CN113701684B - Transformer winding state detection method, device, equipment and storage medium - Google Patents
Transformer winding state detection method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN113701684B CN113701684B CN202110898223.4A CN202110898223A CN113701684B CN 113701684 B CN113701684 B CN 113701684B CN 202110898223 A CN202110898223 A CN 202110898223A CN 113701684 B CN113701684 B CN 113701684B
- Authority
- CN
- China
- Prior art keywords
- vibration signal
- transformer winding
- modal
- mode
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B17/00—Measuring arrangements characterised by the use of infrasonic, sonic or ultrasonic vibrations
- G01B17/04—Measuring arrangements characterised by the use of infrasonic, sonic or ultrasonic vibrations for measuring the deformation in a solid, e.g. by vibrating string
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Abstract
The application relates to a method, a device, equipment and a storage medium for detecting the state of a transformer winding, wherein the method comprises the following steps: performing characteristic extraction according to the vibration signal of the target transformer winding to obtain a characteristic matrix corresponding to the vibration signal of the target transformer winding; inputting the characteristic matrix into a preset machine learning model to obtain a characteristic label of the target transformer winding; different characteristic labels represent different deformation levels of the transformer winding; and determining the state of the target transformer winding according to the characteristic label. The technical scheme provided by the embodiment of the application can improve the flexibility of detecting the state of the transformer winding.
Description
Technical Field
The present application relates to the field of high voltage direct current transmission technologies, and in particular, to a method, an apparatus, a device, and a storage medium for detecting a state of a transformer winding.
Background
The transformer is one of core devices in the high-voltage direct-current transmission project, and the operation state of the transformer directly influences the operation of the whole power system. When the short-circuit current is generated in the transformer, the transformer winding is easy to deform, so that the transformer has mechanical failure. If the mechanical fault of the transformer cannot be found and processed in time, the whole power system is threatened greatly. In practical application, whether the transformer has a fault or not can be effectively diagnosed by detecting the state of the transformer winding.
In the related art, when the state of the transformer winding is detected, usually, the voltage signal of the transformer is collected in the transformer power-off state, the collected voltage signal is processed, and the processing result is compared with a normal value, so that whether the transformer winding deforms or not is judged according to the comparison result.
However, the related art has poor flexibility in detecting the state of the transformer winding.
Disclosure of Invention
Based on this, the embodiments of the present application provide a method, an apparatus, a device, and a storage medium for detecting a state of a transformer winding, which can improve flexibility of detecting the state of the transformer winding.
In a first aspect, a method for detecting a winding state of a transformer is provided, the method comprising:
performing characteristic extraction according to the vibration signal of the target transformer winding to obtain a characteristic matrix corresponding to the vibration signal of the target transformer winding; inputting the characteristic matrix into a preset machine learning model to obtain a characteristic label of the target transformer winding; different feature labels represent different deformation levels of the transformer winding; and determining the state of the target transformer winding according to the characteristic label.
In one embodiment, the extracting the feature according to the vibration signal of the target transformer winding to obtain the feature matrix corresponding to the vibration signal of the target transformer winding includes:
respectively acquiring the modal energy ratio, euclidean distance and frequency spectrum complexity of the vibration signal according to the vibration signal of the target transformer winding; and generating a characteristic matrix corresponding to the vibration signal of the target transformer winding based on the modal energy ratio, the Euclidean distance and the frequency spectrum complexity.
In one embodiment, obtaining the modal energy ratio of the vibration signal comprises:
performing empirical mode decomposition on the vibration signal to obtain a mode decomposition result of the vibration signal; the modal decomposition result comprises a plurality of intrinsic modal components and a residual component; reconstructing a modal decomposition result of the vibration signal to obtain a reconstruction modal; the reconstruction modes comprise a high-frequency mode, a medium-frequency mode and a low-frequency mode; and calculating the energy of the reconstruction mode, and determining the mode energy ratio of the vibration signal according to the energy of the reconstruction mode.
In one embodiment, reconstructing a mode corresponding to the vibration signal to obtain a reconstructed mode corresponding to the mode includes:
acquiring the zero crossing rate of each component in the modal decomposition result of the vibration signal; determining the sum of components with zero crossing rates larger than a first zero crossing rate threshold value in the zero crossing rates of the components as a high-frequency mode; determining the sum of components of which the zero crossing rate is less than or equal to a first zero crossing rate threshold and is greater than or equal to a second zero crossing rate threshold as an intermediate frequency mode; and determining the sum of the components with the zero crossing rate smaller than a second zero crossing rate threshold value in the zero crossing rates of the components as a low-frequency mode.
In one embodiment, the Euclidean distance of the vibration signal is acquired, and the Euclidean distance comprises the following steps:
acquiring the reference modal energy ratio of the vibration signal of the normal transformer winding; and calculating the Euclidean distance of the vibration signal according to the modal energy ratio of the vibration signal of the target transformer winding and the reference modal energy ratio.
In one embodiment, obtaining the spectral complexity of the vibration signal comprises:
acquiring the frequency proportion of the vibration signal; the frequency proportion is used for representing the proportion of harmonic components of the vibration signal at the target frequency; calculating the frequency spectrum complexity of the vibration signal based on the frequency proportion of the vibration signal; the spectral complexity is used to characterize the complexity of the target frequency.
In one embodiment, the method further includes:
acquiring a training sample set; the training sample set comprises a feature matrix corresponding to vibration signals of a plurality of transformer windings and a target feature label corresponding to each transformer winding; inputting the training sample set into an initial machine learning model to obtain a prediction characteristic label of each transformer winding; and updating the parameters of the initial machine learning model based on the predicted characteristic label and the target characteristic label until a preset convergence condition is reached, and generating a preset machine learning model based on the updated parameters.
