result output by the partial discharge classification model;

[0115] The partial discharge classification model is obtained after training based on all GAF sample images and the partial discharge classification results corresponding to each GAF sample image, wherein the partial discharge classification results include corona discharge, particle discharge, suspension discharge and cavitation discharge;

[0116] The first preset duration is the same as or different from the second preset duration, and the second preset duration is the same as or different from the third preset duration.

[0117] Optionally, after determining that the local discharge prediction result is severe discharge, the ultrasonic signal of a specific time period is obtained, the signal is converted into an image using the Gramian Angular Field (GAF) conversion technology, and the specific local discharge type is identified with the help of a trained local discharge classification model. Specifically, in order to avoid noise or interference in the signal, or to wait for further development of the discharge phenomenon, after waiting for the second preset time, the target ultrasonic signal of the third preset time is obtained to ensure that stable and representative discharge characteristics can be captured, and the target ultrasonic signal of the third preset time is converted into a Gramian Angular Field (GAF).

[0118] Optionally, GAF is a method for converting time series data into image representation, by calculating the cumulative sum of the time series and mapping it into a polar coordinate system, and then converting it into a two-dimensional image. This method can retain the dynamic characteristics and correlation of the time series, and is easy to analyze using image processing technology. The obtained GAF ​​image is input into a partial discharge classification model. This model is trained based on a large number of GAF sample images and their corresponding partial discharge classification results. It can automatically identify the discharge features in the image and output the corresponding partial discharge classification results. The partial discharge classification results usually include types such as corona discharge, particle discharge, suspension discharge, and cavitation discharge. These types are of great significance for diagnosing equipment failures and formulating maintenance strategies.

[0119] Optionally, the prediction result of identifying severe discharge is used as a pre-step to identify the type of partial discharge. By combining Gram angle field transformation and the partial discharge classification model, the type of partial discharge can be identified more accurately and the possibility of identification failure or inaccurate identification can be reduced.

[0120] Figure 2 It is a structural schematic diagram of an ultrasonic-based partial discharge identification device provided by the present invention, wherein the ultrasonic-based partial discharge identification device comprises a determination unit 1, wherein the determination unit is used to calibrate an ultrasonic signal of a first preset duration, and determine state information corresponding to each identification feature at each moment in the first preset duration, wherein the identification feature comprises at least one of an amplitude feature, a frequency feature, and a phase feature, and the ultrasonic signal of the first preset duration is obtained from any ultrasonic sensor on the GIS device in a current preset time window, and the working principle of the determination unit 1 may refer to the aforementioned step 101 and will not be elaborated herein.

[0121] The ultrasonic-based partial discharge identification device also includes a construction unit 2, which is used to construct an initial state transfer matrix for each identification feature, calculate the transition probability of each state information transferring to any state information in each state information corresponding to different times, obtain all state transfer probabilities, fill all state transfer probabilities into the initial state transfer matrix, obtain a state transfer probability matrix, traverse all identification features, and obtain a state transfer probability matrix corresponding to each identification feature. The working principle of the construction unit 2 can refer to the aforementioned step 102 and will not be repeated here.

[0122] The ultrasonic-based partial discharge identification device further includes an input unit 3, which is used to input all state transition probability matrices into a preset partial discharge identification model to obtain a partial discharge prediction result output by the preset partial discharge identification model. The working principle of the input unit 3 can be referred to the aforementioned step 103 and will not be described in detail herein.

[0123] The preset partial discharge recognition model is determined after training based on all state transition probability matrices corresponding to all ultrasonic signal samples and the partial discharge prediction results corresponding to each ultrasonic signal sample, and the partial discharge prediction results include no discharge, slight discharge and severe discharge.

[0124] The present invention realizes accurate identification and classification of partial discharge phenomena in GIS equipment through detailed feature analysis and construction of state transition probability matrix, combined with a preset partial discharge identification model. Compared with the prior art, the present invention can quickly screen out data that does not require in-depth analysis by first judging the severity of the discharge, thereby reducing unnecessary processing of such data and improving overall identification efficiency; after confirming the existence of severe discharge, detailed discharge type identification is performed to ensure that limited computing resources and time are invested in the most needed areas, thereby optimizing resource allocation; for severe discharge, since its signal characteristics are usually more obvious and complex, the present invention can obtain higher accuracy when identifying the discharge type, and the early determination of no discharge or slight discharge can avoid unnecessary complex analysis of them, thereby reducing misjudgment; the judgment criteria and subsequent processing procedures of severe discharge can be adjusted according to actual needs to better adapt to different application scenarios.