In a second aspect, there is provided a transformer winding state detection apparatus, the apparatus comprising:
the characteristic extraction module is used for extracting characteristics according to the vibration signals of the target transformer winding to obtain a characteristic matrix corresponding to the vibration signals of the target transformer winding;
the first input module is used for inputting the characteristic matrix into a preset machine learning model to obtain a characteristic label of a target transformer winding; different feature labels represent different deformation levels of the transformer winding;
and the determining module is used for determining the state of the target transformer winding according to the characteristic label.
In a third aspect, a computer device is provided, comprising a memory and a processor, the memory storing a computer program, and the computer program, when executed by the processor, implementing the method steps in any of the embodiments of the first aspect.
In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored, which computer program, when being executed by a processor, carries out the method steps of any of the embodiments of the first aspect described above.
According to the transformer winding state detection method, the device, the equipment and the storage medium, feature extraction is carried out according to the vibration signal of the target transformer winding, so that a feature matrix corresponding to the vibration signal of the target transformer winding is obtained; inputting the characteristic matrix into a preset machine learning model to obtain a characteristic label of the target transformer winding; and determining the state of the target transformer winding according to the characteristic label. In the technical scheme provided by the embodiment of the application, compared with the traditional method, the collected transformer winding vibration signals can be analyzed in real time, and the characteristic matrix corresponding to the vibration signals is calculated according to the pre-trained machine learning model, so that whether the transformer winding is deformed or not is judged, and the flexibility of detecting the state of the transformer winding is improved.
Drawings
FIG. 1 is a block diagram of a computer device provided by an embodiment of the present application;
fig. 2 is a flowchart of a method for detecting a state of a transformer winding according to an embodiment of the present disclosure;
fig. 3 is a flowchart of generating a feature matrix according to an embodiment of the present application;
fig. 4 is a flowchart of determining a modal energy ratio of a vibration signal according to an embodiment of the present application;
fig. 5 is a flowchart for determining a reconstruction modality according to an embodiment of the present application;
fig. 6 is a flowchart for calculating a euclidean distance of a vibration signal according to an embodiment of the present disclosure;
FIG. 7 is a flowchart for calculating the complexity of the frequency spectrum of the vibration signal according to an embodiment of the present disclosure;
fig. 8 is a flowchart of generating a preset machine learning model according to an embodiment of the present disclosure;
fig. 9 is a flowchart of a method for detecting a state of a transformer winding according to an embodiment of the present disclosure;
fig. 10 is a block diagram of a transformer winding state detection apparatus according to an embodiment of the present application.
Detailed Description
To make the objects, technical solutions and advantages of the present application more clear, the following detailed description of the embodiments of the present application will be made with reference to the accompanying drawings.
The transformer winding state detection method provided by the application can be applied to computer equipment, the computer equipment can be a server or a terminal, the server can be one server or a server cluster consisting of a plurality of servers, the method is not specifically limited in this embodiment, and the terminal can be but not limited to various personal computers, notebook computers, smart phones, tablet computers and portable wearable equipment.
Taking the example of a computer device being a server, FIG. 1 shows a block diagram of a server, which may include a processor and memory connected by a system bus, as shown in FIG. 1. Wherein the processor of the server is configured to provide computing and control capabilities. The memory of the server comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operating system and the computer program to run on the non-volatile storage medium. The computer program is executed by a processor to implement a transformer winding state detection method.
Those skilled in the art will appreciate that the architecture shown in fig. 1 is a block diagram of only a portion of the architecture associated with the subject application, and does not constitute a limitation on the servers to which the subject application applies, and that servers may alternatively include more or fewer components than those shown, or combine certain components, or have a different arrangement of components.
The execution subject of the embodiments of the present application may be a computer device, or may be a transformer winding state detection apparatus, and the following method embodiments will be described with reference to the computer device as the execution subject.
In one embodiment, as shown in fig. 2, a flowchart of a method for detecting a winding state of a transformer according to an embodiment of the present application is shown, where the method may include the following steps:
and step 220, extracting characteristics according to the vibration signals of the target transformer winding to obtain a characteristic matrix corresponding to the vibration signals of the target transformer winding.
In the operation process of the transformer, the transformer winding can vibrate under the action of radial electromagnetic force and axial electromagnetic force, and whether the transformer winding deforms or not can be judged through vibration signals of the transformer winding. The vibration signal of the target transformer winding can be acquired by a sensor arranged on the surface of the transformer oil tank, and the vibration signal can be acquired in real time or at fixed time intervals.
After the vibration signal of the target transformer winding is acquired, the characteristic extraction can be performed on the vibration signal according to a preset characteristic extraction algorithm, so that a characteristic matrix corresponding to the vibration signal of the target transformer winding is obtained. The feature matrix is a matrix formed by a plurality of feature parameters of the vibration signal, the feature parameters of the vibration signal may include parameters such as a modal energy ratio, a euclidean distance, and a spectrum complexity, and may also include other feature parameters, for example, parameters such as a power of the vibration signal, a parity-odd harmonic ratio, and the like, which is not specifically limited in this embodiment.
The preset machine learning model is a model used for calculating the characteristic matrix and is trained in advance through the characteristic matrix of the vibration signal and the corresponding characteristic label. The characteristic labels are used for representing the deformation levels of the transformer windings, different characteristic labels represent different deformation levels of the transformer windings, the deformation levels of the transformer windings are the deformation degrees of the transformer windings, the deformation degrees can be distinguished by the number of winding deformation positions of the transformer, and the deformation degrees can also be distinguished by other conditions. The feature labels may be represented by numbers, letters, symbols, or other means, and this embodiment is not limited in this respect. For example, the feature labels are represented by numbers, and as the level of distortion increases, the number representing the feature label may increase.
And step 260, determining the state of the target transformer winding according to the characteristic label.