[0125] Figure 3 Schematic diagram of the structure of the electronic device provided by the present invention. Figure 3As shown, the electronic device may include: a processor (processor) 310, a communication interface (CommunicationsInterface) 320, a memory (memory) 330 and a communication bus 340, wherein the processor 310, the communication interface 320, and the memory 330 communicate with each other through the communication bus 340. The processor 310 can call the logic instructions in the memory 330 to execute the ultrasonic-based partial discharge identification method, which includes: calibrating the ultrasonic signal of a first preset time length, determining the state information corresponding to each identification feature at each moment in the first preset time length, the identification feature includes at least one of an amplitude feature, a frequency feature and a phase feature, and the ultrasonic signal of the first preset time length is obtained from any ultrasonic sensor on the GIS equipment in the current preset time window; for each identification feature, constructing an initial state transfer matrix, calculating the transition probability of each state information transferring to any state information in each state information corresponding to different moments, obtaining all state transfer probabilities, filling all state transfer probabilities into the initial state transfer matrix, obtaining a state transfer probability matrix, traversing all identification features, and obtaining a state transfer probability matrix corresponding to each identification feature; inputting all state transfer probability matrices into a preset partial discharge identification model to obtain a partial discharge prediction result output by the preset partial discharge identification model; wherein the preset partial discharge identification model is determined after training based on all state transfer probability matrices corresponding to all ultrasonic signal samples and the partial discharge prediction result corresponding to each ultrasonic signal sample, and the partial discharge prediction result includes no discharge, slight discharge and severe discharge.

[0126] In addition, the logic instructions in the above-mentioned memory 330 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0127] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute a partial discharge identification method based on ultrasound provided by the above methods, the method comprising: calibrating an ultrasonic signal of a first preset duration, determining state information corresponding to each identification feature at each moment in the first preset duration, the identification feature comprising at least one of an amplitude feature, a frequency feature, and a phase feature, the ultrasonic signal of the first preset duration is obtained from any ultrasonic sensor on the GIS device in the current preset time window; for each identification feature, constructing an initial state transfer matrix, In each state information corresponding to different moments, the transition probability of each state information transferring to any state information is calculated to obtain all state transition probabilities, all state transition probabilities are filled into the initial state transfer matrix to obtain the state transfer probability matrix, all identification features are traversed to obtain the state transfer probability matrix corresponding to each identification feature; all state transfer probability matrices are input into a preset partial discharge recognition model to obtain the partial discharge prediction result output by the preset partial discharge recognition model; wherein the preset partial discharge recognition model is determined after training based on all state transfer probability matrices corresponding to all ultrasonic signal samples and the partial discharge prediction result corresponding to each ultrasonic signal sample, and the partial discharge prediction result includes no discharge, slight discharge and severe discharge.

[0128] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the ultrasonic-based partial discharge identification method provided by the above-mentioned methods, the method comprising: calibrating an ultrasonic signal of a first preset duration, determining state information corresponding to each identification feature at each moment in the first preset duration, the identification feature comprising at least one of an amplitude feature, a frequency feature, and a phase feature, the ultrasonic signal of the first preset duration being obtained from any ultrasonic sensor on the GIS device in a current preset time window; for each identification feature, constructing an initial state transfer matrix, and in each state information corresponding to different moments Calculate the transition probability of each state information transferring to any state information to obtain all state transition probabilities, fill all state transition probabilities into the initial state transfer matrix to obtain the state transfer probability matrix, traverse all identification features to obtain the state transfer probability matrix corresponding to each identification feature; input all state transfer probability matrices into a preset partial discharge recognition model to obtain the partial discharge prediction result output by the preset partial discharge recognition model; wherein the preset partial discharge recognition model is determined after training based on all state transfer probability matrices corresponding to all ultrasonic signal samples and the partial discharge prediction result corresponding to each ultrasonic signal sample, and the partial discharge prediction result includes no discharge, slight discharge and severe discharge.

[0129] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0130] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0131] Finally, 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 aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A partial discharge identification method based on ultrasound, characterized in that: include: Calibrate an ultrasonic signal of a first preset duration, determine state information corresponding to each identification feature at each moment in the first preset duration, the identification feature comprising at least one of an amplitude feature, a frequency feature, and a phase feature, and the ultrasonic signal of the first preset duration is obtained from any ultrasonic sensor on the GIS device in a current preset time window; For each identification feature, an initial state transfer matrix is ​​constructed, and the transition probability of each state information transferring to any state information is calculated in each state information corresponding to different times, and all state transfer probabilities are obtained. All state transfer probabilities are filled into the initial state transfer matrix to obtain a state transfer probability matrix, and all identification features are traversed to obtain a state transfer probability matrix corresponding to each identification feature; Inputting all state transition probability matrices into a preset partial discharge identification model to obtain a partial discharge prediction result output by the preset partial discharge identification model; The preset partial discharge recognition model is determined after training based on all state transition probability matrices corresponding to all ultrasonic signal samples and the partial discharge prediction results corresponding to each ultrasonic signal sample, and the partial discharge prediction results include no discharge, slight discharge and severe discharge.

2. The ultrasonic-based partial discharge identification method according to claim 1, characterized in that: In the case where the identification feature is an amplitude feature, the calibrating of the ultrasonic signal of the first preset duration, determining the state information corresponding to each identification feature at each moment in the first preset duration, includes: For each moment, determining the amplitude value at said moment; When the amplitude value is less than the first preset amplitude value, determining that the amplitude characteristic at the moment is in a low amplitude state; When the amplitude value is greater than or equal to the first preset amplitude value and less than or equal to the second preset amplitude value, determining that the amplitude characteristic at the moment is in a medium amplitude state; When the amplitude value is greater than the second preset amplitude value, it is determined that the amplitude characteristic at the moment is in a high amplitude state.