The state of the target transformer winding can include a normal state and a deformation state, and the feature label of the target transformer winding is obtained after the feature matrix of the vibration signal is calculated through a preset machine learning model. The number of the characteristic tags can be multiple, and the states of the transformer windings corresponding to the characteristic tags are a normal state and a deformation state, so that the state of the target transformer winding can be determined according to the characteristic tags. For example, the characteristic labels are represented by numbers 1, 2, ·, and m +1, where "1" represents a transformer winding with zero deformation, and if the characteristic label of the target transformer winding is 1, the state of the target transformer winding can be determined to be a normal state; if the characteristic label of the target transformer winding is any one of values from 2 to m +1, the state of the target transformer winding can be determined to be an abnormal state.
In the embodiment, the characteristic extraction is carried out according to the vibration signal of the target transformer winding to obtain a characteristic matrix corresponding to the vibration signal of the target transformer winding; inputting the characteristic matrix into a preset machine learning model to obtain a characteristic label of the target transformer winding; and determining the state of the target transformer winding according to the characteristic label. Compared with the traditional method, the collected transformer winding vibration signals can be analyzed in real time, and the characteristic matrix corresponding to the vibration signals is calculated according to the machine learning model trained in advance, so that whether the transformer winding is deformed or not is judged, and the flexibility of detecting the state of the transformer winding is improved.
In one embodiment, as shown in fig. 3, which illustrates a flowchart of a method for detecting a winding state of a transformer according to an embodiment of the present application, specifically, a possible process for generating a feature matrix, the method may include the following steps:
and step 320, respectively obtaining the modal energy ratio, euclidean distance and frequency spectrum complexity of the vibration signal according to the vibration signal of the target transformer winding.
And 340, generating a characteristic matrix corresponding to the vibration signal of the target transformer winding based on the modal energy ratio, the Euclidean distance and the frequency spectrum complexity.
The modal energy ratio of the vibration signal is used to represent a ratio of modal energy of different frequencies to total modal energy, and before calculating the modal energy, a preset Decomposition algorithm needs to be used to decompose the acquired vibration signal to obtain a plurality of modes, where the preset Decomposition algorithm may be an Empirical Mode Decomposition algorithm (EMD), a wavelet Decomposition algorithm, or other Decomposition algorithms, and this embodiment is not particularly limited. The Euclidean distance is used for representing the distance between the modal energy ratio of the vibration signal of the target transformer winding and the modal energy ratio of the vibration signal of the normal transformer winding. The frequency spectrum complexity is used for representing the complexity of frequency components in the frequency of the vibration signal, and the lower the value of the frequency complexity is, the more concentrated the energy in the frequency spectrum of the vibration signal is at certain specific frequencies; the higher the value of the frequency complexity, the more dispersed the energy in the spectrum representing the vibration signal. The modal energy occupancy, the Euclidean distance and the spectrum complexity of the vibration signal are obtained, and can be used as characteristic parameters of the vibration signal, so that a characteristic matrix is generated based on the characteristic parameters.
In the embodiment, the modal energy ratio, the Euclidean distance and the frequency spectrum complexity of the vibration signal are respectively obtained according to the vibration signal of the target transformer winding; and generating a characteristic matrix corresponding to the vibration signal of the target transformer winding based on the modal energy ratio, the Euclidean distance and the frequency spectrum complexity. Because the modal energy ratio, the Euclidean distance and the frequency spectrum complexity can accurately reflect the vibration characteristics of the vibration signals, the state of the target transformer winding can be accurately determined based on the characteristic matrix generated by the values.
In one embodiment, as shown in fig. 4, which illustrates a flowchart of a method for detecting a state of a winding of a transformer according to an embodiment of the present application, specifically, a possible process for determining a modal energy ratio of a vibration signal may include the following steps:
And 460, calculating the energy of the reconstruction mode, and determining the mode energy ratio of the vibration signal according to the energy of the reconstruction mode.
The acquired vibration signal can be decomposed by an empirical mode decomposition algorithm to obtain a mode decomposition result of the vibration signal, wherein the mode decomposition result includes a plurality of intrinsic mode components and a residual component, and the intrinsic mode components are also called as IMF components. The specific process of performing empirical mode decomposition on the vibration signal is as follows:
firstly, determining all maximum points and minimum points of a vibration signal x (t); fitting a maximum value point by adopting a cubic spline interpolation function to form an upper envelope line of the vibration signal, fitting a minimum value point to form a lower envelope line of the vibration signal, and calculating a mean value a of the upper envelope line and the lower envelope line 1 (t); calculating the difference d between the vibration signal and the average value 1 (t)＝x(t)-a 1 (t); judging whether the obtained difference value meets the IMF requirement, if so, determining the difference value d 1 (t) as the first IMF component, while the residual component is denoted as r 1 (t)＝x(t)-d 1 (t) if the IMF requirement is not satisfied, using the difference d 1 (t) replacing the vibration signal x (t), and repeating the previous steps until the difference value meets the IMF requirement, wherein the IMF requirement comprises two, namely, in the whole data segment, the number of the extreme points and the number of the zero-crossing points must be equal or the difference cannot exceed one at most, and the average value of an upper envelope line formed by the extreme points and a lower envelope line formed by the minimum points is zero at any time, namely, the upper envelope line and the lower envelope line are symmetrical relative to a time axis; continued decomposition residual classification r 1 (t), and repeating the previous steps until the final residual component satisfies a predetermined condition, which may be that the final residual component is a monotonic function, or that the difference between the maximum amplitude and the minimum amplitude thereof is less than a predetermined value, and finally decomposing the vibration signal into a sum of n IMF components and one residual component, which can be represented by formula (1):
wherein x (t) is a vibration signal; d i (t) is the IMF component; r is n (t) is the final residual component.
And reconstructing a modal decomposition result of the vibration signal to obtain a reconstruction mode, wherein the reconstruction mode can comprise a high-frequency mode, a medium-frequency mode and a low-frequency mode. Optionally, a zero-crossing rate method may be adopted to reconstruct a modal decomposition result of the vibration signal, as shown in fig. 5, which shows a flowchart of a transformer winding state detection method provided in an embodiment of the present application, and specifically relates to a possible process of determining a reconstruction modal, where the method may include the following steps:
and step 520, acquiring the zero crossing rate of each component in the modal decomposition result of the vibration signal.
The zero crossing rate is used for representing the ratio of the times of crossing the x axis by each component to the component length, and can be expressed by a formula (2):
wherein n is zero Is the zero-crossing rate; n is a radical of zero The number of times a component crosses the x-axis; n is the component length.
After the zero-crossing rate of each component is obtained through calculation, the sum of the components of which the zero-crossing rate is greater than a first zero-crossing rate threshold value in the zero-crossing rates of the components can be determined as a high-frequency mode; determining the sum of components with zero crossing rates smaller than or equal to a first zero crossing rate threshold value and larger than or equal to a second zero crossing rate threshold value as an intermediate frequency mode; and determining the sum of the components with the zero crossing rate smaller than a second zero crossing rate threshold value in the zero crossing rates of the components as a low-frequency mode. Alternatively, the first and second zero-crossing rate thresholds may be set manually according to experience, for example, the first zero-crossing rate threshold may be set to 0.01, and the second zero-crossing rate threshold may be set to 0.005. The modality obtained by the reconstruction can be expressed by formula (3):
wherein d is 1 、d 2 、d 3 Respectively a high-frequency mode, a medium-frequency mode and a low-frequency mode obtained after reconstruction; t, m, l represent the number of modes.
After the reconstruction mode is obtained, the energy of the reconstruction mode can be calculated through a formula (4), when the energy of the reconstruction mode is calculated, because the high-frequency mode, the medium-frequency mode and the low-frequency mode are all a time sequence, the energy of the reconstruction mode can be obtained by summing after values in the time sequence are squared, and then the modal energy proportion of the vibration signal is determined through a formula (5) according to the energy of the reconstruction mode.
Wherein E is 1 、E 2 、E 3 Energy of a high-frequency mode, a medium-frequency mode and a low-frequency mode respectively; q 1 、Q 2 、Q 3 The modal energy ratios of the high-frequency mode, the medium-frequency mode and the low-frequency mode are respectively.
In the embodiment, the mode decomposition result of the vibration signal is obtained by performing empirical mode decomposition on the vibration signal; reconstructing a modal decomposition result of the vibration signal by adopting a zero crossing rate method to obtain a reconstruction mode; and calculating the energy of the reconstruction mode, and determining the mode energy ratio of the vibration signal according to the energy of the reconstruction mode. The vibration signal can be effectively decomposed by adopting an empirical mode decomposition method, so that a plurality of components can be accurately obtained, and the accuracy of determining the modal energy ratio of the vibration signal is further improved.
In one embodiment, as shown in fig. 6, which shows a flowchart of a method for detecting a winding state of a transformer provided in an embodiment of the present application, specifically, a possible process for calculating a euclidean distance of a vibration signal may include the following steps:
and step 620, acquiring the reference modal energy ratio of the vibration signal of the normal transformer winding.
And step 640, calculating the Euclidean distance of the vibration signal according to the modal energy proportion of the vibration signal of the target transformer winding and the reference modal energy proportion.
The reference modal energy ratio of the vibration signal of the normal transformer winding can also be obtained by the process of calculating the modal energy ratio in the above embodiment, and details are not repeated here. And calculating the Euclidean distance of the vibration signal by adopting a formula (6) according to the modal energy ratio of the vibration signal of the target transformer winding and the reference modal energy ratio.
Wherein L is the Euclidean distance; q' is the reference modal energy fraction of the vibration signal of the normal transformer winding.
In the embodiment, the reference modal energy ratio of the vibration signal of the normal transformer winding is obtained; according to the modal energy ratio of the vibration signal of the target transformer winding and the reference modal energy ratio, the Euclidean distance of the vibration signal is calculated, the calculation mode is simple and easy to realize, and the efficiency of obtaining the characteristic parameter of the vibration signal is improved.
In one embodiment, as shown in fig. 7, which illustrates a flowchart of a method for detecting a winding state of a transformer according to an embodiment of the present application, specifically, a possible process for calculating a spectrum complexity of a vibration signal may include the following steps:
The frequency proportion of the vibration signal represents the proportion of the harmonic component of the vibration signal at the target frequency, the vibration power spectral density of the vibration signal is calculated firstly after the vibration signal is subjected to Fourier transform, and then the frequency proportion of the vibration signal is obtained according to the vibration amplitude of each frequency point of the vibration signal and the vibration power spectral density. And calculating the frequency spectrum complexity of the vibration signal through a formula (7) based on the frequency proportion of the vibration signal.
Wherein H is the spectral complexity; p is a radical of formula f Is the frequency proportion of the vibration signal; f is a target frequency and can be customized in 50-5000, for example, f can be 50, 100, 150, 200.
In the embodiment, the frequency proportion of the vibration signal is obtained; the frequency spectrum complexity of the vibration signal is calculated based on the frequency proportion of the vibration signal, the calculation mode is simple, and the frequency spectrum complexity of the vibration signal can be accurately calculated.
In one embodiment, as shown in fig. 8, which illustrates a flowchart of a transformer winding state detection method provided in an embodiment of the present application, specifically, related to a possible process of generating a preset machine learning model, the method may include the following steps:
And 840, inputting the training sample set into the initial machine learning model to obtain the prediction characteristic labels of the transformer windings.
And 860, updating the parameters of the initial machine learning model based on the predicted characteristic label and the target characteristic label until a preset convergence condition is reached, and generating a preset machine learning model based on the updated parameters.