3. The ultrasonic-based partial discharge identification method according to claim 1, characterized in that: In the case where the identification feature is a frequency feature, the calibration of the ultrasonic signal of the first preset duration, determining the state information corresponding to each identification feature at each moment in the first preset duration, includes: Determine the frequency value corresponding to each moment; When the frequency difference at any adjacent time is less than or equal to the preset frequency, determining that the frequency characteristic at the time is a frequency stable state; When the frequency difference at any adjacent time moments is greater than the preset frequency, it is determined that the frequency characteristic at the time moment is in a frequency fluctuation state.

4. The ultrasonic-based partial discharge identification method according to claim 1, characterized in that: In the case where the identification feature is a phase feature, the calibrating of the ultrasonic signal of the first preset time length, determining the state information corresponding to each identification feature at each moment in the first preset time length, includes: When the phase at any moment remains unchanged from that at the previous moment, determining that the phase characteristic at the moment is in a phase-unchanged state; When the phase at any moment changes from that at the previous moment, the phase characteristic at the moment is determined to be a phase change state.

5. The ultrasonic-based partial discharge identification method according to claim 1, characterized in that: The constructing of the initial state transfer matrix comprises: When the identification feature is an amplitude feature, constructing a 3×3 initial state transfer matrix according to the amount of state information of the amplitude feature; In the case where the identification feature is a frequency feature or a phase feature, a 2×2 initial state transfer matrix is ​​constructed according to the amount of state information of the frequency feature or the phase feature.

6. The ultrasonic-based partial discharge identification method according to claim 1, characterized in that: The calculation of the transition probability of each state information to any state information in each state information corresponding to different time instants, and obtaining all state transition probabilities, includes: For each state information, obtaining the number of transitions from the state information to any state information, and determining the state transition probability according to the number of transitions and the total number of transitions of the state information; Traverse all state information and obtain all state transition probabilities.

7. The ultrasonic-based partial discharge identification method according to claim 6, characterized in that: The determining the state transition probability according to the number of transitions and the total number of times the state information is transitioned includes: Wherein, P(j|i) is the state transition probability from state information i to state information j, M is the number of transitions from state information i to state information j, and N is the total number of transitions from state information i to any state.

8. The ultrasonic-based partial discharge identification method according to claim 1, characterized in that: After obtaining the partial discharge prediction result output by the preset partial discharge identification model, the method further includes: When it is determined that the partial discharge prediction result is a slight discharge, obtaining a next ultrasonic signal corresponding to a next time window; When it is determined that the local discharge prediction result corresponding to the next ultrasonic signal is no discharge, updating the local discharge prediction result to no discharge; When it is determined that the local discharge prediction result corresponding to the next ultrasonic signal is slight discharge or severe discharge, the local discharge prediction result is updated to severe discharge.

9. The ultrasonic-based partial discharge identification method according to claim 8, characterized in that: In the case where it is determined that the partial discharge prediction result is severe discharge, the method further includes: After acquiring the ultrasonic signal of the first preset time length, acquiring the target ultrasonic signal of the third preset time length after an interval of the second preset time length; Processing the target ultrasonic signal according to Gram angular field conversion to obtain a GAF image; Inputting the GAF image into a partial discharge classification model to obtain a partial discharge classification result output by the partial discharge classification model; The partial discharge classification model is obtained after training based on all GAF sample images and the partial discharge classification results corresponding to each GAF sample image, wherein the partial discharge classification results include corona discharge, particle discharge, suspension discharge and cavitation discharge; The first preset duration is the same as or different from the second preset duration, and the second preset duration is the same as or different from the third preset duration.

10. A partial discharge identification device based on ultrasound, characterized in that: include: A determination unit, the determination unit is used to calibrate an ultrasonic signal of a first preset duration, determine state information corresponding to each identification feature at each moment in the first preset duration, the identification feature comprising at least one of an amplitude feature, a frequency feature, and a phase feature, and the ultrasonic signal of the first preset duration is obtained from any ultrasonic sensor on the GIS device in a current preset time window; A construction unit, wherein the construction unit is used to construct an initial state transfer matrix for each identification feature, calculate the transition probability of each state information transferring to any state information in each state information corresponding to different time periods, obtain all state transfer probabilities, fill all state transfer probabilities into the initial state transfer matrix, obtain a state transfer probability matrix, traverse all identification features, and obtain a state transfer probability matrix corresponding to each identification feature; An input unit, the input unit is used to input all state transition probability matrices into a preset partial discharge identification model to obtain a partial discharge prediction result output by the preset partial discharge identification model; The preset partial discharge recognition model is determined after training based on all state transition probability matrices corresponding to all ultrasonic signal samples and the partial discharge prediction results corresponding to each ultrasonic signal sample, and the partial discharge prediction results include no discharge, slight discharge and severe discharge.

Images