The training sample set is historical data and can be stored in a winding mechanical state fingerprint library, the training sample set comprises a feature matrix corresponding to vibration signals of a plurality of transformer windings and target feature labels corresponding to the transformer windings, and the vibration signals of the plurality of transformer windings can comprise vibration signals of a normal transformer winding and vibration signals of a plurality of transformer windings with different deformation degrees. Inputting the feature matrix corresponding to the vibration signal and the target feature labels corresponding to the transformer windings into the initial machine learning model to obtain the predicted feature labels of the transformer windings, substituting the predicted feature labels and the target feature labels into a preset loss function to update parameters of the initial machine learning model until a preset convergence condition is reached, and generating the preset machine learning model based on the updated parameters. The preset convergence condition may be that a difference between the predicted feature tag and the target feature tag is smaller than a preset value, or that a preset iteration number is reached, or that another convergence condition is reached, which is not specifically limited in this embodiment.
In the embodiment, a training sample set is obtained; inputting the training sample set into an initial machine learning model to obtain a prediction characteristic label of each transformer winding; the parameters of the initial machine learning model are updated based on the predicted feature labels and the target feature labels until a preset convergence condition is reached, the preset machine learning model is generated based on the updated parameters, the parameters of the initial machine learning model are updated based on the predicted feature labels and the target feature labels, the accuracy of training the machine learning model is improved, and therefore the accuracy of calculating the preset machine learning model is skipped.
In one embodiment, as shown in fig. 9, which illustrates a flowchart of a method for detecting a winding state of a transformer provided in an embodiment of the present application, the method may include the following steps:
and 901, performing empirical mode decomposition on the vibration signal to obtain a mode decomposition result of the vibration signal.
And step 902, acquiring the zero crossing rate of each component in the modal decomposition result of the vibration signal.
And 904, calculating the energy of the reconstruction mode, and determining the mode energy ratio of the vibration signal according to the energy of the reconstruction mode.
And step 905, acquiring the reference modal energy ratio of the vibration signal of the normal transformer winding.
And 906, calculating the Euclidean distance of the vibration signal according to the modal energy ratio of the vibration signal of the target transformer winding and the reference modal energy ratio.
And step 907, acquiring the frequency proportion of the vibration signal.
And step 909, generating a characteristic matrix corresponding to the vibration signal of the target transformer winding based on the modal energy ratio, the Euclidean distance and the frequency spectrum complexity.
And step 910, inputting the feature matrix into a preset machine learning model to obtain a feature label of the target transformer winding.
And step 911, determining the state of the target transformer winding according to the feature tag.
The implementation principle and technical effect of each step in the method for detecting the state of the transformer winding provided by this embodiment are similar to those in the previous embodiments of the method for detecting the state of the transformer winding, and are not described again here. The implementation manner of each step in the embodiment of fig. 9 is only an example, and is not limited to this, and the order of each step may be adjusted in practical application as long as the purpose of each step can be achieved.
According to the technical scheme, the collected transformer winding vibration signals can be analyzed in real time, and the feature matrix corresponding to the vibration signals is calculated according to the machine learning model trained in advance, so that whether the transformer winding deforms or not is judged, and the flexibility of detecting the state of the transformer winding is improved.
It should be understood that although the various steps in the flow diagrams of fig. 2-9 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not limited to being performed in the exact order illustrated and, unless explicitly stated herein, may be performed in other orders. Moreover, at least some of the steps in fig. 2-9 may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of performing the steps or stages is not necessarily sequential, but may be performed alternately or alternatingly with other steps or at least some of the other steps or stages.
Referring to fig. 10, a block diagram of a transformer winding state detection apparatus 1000 according to an embodiment of the present application is shown. As shown in fig. 10, the transformer winding state detection apparatus 1000 may include: a feature extraction module 1002, a first input module 1004, and a determination module 1006, wherein:
the feature extraction module 1002 is configured to perform feature extraction according to the vibration signal of the target transformer winding to obtain a feature matrix corresponding to the vibration signal of the target transformer winding;
the first input module 1004 is used for inputting the feature matrix into a preset machine learning model to obtain a feature tag of a target transformer winding; different characteristic labels represent different deformation levels of the transformer winding;
a determining module 1006, configured to determine a state of the target transformer winding according to the feature tag.
In one embodiment, the feature extraction module 1002 includes an obtaining unit and a generating unit, where the obtaining unit is configured to obtain a modal energy ratio, an euclidean distance, and a spectrum complexity of a vibration signal according to the vibration signal of a target transformer winding; the generating unit is used for generating a characteristic matrix corresponding to the vibration signal of the target transformer winding based on the modal energy ratio, the Euclidean distance and the frequency spectrum complexity.
In an embodiment, the obtaining unit is specifically configured to perform empirical mode decomposition on the vibration signal to obtain a modal decomposition result of the vibration signal; the modal decomposition result comprises a plurality of intrinsic modal components and a residual component; reconstructing a modal decomposition result of the vibration signal to obtain a reconstruction modal; the reconstruction modes comprise a high-frequency mode, a medium-frequency mode and a low-frequency mode; and calculating the energy of the reconstruction mode, and determining the modal energy ratio of the vibration signal according to the energy of the reconstruction mode.
In one embodiment, the obtaining unit is further configured to obtain a zero crossing rate of each component in the modal decomposition result of the vibration signal; determining the sum of components with zero crossing rates larger than a first zero crossing rate threshold value in the zero crossing rates of the components as a high-frequency mode; determining the sum of components of which the zero crossing rate is less than or equal to a first zero crossing rate threshold and is greater than or equal to a second zero crossing rate threshold as an intermediate frequency mode; and determining the sum of the components with the zero crossing rate smaller than a second zero crossing rate threshold value in the zero crossing rates of the components as a low-frequency mode.
In one embodiment, the obtaining unit is further configured to obtain a reference modal energy ratio of the vibration signal of the normal transformer winding; and calculating the Euclidean distance of the vibration signal according to the modal energy ratio of the vibration signal of the target transformer winding and the reference modal energy ratio.
In one embodiment, the obtaining unit is further configured to obtain a frequency proportion of the vibration signal; the frequency proportion is used for representing the proportion of harmonic components of the vibration signal at the target frequency; calculating the frequency spectrum complexity of the vibration signal based on the frequency proportion of the vibration signal; the spectral complexity is used to characterize the complexity of the target frequency.
In one embodiment, the transformer winding state detection apparatus 1000 further includes an obtaining module 1008, a second input module 1010, and a generating module 1012, wherein:
an obtaining module 1008, configured to obtain a training sample set; the training sample set comprises a feature matrix corresponding to vibration signals of a plurality of transformer windings and a target feature label corresponding to each transformer winding;
the second input module 1010 is configured to input the training sample set into the initial machine learning model to obtain a predicted feature label of each transformer winding;
a generating module 1012, configured to update parameters of the initial machine learning model based on the predicted feature labels and the target feature labels until a preset convergence condition is reached, and generate a preset machine learning model based on the updated parameters.
For specific limitations of the transformer winding state detection device, reference may be made to the above limitations of the transformer winding state detection method, and details thereof are not repeated here. The modules in the transformer winding state detection device can be wholly or partially realized by software, hardware and a combination thereof. The modules can be embedded in hardware or independent of a processor in the computer device, or can be stored in a memory in the computer device in software, so that the processor can call and execute operations of the modules.
In one embodiment of the present application, there is provided a computer device comprising a memory and a processor, the memory having stored therein a computer program, the processor implementing the following steps when executing the computer program:
performing characteristic extraction according to the vibration signal of the target transformer winding to obtain a characteristic matrix corresponding to the vibration signal of the target transformer winding; inputting the characteristic matrix into a preset machine learning model to obtain a characteristic label of the target transformer winding; different characteristic labels represent different deformation levels of the transformer winding; and determining the state of the target transformer winding according to the characteristic label.
In one embodiment of the application, the processor when executing the computer program further performs the steps of:
respectively acquiring the modal energy ratio, euclidean distance and frequency spectrum complexity of the vibration signal according to the vibration signal of the target transformer winding; and generating a characteristic matrix corresponding to the vibration signal of the target transformer winding based on the modal energy ratio, the Euclidean distance and the frequency spectrum complexity.
In one embodiment of the application, the processor when executing the computer program further performs the steps of:
carrying out empirical mode decomposition on the vibration signal to obtain a mode decomposition result of the vibration signal; the modal decomposition result comprises a plurality of intrinsic modal components and a residual component; reconstructing a modal decomposition result of the vibration signal to obtain a reconstruction modal; the reconstruction modes comprise a high-frequency mode, a medium-frequency mode and a low-frequency mode; and calculating the energy of the reconstruction mode, and determining the modal energy ratio of the vibration signal according to the energy of the reconstruction mode.
In one embodiment of the application, the processor when executing the computer program further performs the steps of:
acquiring the zero crossing rate of each component in the modal decomposition result of the vibration signal; determining the sum of components with zero crossing rates larger than a first zero crossing rate threshold value in the zero crossing rates of the components as a high-frequency mode; determining the sum of components of which the zero crossing rate is less than or equal to a first zero crossing rate threshold and is greater than or equal to a second zero crossing rate threshold as an intermediate frequency mode; and determining the sum of the components with the zero crossing rate smaller than a second zero crossing rate threshold value in the zero crossing rates of the components as a low-frequency mode.
In one embodiment of the application, the processor when executing the computer program further performs the steps of:
acquiring the reference modal energy ratio of the vibration signal of the normal transformer winding; and calculating the Euclidean distance of the vibration signal according to the modal energy ratio of the vibration signal of the target transformer winding and the reference modal energy ratio.
In one embodiment of the application, the processor when executing the computer program further performs the steps of:
acquiring the frequency proportion of the vibration signal; the frequency proportion is used for representing the proportion of harmonic components of the vibration signal at the target frequency; calculating the frequency spectrum complexity of the vibration signal based on the frequency proportion of the vibration signal; the spectral complexity is used to characterize the complexity of the target frequency.
In one embodiment of the application, the processor when executing the computer program further performs the steps of:
acquiring a training sample set; the training sample set comprises a feature matrix corresponding to vibration signals of a plurality of transformer windings and a target feature label corresponding to each transformer winding; inputting the training sample set into an initial machine learning model to obtain a prediction characteristic label of each transformer winding; and updating the parameters of the initial machine learning model based on the predicted characteristic label and the target characteristic label until a preset convergence condition is reached, and generating a preset machine learning model based on the updated parameters.
The implementation principle and technical effect of the computer device provided in the embodiment of the present application are similar to those of the method embodiment described above, and are not described herein again.
In an embodiment of the application, a computer-readable storage medium is provided, on which a computer program is stored, which computer program, when being executed by a processor, carries out the steps of:
performing characteristic extraction according to the vibration signal of the target transformer winding to obtain a characteristic matrix corresponding to the vibration signal of the target transformer winding; inputting the characteristic matrix into a preset machine learning model to obtain a characteristic label of the target transformer winding; different feature labels represent different deformation levels of the transformer winding; and determining the state of the target transformer winding according to the characteristic label.
In one embodiment of the application, the computer program when executed by the processor further performs the steps of:
respectively acquiring modal energy ratio, euclidean distance and frequency spectrum complexity of the vibration signal according to the vibration signal of the target transformer winding; and generating a characteristic matrix corresponding to the vibration signal of the target transformer winding based on the modal energy ratio, the Euclidean distance and the frequency spectrum complexity.
In one embodiment of the application, the computer program when executed by the processor further performs the steps of:
performing empirical mode decomposition on the vibration signal to obtain a mode decomposition result of the vibration signal; the modal decomposition result comprises a plurality of intrinsic modal components and a residual component; reconstructing a modal decomposition result of the vibration signal to obtain a reconstruction modal; the reconstruction modes comprise a high-frequency mode, a medium-frequency mode and a low-frequency mode; and calculating the energy of the reconstruction mode, and determining the mode energy ratio of the vibration signal according to the energy of the reconstruction mode.
In one embodiment of the application, the computer program when executed by the processor further performs the steps of:
acquiring the zero crossing rate of each component in the modal decomposition result of the vibration signal; determining the sum of components with zero crossing rates larger than a first zero crossing rate threshold value in the zero crossing rates of the components as a high-frequency mode; determining the sum of components of which the zero crossing rate is less than or equal to a first zero crossing rate threshold and is greater than or equal to a second zero crossing rate threshold as an intermediate frequency mode; and determining the sum of the components with the zero crossing rate smaller than a second zero crossing rate threshold value in the zero crossing rates of the components as a low-frequency mode.
In one embodiment of the application, the computer program when executed by the processor further performs the steps of:
acquiring the reference modal energy ratio of a vibration signal of a normal transformer winding; and calculating the Euclidean distance of the vibration signal according to the modal energy ratio of the vibration signal of the target transformer winding and the reference modal energy ratio.
In one embodiment of the application, the computer program when executed by the processor further performs the steps of:
acquiring the frequency proportion of the vibration signal; the frequency proportion is used for representing the proportion of harmonic components of the vibration signal at the target frequency; calculating the frequency spectrum complexity of the vibration signal based on the frequency proportion of the vibration signal; the spectral complexity is used to characterize the complexity of the target frequency.
In one embodiment of the application, the computer program when executed by the processor further performs the steps of:
acquiring a training sample set; the training sample set comprises a feature matrix corresponding to vibration signals of a plurality of transformer windings and a target feature label corresponding to each transformer winding; inputting the training sample set into an initial machine learning model to obtain a prediction characteristic label of each transformer winding; and updating the parameters of the initial machine learning model based on the predicted characteristic label and the target characteristic label until a preset convergence condition is reached, and generating a preset machine learning model based on the updated parameters.
The implementation principle and technical effect of the computer-readable storage medium provided by this embodiment are similar to those of the above-described method embodiment, and are not described herein again.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), rambus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above examples only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the claims. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. A method of detecting a state of a winding of a transformer, the method comprising:
respectively obtaining the modal energy ratio, the Euclidean distance and the frequency spectrum complexity of a vibration signal according to the vibration signal of a target transformer winding, and generating a characteristic matrix corresponding to the vibration signal of the target transformer winding based on the modal energy ratio, the Euclidean distance and the frequency spectrum complexity, wherein the modal energy ratio of the vibration signal is used for representing the ratio of modal energy of different frequencies to total modal energy, the Euclidean distance is used for representing the distance between the modal energy ratio of the vibration signal of the target transformer winding and a reference modal energy ratio, and the reference modal energy ratio is the modal energy ratio of the vibration signal of a normal transformer winding;
inputting the characteristic matrix into a preset machine learning model to obtain a characteristic label of the target transformer winding; different characteristic labels represent different deformation levels of the transformer winding;
and determining the state of the target transformer winding according to the characteristic label.
2. The method of claim 1, wherein obtaining the modal energy ratio of the vibration signal comprises:
carrying out empirical mode decomposition on the vibration signal to obtain a mode decomposition result of the vibration signal; the modal decomposition result comprises a plurality of intrinsic modal components and a residual component;
reconstructing a modal decomposition result of the vibration signal to obtain a reconstruction modal; the reconstruction modes comprise a high-frequency mode, a medium-frequency mode and a low-frequency mode;
and calculating the energy of the reconstruction mode, and determining the mode energy ratio of the vibration signal according to the energy of the reconstruction mode.
3. The method according to claim 2, wherein the reconstructing a mode corresponding to the vibration signal to obtain a reconstructed mode corresponding to the mode comprises:
acquiring the zero crossing rate of each component in the modal decomposition result of the vibration signal;
determining the sum of components with zero crossing rates larger than a first zero crossing rate threshold value in the zero crossing rates of the components as the high-frequency mode;
determining the sum of components with zero crossing rates smaller than or equal to the first zero crossing rate threshold and larger than or equal to the second zero crossing rate threshold as the intermediate frequency mode;
and determining the sum of components with zero crossing rates smaller than the second zero crossing rate threshold value in the zero crossing rates of the components as the low-frequency mode.
4. The method according to any one of claims 1-3, wherein obtaining the Euclidean distance of the vibration signal comprises:
acquiring the reference modal energy ratio of the vibration signal of the normal transformer winding;
and calculating the Euclidean distance of the vibration signal according to the modal energy ratio of the vibration signal of the target transformer winding and the reference modal energy ratio.
5. The method according to any one of claims 1-3, wherein obtaining the spectral complexity of the vibration signal comprises:
acquiring the frequency proportion of the vibration signal; the frequency proportion is used for representing the proportion of harmonic components of the vibration signal at a target frequency;
calculating the frequency spectrum complexity of the vibration signal based on the frequency proportion of the vibration signal; the spectral complexity is used to characterize the complexity of the target frequency.
6. The method according to any one of claims 1-3, further comprising:
acquiring a training sample set; the training sample set comprises a feature matrix corresponding to vibration signals of a plurality of transformer windings and a target feature label corresponding to each transformer winding;
inputting the training sample set into an initial machine learning model to obtain a prediction characteristic label of each transformer winding;
and updating the parameters of the initial machine learning model based on the predicted characteristic label and the target characteristic label until a preset convergence condition is reached, and generating the preset machine learning model based on the updated parameters.
7. A transformer winding condition detection apparatus, the apparatus comprising:
the characteristic extraction module is used for respectively acquiring the modal energy duty ratio, the Euclidean distance and the frequency spectrum complexity of a vibration signal of a target transformer winding according to the vibration signal of the target transformer winding, and generating a characteristic matrix corresponding to the vibration signal of the target transformer winding based on the modal energy duty ratio, the Euclidean distance and the frequency spectrum complexity, wherein the modal energy duty ratio of the vibration signal is used for representing the ratio of modal energy of different frequencies to total modal energy, the Euclidean distance is used for representing the distance between the modal energy duty ratio of the vibration signal of the target transformer winding and a reference modal energy duty ratio, and the reference modal energy duty ratio is the modal energy duty ratio of the vibration signal of a normal transformer winding;
the first input module is used for inputting the characteristic matrix into a preset machine learning model to obtain a characteristic label of the target transformer winding; different characteristic labels represent different deformation levels of the transformer winding;
and the determining module is used for determining the state of the target transformer winding according to the characteristic label.
8. The apparatus according to claim 7, wherein the feature extraction module, when obtaining the modal energy ratio of the vibration signal, is specifically configured to:
carrying out empirical mode decomposition on the vibration signal to obtain a mode decomposition result of the vibration signal; the modal decomposition result comprises a plurality of intrinsic modal components and a residual component;
reconstructing a modal decomposition result of the vibration signal to obtain a reconstruction mode; the reconstruction modes comprise a high-frequency mode, a medium-frequency mode and a low-frequency mode;
and calculating the energy of the reconstruction mode, and determining the mode energy ratio of the vibration signal according to the energy of the reconstruction mode.
9. A computer arrangement comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, carries out the steps of the method according to any one of claims 1 to 6.
10. A computer-readable storage medium, characterized in that a computer program is stored thereon, which computer program, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110898223.4A CN113701684B (en) | 2021-08-05 | 2021-08-05 | Transformer winding state detection method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110898223.4A CN113701684B (en) | 2021-08-05 | 2021-08-05 | Transformer winding state detection method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113701684A CN113701684A (en) | 2021-11-26 |
| CN113701684B true CN113701684B (en) | 2022-10-14 |
Family
ID=78651660
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110898223.4A Active CN113701684B (en) | 2021-08-05 | 2021-08-05 | Transformer winding state detection method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113701684B (en) |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105547465A (en) * | 2015-12-08 | 2016-05-04 | 华北电力大学（保定） | Transformer vibration signal winding state feature extraction method |
| CN107132451A (en) * | 2017-05-31 | 2017-09-05 | 广州供电局有限公司 | The winding state detection method and system of transformer |
| CN108169583B (en) * | 2017-11-17 | 2020-11-20 | 国网湖南省电力有限公司 | Autotransformer direct-current magnetic bias discrimination method and system with neutral point grounded through capacitor |
| CN108693437B (en) * | 2018-03-22 | 2020-12-25 | 国网湖南省电力有限公司 | Method and system for judging deformation of transformer winding |
| CN110458248B (en) * | 2019-09-18 | 2022-08-30 | 华北电力大学（保定） | Transformer abnormal state detection method based on multi-point vibration signals |
| CN111273100B (en) * | 2020-02-20 | 2021-04-27 | 浙江大学 | Power transformer winding state evaluation method based on vibration phase |
-
2021
- 2021-08-05 CN CN202110898223.4A patent/CN113701684B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN113701684A (en) | 2021-11-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3515037A1 (en) | Dynamic concurrent learning method to neutralize cyber attacks and faults for industrial asset monitoring nodes | |
| CN110926782B (en) | Circuit breaker fault type judgment method and device, electronic equipment and storage medium | |
| JP6971514B1 (en) | Information processing equipment, information processing methods and programs | |
| CN110826803A (en) | Electricity price prediction method and device for electric power spot market | |
| Wang et al. | DAMER: A novel diagnosis aggregation method with evidential reasoning rule for bearing fault diagnosis | |
| CN113516340A (en) | Intelligent work order pushing method and device | |
| CN113284001A (en) | Power consumption prediction method and device, computer equipment and storage medium | |
| CN110570487B (en) | Undersampling model generation method, image reconstruction method, device and computer equipment | |
| CN110119621B (en) | Attack defense method, system and defense device for abnormal system call | |
| Lee et al. | A feature selection approach hybrid grey wolf and heap-based optimizer applied in bearing fault diagnosis | |
| CN113758652B (en) | Oil leakage detection method and device for converter transformer, computer equipment and storage medium | |
| CN113701684B (en) | Transformer winding state detection method, device, equipment and storage medium | |
| CN113850305A (en) | Generator fault diagnosis and health state assessment method and electronic equipment | |
| CN116106672B (en) | Vehicle network resonance detection method and device based on data driving and engineering knowledge | |
| Fu et al. | PQEventCog: Classification of power quality disturbances based on optimized S-transform and CNNs with noisy labeled datasets | |
| US11727109B2 (en) | Identifying adversarial attacks with advanced subset scanning | |
| CN116975742A (en) | Partial discharge pattern recognition method, apparatus, device, and storage medium | |
| US11748632B2 (en) | Analysis of anomalies in a facility | |
| Zheng et al. | Short-term load forecasting based on Gaussian wavelet SVM | |
| Aggarwal et al. | Recognition of voltage sag causes using vector quantization based orthogonal wavelet | |
| CN117953263A (en) | Sample image generation method, model processing method, device and related equipment | |
| CN115510925B (en) | Mechanical fault diagnosis method, device and medium | |
| Gandhi et al. | Nuisance attribute projection for system independent fault diagnosis of synchronous generator | |
| CN114139766A (en) | Photovoltaic group string current prediction method and system based on CNN and LSTM methods | |
| Santos et al. | Two-stage binary image operator design: An approach based on interaction information |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |