CN112215394B

CN112215394B - Converter transformer vibration signal prediction method, device, equipment and storage medium

Info

Publication number: CN112215394B
Application number: CN202010904359.7A
Authority: CN
Inventors: 汲胜昌; 吴书煜; 祝令瑜; 代双寅; 党永亮; 张壮壮
Original assignee: Xian Jiaotong University; Electric Power Research Institute of State Grid Henan Electric Power Co Ltd
Current assignee: Xian Jiaotong University; Electric Power Research Institute of State Grid Henan Electric Power Co Ltd
Priority date: 2020-09-01
Filing date: 2020-09-01
Publication date: 2023-08-15
Anticipated expiration: 2040-09-01
Also published as: CN112215394A

Abstract

The application discloses a method, a device and equipment for predicting a vibration signal of a converter transformer and a storage medium, and belongs to the technical field of converter transformers. The method comprises the following steps: performing multiple decomposition on the current vibration signal of the converter transformer by adopting at least two modal decomposition methods to obtain a plurality of target modal components corresponding to the current vibration signal; predicting each target modal component by adopting a plurality of prediction methods to obtain a prediction result corresponding to each prediction method; the predicting method is used for predicting the peak value of the vibration signal at the next moment; according to a preset weight distribution rule, distributing weights for all prediction results; and carrying out weighted summation on each prediction result and the weight corresponding to each prediction result to obtain a comprehensive prediction result. The technical scheme provided by the embodiment of the application can improve the prediction precision of the vibration signal.

Description

Converter transformer vibration signal prediction method, device, equipment and storage medium

Technical Field

The application relates to the technical field of converter transformers, in particular to a method, a device and equipment for predicting vibration signals of a converter transformer and a storage medium.

Background

The converter transformer is one of core equipment in the extra-high voltage direct current transmission project, and the running state of the converter transformer directly influences the running of the whole system. According to the actual statistics, the fault of the converter transformer is mainly caused by the iron core defect and winding deformation, and the vibration generated by the iron core defect and winding deformation can cause the abnormal vibration of the converter transformer, so that the fault of the converter transformer can be diagnosed through the vibration characteristics of the converter transformer.

At present, a vibration signal analysis method is mainly adopted for fault diagnosis of the converter transformer, wherein the vibration signal analysis method is to take a vibration signal of the surface of an oil tank for monitoring the converter transformer as a monitoring basis of the states of an iron core and a winding of the converter transformer, and then compare the vibration signal after the fault occurrence with a normal vibration signal to finally obtain a fault diagnosis result. However, the fault diagnosis method belongs to the diagnosis after the fault occurs, and the safe operation of the converter transformer cannot be ensured.

Based on this, researchers have proposed a method of predicting vibration signals of a converter transformer, so that the operation state of the converter transformer can be monitored in real time. At present, a comprehensive prediction method is mainly adopted when the vibration signal is predicted, and the comprehensive prediction method is to obtain a comprehensive prediction result according to the prediction result obtained by each prediction method after the vibration signal is predicted by adopting a plurality of prediction methods.

However, the accuracy of the result predicted by the above-described comprehensive prediction method is low due to the complicated variation of the vibration signal.

Disclosure of Invention

Based on the above, the embodiment of the application provides a method, a device, equipment and a storage medium for predicting a vibration signal of a converter transformer, which can improve the prediction precision of the vibration signal.

In a first aspect, a method for predicting a converter transformer vibration signal is provided, the method comprising:

performing multiple decomposition on the current vibration signal of the converter transformer by adopting at least two modal decomposition methods to obtain a plurality of target modal components corresponding to the current vibration signal; predicting each target modal component by adopting a plurality of prediction methods to obtain a prediction result corresponding to each prediction method; the predicting method is used for predicting the peak value of the vibration signal at the next moment; according to a preset weight distribution rule, distributing weights for all prediction results; and carrying out weighted summation on each prediction result and the weight corresponding to each prediction result to obtain a comprehensive prediction result.

In one embodiment, at least two modal decomposition methods are adopted to perform multiple decomposition on a current vibration signal of the converter transformer respectively, so as to obtain a plurality of target modal components corresponding to the current vibration signal, including:

Multiple decomposition is carried out on the current vibration signal by adopting an empirical wavelet transformation decomposition method, so as to obtain a plurality of first target modal components corresponding to the current vibration signal; and performing multiple decomposition on the current vibration signal by adopting a variation modal decomposition method to obtain a plurality of second target modal components corresponding to the current vibration signal.

In one embodiment, performing multiple decomposition on the current vibration signal by using an empirical wavelet transform decomposition method to obtain a plurality of first target modal components corresponding to the current vibration signal, where the multiple decomposition includes:

multiple decomposition is carried out on the current vibration signal by adopting an empirical wavelet transformation decomposition method, so as to obtain a plurality of first initial modal components; determining the number of first target modal components according to a preset maximum value threshold and a first correlation coefficient threshold; a plurality of first target modal components is determined from the plurality of first initial modal components according to the number of first target modal components.

In one embodiment, determining the number of first target modal components according to the preset maximum threshold and the first correlation coefficient threshold includes:

counting the number of first initial modal components with local maxima larger than a maximum threshold as a first number; counting the number of first initial modal components with the first correlation coefficient larger than a first correlation coefficient threshold as a second number; the first correlation coefficient represents the degree of correlation between the first initial modal component and the corresponding vibration peak value; the minimum of the first number and the second number is determined as the number of first target modal components.

In one embodiment, multiple decomposition is performed on the current vibration signal by using a variational mode decomposition method to obtain a plurality of target second mode components corresponding to the current vibration signal, including:

multiple decomposition is carried out on the current vibration signal by adopting a variable-decomposition mode decomposition method, so as to obtain a plurality of second initial mode components; determining the number of second target modal components according to a preset center frequency threshold and a second correlation number threshold; a plurality of second target modal components is determined from the plurality of second initial modal components according to the number of second target modal components.

In one embodiment, determining the number of second target modal components according to the preset center frequency threshold and the second phase relation threshold includes:

calculating the difference ratio of the center frequencies of the adjacent second initial modal components, and counting the number of the second initial modal components with the difference ratio of the center frequencies larger than the center frequency threshold as a third number; counting the number of second initial modal components with the second phase relation number larger than a second phase relation number threshold value as a fourth number; the second correlation coefficient represents the degree of correlation between the second initial modal component and the corresponding vibration peak; the minimum of the third number and the fourth number is determined as the number of second target modal components.

In one embodiment, predicting each modal component by using a plurality of prediction methods to obtain a prediction result corresponding to each prediction method includes:

for each prediction method, predicting each first target modal component by adopting the prediction method to obtain a component prediction result of each first target modal component; adding the component prediction results of all the first target modal components to obtain a first prediction result corresponding to the prediction method; for each prediction method, predicting each second target modal component by adopting the prediction method to obtain a component prediction result of each second target modal component; and adding the component prediction results of all the second target modal components to obtain a second prediction result corresponding to the prediction method.

In one embodiment, according to a preset weight distribution rule, distributing weights to each prediction result includes:

generating a predictive evaluation index matrix according to a predictive evaluation standard, wherein the predictive evaluation index matrix comprises an average absolute error, a mean square error, a root mean square error, an average absolute percentage error and a symmetrical average absolute percentage error, and the predictive evaluation index matrix is a matrix formed by the average absolute error, the mean square error, the root mean square error, the average absolute percentage error and the symmetrical average absolute percentage error; and (3) adopting a fuzzy analytic hierarchy process and an entropy weight process, and distributing weights to all prediction results according to the prediction evaluation index matrix.

In a second aspect, there is provided a predicting apparatus for a vibration signal of a converter transformer, the predicting apparatus comprising:

the multiple decomposition module is used for performing multiple decomposition on the current vibration signal of the converter transformer by adopting at least two modal decomposition methods respectively to obtain a plurality of target modal components corresponding to the current vibration signal;

the prediction module is used for predicting each modal component by adopting a plurality of prediction methods to obtain a prediction result corresponding to each prediction method; the predicting method is used for predicting the peak value of the vibration signal at the next moment;

the distribution module is used for distributing weights for all prediction results according to preset weight distribution rules;

and the weighted summation module is used for weighted summation of each prediction result and the weight corresponding to each prediction result to obtain a comprehensive prediction result.

In a third aspect, a computer device is provided, comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, implements a method of predicting a converter transformer vibration signal as described in any one of the first aspects above.

In a fourth aspect, a computer readable storage medium is provided, on which a computer program is stored which, when executed by a processor, implements a method for predicting a converter transformer vibration signal according to any one of the first aspects above.

The technical scheme provided by the embodiment of the application has the beneficial effects that at least:

performing multiple decomposition on the current vibration signal of the converter transformer by adopting at least two modal decomposition methods to obtain a plurality of target modal components corresponding to the current vibration signal; predicting each target modal component by adopting a plurality of prediction methods to obtain a prediction result corresponding to each prediction method; the predicting method is used for predicting the peak value of the vibration signal at the next moment; according to a preset weight distribution rule, distributing weights for all prediction results; and carrying out weighted summation on each prediction result and the weight corresponding to each prediction result to obtain a comprehensive prediction result. In the technical scheme provided by the embodiment of the application, the current vibration signal with the complex characteristic can be decomposed into a plurality of simple target modal components which are convenient to predict due to the fact that at least two decomposition methods are adopted to carry out multiple decomposition on the current vibration signal, so that the prediction precision of the current vibration signal is improved. And each target modal component is respectively predicted by adopting a plurality of prediction methods, and the prediction advantages of different prediction methods can be combined for prediction, so that the prediction precision is improved. And the weight is distributed to each prediction result, so that the prediction advantages of different prediction methods can be further fully exerted, and the prediction precision of the current vibration signal is further improved.

Drawings

FIG. 1 is a block diagram of a server according to an embodiment of the present application;

FIG. 2 is a flowchart of a method for predicting vibration signals according to an embodiment of the present application;

FIG. 3 is a flow chart of multiple decomposition of a current vibration signal according to an embodiment of the present application;

FIG. 4 is a flow chart of multiple decomposition of a current vibration signal according to an embodiment of the present application;

FIG. 5 is a flow chart of multiple decomposition of a current vibration signal according to an embodiment of the present application;

FIG. 6 is a flowchart illustrating a method for determining a number of first target modal components according to an embodiment of the application;

FIG. 7 is a schematic diagram of seven first target modal components according to an embodiment of the present application;

FIG. 8 is a flowchart illustrating a method for determining the number of second target modal components according to an embodiment of the application;

FIG. 9 is a schematic diagram of seven second target modal components according to an embodiment of the present application;

FIG. 10 is a flowchart of predicting each first target modal component and each second target modal component according to an embodiment of the application;

FIG. 11 is a schematic diagram of prediction results corresponding to a plurality of prediction methods according to an embodiment of the present application;

Fig. 12 is a flowchart of a weight distribution method according to an embodiment of the present application;

FIG. 13 is a diagram illustrating a method for predicting vibration signals according to an embodiment of the present application;

fig. 14 is a block diagram of a vibration signal prediction apparatus according to an embodiment of the present application;

FIG. 15 is a block diagram of a vibration signal prediction apparatus according to an embodiment of the present application;

FIG. 16 is a block diagram of a vibration signal prediction apparatus according to an embodiment of the present application;

fig. 17 is a block diagram of a vibration signal prediction apparatus according to an embodiment of the present application.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the present application more apparent, the embodiments of the present application will be described in further detail with reference to the accompanying drawings.

The method for predicting the vibration signal of the converter transformer provided by the embodiment of the application can be applied to a server, wherein the server can be one server or a server cluster consisting of a plurality of servers, and the embodiment of the application is not particularly limited.

Referring to fig. 1, a block diagram of a server according to an embodiment of the present application is shown, where, as shown in fig. 1, the server may include a processor and a memory connected through a system bus. Wherein the processor of the server is configured to provide computing and control capabilities. The memory of the server includes nonvolatile storage medium and internal memory. The nonvolatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The computer program is executed by a processor to implement a method of predicting a converter transformer vibration signal.

It will be appreciated by those skilled in the art that the structure shown in fig. 1 is merely a block diagram of a portion of the structure associated with the present inventive arrangements and does not constitute a limitation of the servers to which the present inventive arrangements are applied, alternatively the servers may include more or less components than those shown, or may combine certain components, or have different arrangements of components.

Referring to fig. 2, a flowchart of a method for predicting a vibration signal of a converter transformer according to an embodiment of the present application is shown, where the method for predicting a vibration signal may be applied to a server. As shown in fig. 2, the method for predicting the vibration signal may include the steps of:

and 201, performing multiple decomposition on the current vibration signal of the converter transformer by adopting at least two modal decomposition methods to obtain a plurality of target modal components corresponding to the current vibration signal.

The current vibration signal is a vibration signal acquired by detecting the state of the converter transformer, and can be expressed as a time-sequential parameter sequence, wherein the parameter can be a vibration peak value of the current vibration signal, and the vibration peak value represents acceleration. The decomposition method for performing multiple decomposition on the current vibration signal may be an empirical wavelet transform (Empirical Wavelet Transform, EWT) decomposition method and a variational modal decomposition method (Variational Mode Decomposition, VMD), or may be other multiple decomposition methods, which are not limited in the embodiment of the present application.

The plurality of target modal components are target modal components corresponding to the current vibration signal obtained by carrying out multiple decomposition on the current vibration signal. And decomposing the current vibration signal by adopting a plurality of methods to obtain a plurality of modal components obtained by decomposing the methods. For example, the current vibration signal is decomposed by two methods, 10 modal components are obtained after the decomposition by the first method, 8 modal components are obtained after the decomposition by the second method, and the modal components can be directly used as target modal components or part of the modal components can be screened out from the modal components to be used as target modal components.

And 202, respectively predicting each target modal component by adopting a plurality of prediction methods to obtain a prediction result corresponding to each prediction method.

The prediction method is used for predicting the peak value of the vibration signal at the next moment. And predicting each target modal component by adopting a plurality of prediction methods, wherein the target modal component is decomposed according to the current vibration signal, so that the prediction result corresponding to each prediction method can be finally obtained according to the prediction result of each target modal component after each target modal component is predicted. For example, there are three prediction methods, namely, an autoregressive moving average method (Autoregressive Moving Average, ARIMA), a BP neural network method (BP Neural Network, BP) and a least squares support vector machine (Least Squares Support Vector Machine, LSSVM), which may be used to predict each target modal component separately to obtain a prediction result, or may be used to predict each target modal component after any combination of the three methods to obtain a prediction result.

Step 203, according to a preset weight distribution rule, a weight is distributed to each prediction result.

The preset weight distribution rule is a distribution rule when the weight is distributed to each prediction result obtained by the method, and the weight distribution rule can be that each prediction result corresponds to one weight and can be that each prediction result corresponding to each prediction method is distributed with one weight; alternatively, the same prediction method may have different weights for the prediction results for each modal component.

And 204, carrying out weighted summation on each prediction result and the weight corresponding to each prediction result to obtain a comprehensive prediction result.

The final comprehensive prediction result is obtained by weighting and summing each prediction result and the weight allocated to the prediction result, namely, after multiplying each prediction result by the weight corresponding to each prediction result, all multiplied results are added to obtain the comprehensive prediction result of the current vibration signal. For example, three different prediction methods may be adopted to predict each modal component to obtain a prediction result X corresponding to each prediction method ₁₁ 、X ₁₂ 、X ₁₃ And the weight eta of each prediction result ₁₁ 、η ₁₂ 、η ₁₃ The comprehensive prediction result of the current vibration signal can be obtained according to the formula (1).

Wherein eta ₁₁ Is the prediction result X ₁₁ Corresponding weight, eta ₁₂ Is the prediction result X ₁₂ Corresponding weight, eta ₁₃ Is the prediction result X ₁₃ And (5) corresponding weight. η (eta) ₂₁ Is the prediction result X ₂₁ Corresponding weight, eta ₂₂ Is the prediction result X ₂₂ Corresponding weight, eta ₂₃ Is the prediction result X ₂₃ And (5) corresponding weight.

Performing multiple decomposition on the current vibration signal of the converter transformer by adopting at least two modal decomposition methods to obtain a plurality of target modal components corresponding to the current vibration signal; predicting each target modal component by adopting a plurality of prediction methods to obtain a prediction result corresponding to each prediction method; according to a preset weight distribution rule, distributing weights for all prediction results; and carrying out weighted summation on each prediction result and the weight corresponding to each prediction result to obtain a comprehensive prediction result. In the technical scheme provided by the embodiment of the application, the current vibration signal is subjected to multiple decomposition, so that the current vibration signal with the complex characteristic of variation is decomposed into a plurality of simple target modal components which are convenient to predict, and the prediction precision of the current vibration signal can be improved. And each target modal component is respectively predicted by adopting a plurality of prediction methods, so that the prediction advantages of different prediction methods can be combined, and the prediction precision can be improved. Finally, weights are distributed to all the prediction results, so that the prediction advantages of different prediction methods can be further fully exerted, and the prediction precision of the current vibration signal is further improved.

In one embodiment, the application provides a method for respectively carrying out multiple decomposition on the current vibration signal of the converter transformer by adopting two modal decomposition methods to obtain a plurality of target modal components corresponding to the current vibration signal. For example, the current vibration signal is subjected to multiple decomposition by respectively adopting an empirical wavelet transformation decomposition method and a variation modal decomposition method, so as to obtain a plurality of target modal components corresponding to the current vibration signal.

Referring to fig. 3, a flow chart of multiple decomposition of a current vibration signal according to an embodiment of the present application is shown, and the method can be applied to a server. As shown in fig. 3, the multiple decomposition method may include the steps of:

and 301, performing multiple decomposition on the current vibration signal by using an empirical wavelet transformation decomposition method to obtain a plurality of first target modal components corresponding to the current vibration signal.

In this embodiment, an empirical wavelet transform decomposition method may be used to perform multiple decomposition on the current vibration signal, and multiple initial modal components obtained by decomposition are directly used as first target modal components; alternatively, a mode component of a portion selected from the plurality of initial mode components obtained by decomposition may be used as the first target mode component.

As shown in fig. 4, a specific implementation of step 301 may include the following steps:

and 401, performing multiple decomposition on the current vibration signal by using an empirical wavelet transformation decomposition method to obtain a plurality of first initial modal components.

Wherein the first initial modal component f ₀ (t) and f _i The (t) can be calculated by a formula (2) -a formula (9), and the calculation process is specifically as follows: first, the frequency range [0, pi ]]Is divided into N consecutive areas S _i Forming n+1 boundaries ω, wherein ω ₀ =0 and ω _N Pi is a known boundary, and the remaining N-1 boundaries are determined by the number M of local maxima in descending order in the spectrum. If M>N, the first N-1 maximum values are reserved; if M<N, then all maxima are retained and N is adjusted to the appropriate value, optionally, n=m. Finally, taking the central frequency omega of each two continuous local maxima _i As a boundary of two regions;

second, for the N regions S determined above _i According to Littlewood-Paley and Meyer wavelet construction ideas, an empirical scale function within each region is built according to formulas (2) to (5)And empirical wavelet->

Wherein β (x) and γ are functions that establish an empirical scale within each regionAnd empirical wavelet- >Parameters that are needed when needed.

Then, based on the formulas (2) to (5), the detail coefficients of the EWT decomposition are calculated according to the formulas (6) and (7)And approximation coefficient->

Wherein, psi (&) is equal toAre conjugate functions; (. Cndot. ^∨ Is an inverse fourier transform.

Finally, EWT-based detail coefficientsAnd approximation coefficient->The signal f (t) is reconstructed according to equation (8).

Where x represents the convolution.

Therefore, after the current vibration signal is subjected to multiple decomposition by using the EWT decomposition method, a plurality of first initial modal components f are obtained according to the formula (9) ₀ (t) and f _i (t) and the number of first initial modal components is N.

Step 402, determining the number of first target modal components according to a preset maximum threshold and a first correlation coefficient threshold.

The preset maximum value threshold is a maximum value threshold preset for determining the number of the first target modal components, and the preset first correlation coefficient threshold is a correlation coefficient threshold preset for determining the number of the first target modal components. The first target modal component is a modal component determined from the first initial modal component. Optionally, the number of the first initial modal components, of which the vibration peak spectrum local maxima of the first initial modal components are larger than a preset maximum threshold and the first correlation coefficient is larger than a preset first correlation coefficient threshold, is counted as the number of the first target modal components, wherein the first correlation coefficient is the degree of correlation between the first initial modal components and the vibration peaks corresponding to the first initial modal components.

Step 403, determining a plurality of first target modal components from a plurality of first initial modal components according to the number of first target modal components.

Wherein after the number of first target modal components is determined, a plurality of first target modal components may be determined from the plurality of first initial modal components. Alternatively, the number of first initial modal components may be the same as the number of first target modal components, and the number of first initial modal components may also be greater than the number of first target modal components.

The number of the first target modal components is determined by presetting a maximum threshold and a first correlation coefficient threshold, and then the first target modal components are determined from the first initial modal components, so that the problem of excessive decomposition when the current vibration signal is subjected to multiple decomposition by adopting an EWT decomposition method can be avoided.

And 302, performing multiple decomposition on the current vibration signal by adopting a variation modal decomposition method to obtain a plurality of second target modal components corresponding to the current vibration signal.

In this embodiment, a variation mode decomposition method may be used to perform multiple decomposition on the current vibration signal, and multiple initial mode components obtained by decomposition are directly used as second target mode components; alternatively, a mode component of a portion selected from the plurality of initial mode components obtained by decomposition may be selected as the second target mode component.

As shown in fig. 5, a specific implementation of step 302 may include the steps of:

and 501, performing multiple decomposition on the current vibration signal by adopting a variation modal decomposition method to obtain a plurality of second initial modal components.

Wherein the second initial modal component u _k The method can be calculated by a formula (10) -a formula (13), and the calculation process is specifically as follows: firstly, presetting the number of second initial modal components as K; next, the variation problem is constructed according to equation (10).

Wherein u is _k For the second initial modal component obtained after VMD decomposition, { u _K }＝{u ₁ ,u ₂ ,…,u _K And k is the center frequency of the corresponding mode, { omega } _K }＝{ω ₁ ,ω ₂ ,…,ω _K Delta (t) is a dirac function,l as gradient ² Norms.

Then, a secondary penalty factor alpha and a Lagrangian multiplier lambda (t) are introduced in the process of solving the variation problem, and the constraint problem is converted into an unconstrained problem, so that an augmented Lagrangian formula (11) is obtained:

the equation (11) is solved according to an iterative optimization method of a multiplier alternating direction method, and in the iterative optimization process, the minimization problem of the equation (10) is converted into a saddle point for solving the augmented Lagrange of the equation (11).

Finally, an optimization solution is obtained according to the formula (12) and the formula (13):

wherein:to obtain a second initial modal component u obtained after VMD decomposition by performing Fourier inversion conversion on the residual error to obtain a real part _k 。

Step 502, determining the number of second target modal components according to a preset central frequency threshold and a second phase relation number threshold.

The preset central frequency threshold is a central frequency threshold preset for determining the number of the second target modal components. The second correlation coefficient threshold is a correlation coefficient threshold set in advance for determining the number of second target modal components. The second target modal component is a modal component determined from the second initial modal component. Optionally, counting the number of second initial modal components with a center frequency greater than a preset center frequency threshold and a second correlation number greater than a preset second correlation number threshold, and taking the number as the number of second target modal components, wherein the second correlation coefficient is a correlation degree between the second initial modal components and the corresponding vibration peaks thereof.

Step 503, according to the number of second target modal components, the server determines a plurality of second target modal components from a plurality of second initial modal components.

Wherein after the number of second target modal components is determined, a plurality of second target modal components may be determined from the plurality of second initial modal components. Alternatively, the number of second initial modal components may be the same as the number of second target modal components, and the number of second initial modal components may also be greater than the number of first target modal components.

The number of the second target modal components is determined by presetting the central frequency threshold and the second phase relation threshold, and the second target modal components are determined from the second initial modal components, so that the problem of excessive decomposition when the VMD decomposition method is adopted to carry out multiple decomposition on the current vibration signal can be avoided, and the multiple decomposition is more reasonable.

The method comprises the steps of carrying out multiple decomposition on a current vibration signal by adopting an EWT decomposition method to obtain a plurality of first target modal components corresponding to the current vibration signal, carrying out multiple decomposition on the current vibration signal by adopting a VMD decomposition method to obtain a plurality of second target modal components corresponding to the current vibration signal, wherein the EWT decomposition method can realize self-adaptive decomposition aiming at the characteristics of different signals, and the VMD decomposition method has stronger robustness on noise in the signals, namely, is not easily influenced by the noise during multiple decomposition, so that the EWT decomposition method and the VMD decomposition method are adopted to carry out multiple decomposition on the current vibration signal, and the current vibration signal is decomposed more accurately, thereby improving the prediction precision of the current vibration signal.

The following description will explain the process of determining the number of the first target modal components according to the preset maximum threshold and the first correlation coefficient threshold, which are described above in the embodiments of the present application:

In one embodiment, please refer to fig. 6, which illustrates a flowchart for determining the number of first target modal components according to an embodiment of the present application, where the method may be applied to a server. As shown in fig. 6, the method for determining the number of first target modal components may include the steps of:

step 601, counting the number of first initial modal components with local maxima larger than a maximum threshold as a first number.

When the number of the first target modal components is determined by adopting a preset maximum threshold value, searching the local maximum value of the vibration peak frequency spectrum and arranging the local maximum value according to a descending order, and marking as: a is that ₁ ≥A ₂ ≥…A _M Take A _t ＝A _M +ε(A ₁ -A _M ) Is a preset maximum threshold value, and all local maximum values larger than the preset maximum threshold value are reserved, and the number is recorded as M ₁ ，M ₁ Is a first number. Alternatively, ε may be 0.3.

Step 602, counting the number of first initial modal components with the first correlation coefficient larger than the first correlation coefficient threshold as the second number.

Wherein the first correlation coefficient represents a degree of correlation between the first initial modal component and the corresponding vibration peak. When the number of first target modal components is determined using the first correlation coefficient threshold, the number M of first target modal components is calculated according to equation (14) ₂ ，M ₂ A second number.

Wherein Y is _i X as a first initial modal component _i The first initial modal component and the corresponding vibration peak value;is the mean value of the first initial modal component, +.>The first initial modal component and the corresponding vibration peak value mean value; m is M ₂ Is the number of first target modal components. Optionally, ρ is not less than 0.2.

Step 603, determining the minimum value of the first number and the second number as the number of first target modal components.

Wherein the number of first target modal components is the smallest of the first number and the second number, i.e. M is compared ₁ And M ₂ The smaller value is taken as the number of first target modality components. For example, M ₁ Ratio M ₂ Small, then the number of first target modal components is M ₁ . For example, M ₁ When=7, as shown in fig. 7, fig. 7 is a schematic diagram of seven first target modal components according to an embodiment of the application.

As the minimum value in the first quantity and the second quantity is selected as the quantity of the first target modal components, the problem of excessive decomposition when the current vibration signal is subjected to multiple decomposition by adopting an EWT decomposition method can be avoided, so that the decomposition of the current vibration signal is more reasonable, and the obtained first target modal components are more accurate.

The following description will explain the process of determining the number of the second target modal components according to the preset center frequency threshold and the second correlation number threshold, which are described above in the embodiments of the present application:

in one embodiment, please refer to fig. 8, which illustrates a flowchart for determining the number of second target modal components according to an embodiment of the present application, where the method may be applied to a server. As shown in fig. 8, the method for determining the number of second target modal components may include the steps of:

step 801, calculating a difference ratio of center frequencies of adjacent second initial modal components, and counting the number of the second initial modal components with the difference ratio of the center frequencies larger than a center frequency threshold as a third number.

Wherein the center frequencies of adjacent second initial modal components are noted:the difference ratio of the center frequencies of the adjacent second initial modal components is calculated according to equation (15).

d _i+1 ＝(f _i+1 -f _i )/f _i+1 (15)

Optionally, take d _i+1 The number of the second initial modal components with the difference ratio of the statistical center frequency larger than the center frequency threshold value is more than or equal to 0.2 and is taken as a third number M' ₁ And M' ₁ =i+1. For example, the center frequency threshold is 0.2, and the difference ratio of the center frequencies of the eight second initial modal components is 0.50, 0.45, 0.39, 0.35, 0.28, 0.18, 0.09, it can be seen that the fifth second initial modal component The difference ratio to the center frequency of the sixth second initial modal component is 0.18, which is less than the center frequency threshold of 0.2, and therefore, a third amount M' ₁ ＝i+1＝6。

Screening the center frequencies of adjacent second initial modal components according to equation (15) can avoid excessive decomposition problems generated when the VMD is used to multiply decompose the current vibration signal.

Step 802, counting the number of second initial modal components with the second phase relation number larger than the second phase relation number threshold as a fourth number.

Wherein the second correlation coefficient represents a degree of correlation between the second initial modal component and the corresponding vibration peak. Calculating a second correlation coefficient of each second initial modal component, comparing the second correlation coefficient with a second phase relation threshold, counting the number of the second initial modal components larger than the second phase relation threshold, and taking the number as a fourth number M' ₂ . Fourth quantity M' ₂ And the second quantity M ₂ The method embodiments of the determination of (a) are similar and are not described in detail herein.

Step 803, determining the minimum value of the third number and the fourth number as the number of second target modal components.

Wherein the number of second target modal components is the smallest of the third number and the fourth number, i.e. M 'is compared' ₁ And M' ₂ The smaller value is taken as the number of second target modality components. For example, M' ₁ Ratio M' ₂ Small, then the number of second target modal components is M' ₁ . For example, M' ₁ When=7, as shown in fig. 9, fig. 9 is a schematic diagram of seven second target modal components according to an embodiment of the application.

As the minimum value in the third quantity and the fourth quantity is selected as the quantity of the second target modal components, the problem of excessive decomposition when the current vibration signal is subjected to multiple decomposition by adopting the VMD decomposition method can be avoided, and the decomposition of the current vibration signal is more reasonable.

In the embodiment of the application, in order to improve the prediction precision, a plurality of prediction methods are adopted to respectively predict each first target modal component and each second target modal component to obtain a prediction result corresponding to each prediction method.

Referring to fig. 10, a flowchart of predicting each first target modal component and each second target modal component according to an embodiment of the present application is shown, and the method may be applied to a server. As shown in fig. 10, the prediction method may include the steps of:

step 1001, for each prediction method, predicting each first target modal component by adopting the prediction method to obtain a component prediction result of each first target modal component; and adding the component prediction results of all the first target modal components to obtain a first prediction result corresponding to the prediction method.

The multiple prediction methods can be ARIMA prediction methods, BP neural network prediction methods and LSSVM prediction methods. And respectively predicting each first target modal component by adopting a prediction method aiming at each prediction method, and finally obtaining a component prediction result of the first target modal component predicted by each prediction method.

And adding the component prediction results of all the first target modal components to obtain a first prediction result corresponding to the prediction method. For example, after predicting each first target modal component by adopting an ARIMA prediction method, adding the component prediction results of each first target modal component to obtain a first prediction result corresponding to the ARIMA prediction method; predicting each first target modal component by adopting a BP neural network prediction method, and then adding the component prediction results of each first target modal component to obtain a first prediction result corresponding to the BP neural network prediction method; and predicting each first target modal component by adopting an LSSVM prediction method, and then adding the component prediction results of each first target modal component to obtain a first prediction result corresponding to the LSSVM prediction method.

Because the first target modal components are obtained by EWT multiple decomposition, the component prediction result of each first target modal component may be expressed as EWT-ARIMA, EWT-BP, and EWT-LSSVM, as shown in diagrams (d), (e), and (f) in fig. 11, and fig. 11 is a schematic diagram of a prediction result corresponding to one of multiple prediction methods provided in the embodiment of the present application. The graph (d) in fig. 11 is a schematic diagram of a first prediction result corresponding to the ARIMA prediction method, the graph (e) in fig. 11 is a schematic diagram of a first prediction result corresponding to the BP neural network prediction method, and the graph (f) in fig. 11 is a schematic diagram of a first prediction result corresponding to the LSSVM prediction method.

Step 1002, for each prediction method, predicting each second target modal component by adopting the prediction method to obtain a component prediction result of each second target modal component; and adding the component prediction results of all the second target modal components to obtain a second prediction result corresponding to the prediction method.

The multiple prediction methods are the same as the above embodiment of the prediction method for predicting each first target modal component by using the prediction method. And for each prediction method, respectively predicting each second target modal component by adopting the prediction method, and finally obtaining a component prediction result of each second target modal component.

And adding the component prediction results of all the second target modal components to obtain a second prediction result corresponding to the prediction method. For example, after predicting each second target modal component by adopting an ARIMA prediction method, adding the component prediction results of each second target modal component to obtain a second prediction result corresponding to the ARIMA prediction method; predicting each second target modal component by adopting a BP neural network prediction method, and then adding the component prediction results of each second target modal component to obtain a second prediction result corresponding to the BP neural network prediction method; and predicting each second target modal component by adopting an LSSVM prediction method, and then adding the component prediction results of each second target modal component to obtain a second prediction result corresponding to the LSSVM prediction method.

Since the second target modal components are multiply decomposed by the VMD, the component prediction result of each second target modal component can be expressed as VMD-ARIMA, VMD-BP, VMD-LSSVM, as shown in diagrams (g), (h), (i) in fig. 11. The graph (g) in fig. 11 is a second prediction result schematic diagram corresponding to the ARIMA method, the graph (h) in fig. 11 is a second prediction result schematic diagram corresponding to the BP neural network method, and the graph (i) in fig. 11 is a second prediction result schematic diagram corresponding to the LSSVM method.

Because the ARIMA prediction method has a good prediction effect on the unsteady vibration signals with complex changes, the BP neural network prediction method can automatically learn in training so as to improve the prediction precision, and the LSSVM prediction method has high prediction precision and small calculated amount so as to have high prediction speed. Therefore, the three prediction methods are adopted to respectively predict each first target modal component and each second target modal component, so that the prediction advantages of different prediction methods can be combined to predict each modal component, and the prediction accuracy is improved.

After each prediction result is obtained, a weight is required to be allocated to each prediction result according to a preset weight allocation rule. The following description will be given of the process of assigning weights to each prediction result according to the preset weight assignment rule, which is described above in the embodiment of the present application:

Referring to fig. 12, a flowchart of a weight distribution method according to an embodiment of the present application is shown, and the method may be applied to a server. As shown in fig. 12, the weight allocation method may include the steps of:

step 1201, generating a predictive evaluation index matrix according to the predictive evaluation standard.

The predictive evaluation criteria include average absolute error (Mean Absolute Error, MAE), mean square error (Mean Square Error, MSE), root mean square error (Root Mean Square Error, RMSE), average absolute percent error (Mean Absolute Percentage Error, MAPE) and symmetric average absolute percent error (Symmetry Mean Absolute Percentage Error, SMAPE), and the predictive evaluation index matrix is a matrix formed by MAE, MSE, RMSE, MAPE and SMAPE, and is expressed as: and Z= [ MAE, MSE, RMSE, MAPE, SMAPE ], and calculating the prediction evaluation standard in the prediction evaluation index matrix according to the formulas (16) to (20).

Wherein: x'. _i For the prediction result corresponding to each prediction method,is the mean value of the predicted result.

And 1202, distributing weights to the prediction results according to the prediction evaluation index matrix by adopting a fuzzy analytic hierarchy process and an entropy weight process.

And a fuzzy analytic hierarchy process is adopted, and a first initial weight is distributed to each prediction result according to the prediction evaluation index matrix. And (3) adopting an entropy weight method, and distributing second initial weights to the prediction results according to the prediction evaluation index matrix. And finally, assigning weights to the prediction results according to the first initial weights and the second initial weights.

The following will describe a process of assigning weights to each prediction result according to a prediction evaluation index matrix by using a fuzzy analytic hierarchy process and an entropy weight method, respectively:

firstly, a fuzzy analytic hierarchy process is adopted, and a first initial weight is distributed to each prediction result according to a prediction evaluation index matrix.

First, assume that there are L evaluation target indexes, rawThe formed predictive evaluation index matrix is Z _j ＝[z _j1 ,z _j2 ,…,z _jL ]The expert obtains a fuzzy consistency judgment matrix A according to the comparison judgment of the difference characteristics of the predictive evaluation indexes in pairs and the quantitative representation of the judgment basis that one predictive evaluation index matrix is more important than the other predictive evaluation index matrix _k As shown in formula (21):

wherein a is _ij As index Z _i And index Z _j The compared fuzzy judgment importance degree, K is the kth expert, k=1, 2, …, K, n is the number of evaluation objects. The blur determination importance level is expressed as shown in the formula (22):

next, a fuzzy consistency judgment matrix is formed, and the weight of the evaluation object is calculated by a power method calculation method from the formula (23) to the formula (26).

E _k ＝e _ijk ＝a _ijk /a _jik (23)

If it isThe iteration is stopped and epsilon is the given error.

Wherein e _ijk Is a as _ijk And a _jik Ratio of (E), let E _k ＝e _ijk ，E _k Calculating a power value, eta 'in the weight of the evaluation object for the power method calculation method' _jk Can pass through a _ijk Obtained from n and ζ, the value of ζ is related to n,is eta' _jk With eta' _jk Is used for the ratio of the maximum values of (c),is->And->Ratio of maximum values, ++>For E _k And->Is the product of eta _jk Is->And->The ratio of the summed values, ε, may take the value of 1e-15, i.e., 1X 10 ^-15 。

Finally, through k expert evaluations, a fuzzy analytic hierarchy process is adopted, and a first initial weight eta is distributed to each prediction result according to a prediction evaluation index matrix _j.FAHP As shown in equation (27).

And secondly, adopting an entropy weight method, and distributing second initial weights to the prediction results according to the prediction evaluation index matrix.

Assuming that there are L evaluation target indexes, the generated predictive evaluation index matrix is Z _j ＝[z _j1 ,z _j2 ,…,z _jL ]' weight η of the evaluation object is calculated by the formula (28) _j.EWM 。

/>

Wherein, the liquid crystal display device comprises a liquid crystal display device,is the probability of each predictive evaluation index, H _j Is the information entropy of each prediction evaluation index,i＝1,2,…,L，j＝1,2,…,n。

and thirdly, obtaining a time-varying combination weight according to the first initial weight and the second initial weight, as shown in a formula (29), and according to a preset weight distribution rule, the preset weight distribution rule is the obtained time-varying combination weight, so that each prediction result is distributed with a weight according to the time-varying combination weight Q, as shown in a formula (30).

Wherein eta _1k Is the weight, eta, assigned to the first predictor _2k Is the weight assigned to the second predictor.

Because the weight is allocated to each prediction result after the prediction result corresponding to each prediction method is obtained, the fuzzy analytic hierarchy process is a subjective weight allocation mode, namely the weight is allocated according to expert experience, the entropy weight process is an objective weight allocation mode, namely the weight is allocated according to the correlation between historical data research evaluation indexes, and the two methods are combined to allocate different weights to the prediction result, so that the rationality of the weight allocation can be ensured, the prediction advantages of different prediction methods can be further fully exerted, and the prediction precision of the current vibration signal can be further improved.

Please refer to fig. 11 (a), fig. b, fig. c, and table 1, wherein fig. 11 (a) is a schematic diagram of a prediction result corresponding to an ARIMA method in the prior art for performing a comprehensive prediction on a current vibration signal, fig. 11 (b) is a schematic diagram of a prediction result corresponding to a BP neural network method in the prior art for performing a comprehensive prediction on a current vibration signal, and fig. 11 (c) is a schematic diagram of a prediction result corresponding to an LSSVM method in the prior art for performing a comprehensive prediction on a current vibration signal. Table 1 shows the comparison result between the comprehensive prediction experimental result obtained by the method for predicting the vibration signal according to the embodiment of the present application and the prediction experimental result obtained by comprehensively predicting the current vibration signal by using the prior art.

The graph (j) in fig. 11 is a schematic diagram of a prediction result corresponding to the CPM comprehensive prediction method in the present solution, and according to a comparison result between the schematic diagram of the prediction result and the experimental result, it can be seen that the prediction error of the CPM comprehensive prediction method in the present solution is reduced to less than 5%, compared with the prediction error when the current vibration signal is comprehensively predicted in the prior art, that is, the prediction precision of the current vibration signal is improved.

TABLE 1

As shown in fig. 13, the embodiment of the application further provides a method for predicting a vibration signal of a converter transformer, which includes the following steps:

step 1301, performing multiple decomposition on a current vibration signal by adopting an empirical wavelet transformation decomposition method to obtain a plurality of first initial modal components;

step 1302, determining the number of first target modal components according to a preset maximum threshold and a first correlation coefficient threshold;

wherein step 1302 includes counting a number of first initial modal components having local maxima greater than a maximum threshold as a first number; counting the number of first initial modal components with the first correlation coefficient larger than a first correlation coefficient threshold as a second number; the minimum of the first number and the second number is determined as the number of first target modal components.

Step 1303, determining a plurality of first target modal components from a plurality of first initial modal components according to the number of first target modal components.

1304, performing multiple decomposition on the current vibration signal by adopting a variation modal decomposition method to obtain a plurality of second initial modal components;

step 1305, determining the number of second target modal components according to a preset central frequency threshold and a second phase relation number threshold;

step 1305 includes calculating a difference ratio of center frequencies of adjacent second initial modal components, and counting the number of the second initial modal components with the difference ratio of the center frequencies being greater than a center frequency threshold as a third number; counting the number of second initial modal components with the second phase relation number larger than a second phase relation number threshold value as a fourth number; the second correlation coefficient represents the degree of correlation between the second initial modal component and the corresponding vibration peak; the minimum of the third number and the fourth number is determined as the number of second target modal components.

Step 1306, determining a plurality of second target modal components from the plurality of second initial modal components according to the number of second target modal components.

Step 1307, for each prediction method, predicting each first target modal component by adopting the prediction method to obtain a component prediction result of each first target modal component; adding the component prediction results of all the first target modal components to obtain a first prediction result corresponding to the prediction method;

Step 1308, for each prediction method, predicting each second target modal component by adopting the prediction method to obtain a component prediction result of each second target modal component; and adding the component prediction results of all the second target modal components to obtain a second prediction result corresponding to the prediction method.

Step 1309, generating a predictive evaluation index matrix according to a predictive evaluation standard, wherein the predictive evaluation standard comprises an average absolute error, a mean square error, a root mean square error, an average absolute percentage error and a symmetrical average absolute percentage error, and the predictive evaluation index matrix is a matrix formed by the average absolute error, the mean square error, the root mean square error, the average absolute percentage error and the symmetrical average absolute percentage error;

and 1310, distributing weights to each first prediction result and each second prediction result according to the prediction evaluation index matrix by adopting a fuzzy analytic hierarchy process and an entropy weight process.

Step 1311, performing weighted summation on each first prediction result, each second prediction result and weights corresponding to each prediction result to obtain a comprehensive prediction result.

In the technical scheme provided by the embodiment of the application, the current vibration signal is subjected to multiple decomposition, so that the current vibration signal with the complex characteristic of variation is decomposed into a plurality of simple target modal components which are convenient to predict, and the prediction precision of the current vibration signal can be improved. And each target modal component is respectively predicted by adopting a plurality of prediction methods, so that the prediction advantages of different prediction methods can be combined, and the prediction precision can be improved. Finally, weights are distributed to all the prediction results, so that the prediction advantages of different prediction methods can be further fully exerted, and the prediction precision of the current vibration signal is further improved.

It should be understood that, although the steps in the flowcharts of fig. 2-13 are shown in order as indicated by the arrows, these steps are not necessarily performed in order as indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in FIGS. 2-13 may include multiple steps or stages that are not necessarily performed at the same time, but may be performed at different times, nor does the order in which the steps or stages are performed necessarily performed in sequence, but may be performed alternately or alternately with at least a portion of the steps or stages in other steps or other steps.

Referring to fig. 14, a block diagram of a vibration signal prediction apparatus 140 according to an embodiment of the present application is shown, where the vibration signal prediction apparatus 140 may be configured in a server. As shown in fig. 14, the vibration signal prediction apparatus 140 may include: a multiple decomposition module 141, a prediction module 142, an allocation module 143, and a weighted summation module 144; wherein, the liquid crystal display device comprises a liquid crystal display device,

the multiple decomposition module 141 is configured to perform multiple decomposition on a current vibration signal of the converter transformer by using at least two modal decomposition methods, so as to obtain multiple target modal components corresponding to the current vibration signal;

The prediction module 142 is configured to predict each modal component by using multiple prediction methods, so as to obtain a prediction result corresponding to each prediction method; the predicting method is used for predicting the peak value of the vibration signal at the next moment;

the allocation module 143 is configured to allocate weights to the prediction results according to a preset weight allocation rule;

the weighted summation module 144 is configured to perform weighted summation on each prediction result and the weight corresponding to each prediction result, so as to obtain a comprehensive prediction result.

In one embodiment, as shown in fig. 15, the multiple decomposition module 141 includes a first multiple decomposition unit 1411 and a second multiple decomposition unit 1412;

the first multiple decomposition unit 1411 is configured to perform multiple decomposition on the current vibration signal by using an empirical wavelet transform decomposition method, so as to obtain a plurality of first target modal components corresponding to the current vibration signal;

the second multiple decomposition unit 1412 is configured to perform multiple decomposition on the current vibration signal by using a variation mode decomposition method, so as to obtain a plurality of second target mode components corresponding to the current vibration signal.

In one embodiment, the first multiple-resolution unit includes a first multiple-resolution subunit, a first determination subunit, and a second determination subunit;

The first multiple decomposition subunit is used for performing multiple decomposition on the current vibration signal by adopting an empirical wavelet transformation decomposition method to obtain a plurality of first initial modal components;

the first determining subunit is configured to determine, according to a preset maximum threshold and a first correlation coefficient threshold, the number of first target modal components;

the second determining subunit is configured to determine a plurality of first target modal components from a plurality of first initial modal components according to the number of first target modal components.

In one embodiment, the first determining subunit is specifically configured to: counting the number of first initial modal components with local maxima larger than a maximum threshold as a first number; counting the number of first initial modal components with the first correlation coefficient larger than a first correlation coefficient threshold as a second number; the first correlation coefficient represents the degree of correlation between the first initial modal component and the corresponding vibration peak value; the minimum of the first number and the second number is determined as the number of first target modal components.

In one embodiment, the second multiple-resolution unit includes a second multiple-resolution subunit, a third determination subunit, and a fourth determination subunit;

The second multiple decomposition subunit is configured to perform multiple decomposition on the current vibration signal by using a variation mode decomposition method, so as to obtain a plurality of second initial mode components.

The third determining subunit is configured to determine, according to a preset center frequency threshold and a second correlation number threshold, a number of second target modal components.

The fourth determination subunit is configured to determine a plurality of second target modal components from a plurality of second initial modal components according to the number of second target modal components.

In one embodiment, the third determining subunit is specifically configured to: calculating the difference ratio of the center frequencies of the adjacent second initial modal components, and counting the number of the second initial modal components with the difference ratio of the center frequencies larger than the center frequency threshold as a third number; counting the number of second initial modal components with the second phase relation number larger than a second phase relation number threshold value as a fourth number; the second correlation coefficient represents the degree of correlation between the second initial modal component and the corresponding vibration peak; the minimum of the third number and the fourth number is determined as the number of second target modal components.

In one embodiment, as shown in FIG. 16, the prediction module 142 includes a first prediction unit 1421 and a second prediction unit 1422;

The first prediction unit 1421 is configured to predict, for each prediction method, each first target modal component by using a prediction method, so as to obtain a component prediction result of each first target modal component; and adding the component prediction results of all the first target modal components to obtain a first prediction result corresponding to the prediction method.

The second prediction unit 1422 is configured to predict, for each prediction method, each second target modal component by using the prediction method, so as to obtain a component prediction result of each second target modal component; and adding the component prediction results of all the second target modal components to obtain a second prediction result corresponding to the prediction method.

In one embodiment, as shown in fig. 17, the allocation module 143 includes a generation unit 1431 and an allocation unit 1432;

the generating unit 1431 is configured to generate a prediction evaluation index matrix according to a prediction evaluation criterion, where the prediction evaluation criterion includes a matrix formed by an average absolute error, a mean square error, a root mean square error, an average absolute percentage error, and a symmetrical average absolute percentage error;

The allocation unit 1432 is configured to allocate weights to the prediction results according to the prediction evaluation index matrix by using a fuzzy analytic hierarchy process and an entropy weight process.

The device for predicting the vibration signal provided by the embodiment of the application can realize the method embodiment, and the implementation principle and the technical effect are similar, and are not repeated here.

For a specific definition of the prediction means of the vibration signal, reference may be made to the definition of the prediction method of the vibration signal of the converter transformer hereinabove, and no further description is given here. The various modules in the above-described predictions of the requested vibration signal may be implemented in whole or in part by software, hardware, and combinations thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may invoke and perform the operations of the above modules.

In one embodiment of the present application, there is provided a computer device including a memory and a processor, the memory having stored therein a computer program which when executed by the processor performs the steps of:

In one embodiment of the application, the processor when executing the computer program further performs the steps of: multiple decomposition is carried out on the current vibration signal by adopting an empirical wavelet transformation decomposition method, so as to obtain a plurality of first target modal components corresponding to the current vibration signal; and performing multiple decomposition on the current vibration signal by adopting a variation modal decomposition method to obtain a plurality of second target modal components corresponding to the current vibration signal.

In one embodiment of the application, the processor when executing the computer program further performs the steps of: multiple decomposition is carried out on the current vibration signal by adopting an empirical wavelet transformation decomposition method, so as to obtain a plurality of first initial modal components; determining the number of first target modal components according to a preset maximum value threshold and a first correlation coefficient threshold; a plurality of first target modal components is determined from the plurality of first initial modal components according to the number of first target modal components.

In one embodiment of the application, the processor when executing the computer program further performs the steps of: counting the number of first initial modal components with local maxima larger than a maximum threshold as a first number; counting the number of first initial modal components with the first correlation coefficient larger than a first correlation coefficient threshold as a second number; the first correlation coefficient represents the degree of correlation between the first initial modal component and the corresponding vibration peak value; the minimum of the first number and the second number is determined as the number of first target modal components.

In one embodiment of the application, the processor when executing the computer program further performs the steps of: multiple decomposition is carried out on the current vibration signal by adopting a variable-decomposition mode decomposition method, so as to obtain a plurality of second initial mode components; determining the number of second target modal components according to a preset center frequency threshold and a second correlation number threshold; a plurality of second target modal components is determined from the plurality of second initial modal components according to the number of second target modal components.

In one embodiment of the application, the processor when executing the computer program further performs the steps of: calculating the difference ratio of the center frequencies of the adjacent second initial modal components, and counting the number of the second initial modal components with the difference ratio of the center frequencies larger than the center frequency threshold as a third number; counting the number of second initial modal components with the second phase relation number larger than a second phase relation number threshold value as a fourth number; the second correlation coefficient represents the degree of correlation between the second initial modal component and the corresponding vibration peak; the minimum of the third number and the fourth number is determined as the number of second target modal components.

In one embodiment of the application, the processor when executing the computer program further performs the steps of: for each prediction method, predicting each first target modal component by adopting the prediction method to obtain a component prediction result of each first target modal component; adding the component prediction results of all the first target modal components to obtain a first prediction result corresponding to the prediction method; for each prediction method, predicting each second target modal component by adopting the prediction method to obtain a component prediction result of each second target modal component; and adding the component prediction results of all the second target modal components to obtain a second prediction result corresponding to the prediction method.

In one embodiment of the application, the processor when executing the computer program further performs the steps of: generating a predictive evaluation index matrix according to a predictive evaluation standard, wherein the predictive evaluation index matrix comprises an average absolute error, a mean square error, a root mean square error, an average absolute percentage error and a symmetrical average absolute percentage error, and the predictive evaluation index matrix is a matrix formed by the average absolute error, the mean square error, the root mean square error, the average absolute percentage error and the symmetrical average absolute percentage error; and (3) adopting a fuzzy analytic hierarchy process and an entropy weight process, and distributing weights to all prediction results according to the prediction evaluation index matrix.

The implementation principle and technical effects of the computer device provided by the embodiment of the present application are similar to those of the above method embodiment, and are not described herein.

In one embodiment of the present application, there is provided a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of:

In one embodiment of the application, the computer program when executed by the processor further implements the steps of: multiple decomposition is carried out on the current vibration signal by adopting an empirical wavelet transformation decomposition method, so as to obtain a plurality of first target modal components corresponding to the current vibration signal; and performing multiple decomposition on the current vibration signal by adopting a variation modal decomposition method to obtain a plurality of second target modal components corresponding to the current vibration signal.

In one embodiment, the computer program when executed by the processor further performs the steps of: multiple decomposition is carried out on the current vibration signal by adopting an empirical wavelet transformation decomposition method, so as to obtain a plurality of first initial modal components; determining the number of first target modal components according to a preset maximum value threshold and a first correlation coefficient threshold; a plurality of first target modal components is determined from the plurality of first initial modal components according to the number of first target modal components.

In one embodiment of the application, the computer program when executed by the processor further implements the steps of: counting the number of first initial modal components with local maxima larger than a maximum threshold as a first number; counting the number of first initial modal components with the first correlation coefficient larger than a first correlation coefficient threshold as a second number; the first correlation coefficient represents the degree of correlation between the first initial modal component and the corresponding vibration peak value; the minimum of the first number and the second number is determined as the number of first target modal components.

In one embodiment of the application, the computer program when executed by the processor further implements the steps of: multiple decomposition is carried out on the current vibration signal by adopting a variable-decomposition mode decomposition method, so as to obtain a plurality of second initial mode components; determining the number of second target modal components according to a preset center frequency threshold and a second correlation number threshold; a plurality of second target modal components is determined from the plurality of second initial modal components according to the number of second target modal components.

In one embodiment of the application, the computer program when executed by the processor further implements the steps of: calculating the difference ratio of the center frequencies of the adjacent second initial modal components, and counting the number of the second initial modal components with the difference ratio of the center frequencies larger than the center frequency threshold as a third number; counting the number of second initial modal components with the second phase relation number larger than a second phase relation number threshold value as a fourth number; the second correlation coefficient represents the degree of correlation between the second initial modal component and the corresponding vibration peak; the minimum of the third number and the fourth number is determined as the number of second target modal components.

In one embodiment of the application, the computer program when executed by the processor further implements the steps of: for each prediction method, predicting each first target modal component by adopting the prediction method to obtain a component prediction result of each first target modal component; adding the component prediction results of all the first target modal components to obtain a first prediction result corresponding to the prediction method; for each prediction method, predicting each second target modal component by adopting the prediction method to obtain a component prediction result of each second target modal component; and adding the component prediction results of all the second target modal components to obtain a second prediction result corresponding to the prediction method.

In one embodiment of the application, the computer program when executed by the processor further implements the steps of: generating a predictive evaluation index matrix according to a predictive evaluation standard, wherein the predictive evaluation index matrix comprises an average absolute error, a mean square error, a root mean square error, an average absolute percentage error and a symmetrical average absolute percentage error, and the predictive evaluation index matrix is a matrix formed by the average absolute error, the mean square error, the root mean square error, the average absolute percentage error and the symmetrical average absolute percentage error; and (3) adopting a fuzzy analytic hierarchy process and an entropy weight process, and distributing weights to all prediction results according to the prediction evaluation index matrix.

The computer readable storage medium provided in this embodiment has similar principles and technical effects to those of the above method embodiment, and will not be described herein.

Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile 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), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.

The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.

The foregoing examples illustrate only a few embodiments of the application, which are described in detail and are not to be construed as limiting the scope of the claims. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of protection of the present application is to be determined by the appended claims.

Claims

1. A method for predicting a converter transformer vibration signal, the method comprising:

performing multiple decomposition on a current vibration signal of a converter transformer by adopting at least two modal decomposition methods to obtain a plurality of target modal components corresponding to the current vibration signal;

predicting each target modal component by adopting a plurality of prediction methods to obtain a prediction result corresponding to each prediction method; the prediction method is used for predicting the peak value of the vibration signal at the next moment;

According to a preset weight distribution rule, distributing weights for all the prediction results;

and carrying out weighted summation on each prediction result and the weight corresponding to each prediction result to obtain a comprehensive prediction result.

2. The method according to claim 1, wherein the performing multiple decomposition on the current vibration signal of the converter transformer by using at least two modal decomposition methods to obtain a plurality of target modal components corresponding to the current vibration signal includes:

performing multiple decomposition on the current vibration signal by adopting an empirical wavelet transformation decomposition method to obtain a plurality of first target modal components corresponding to the current vibration signal;

and performing multiple decomposition on the current vibration signal by adopting a variation modal decomposition method to obtain a plurality of second target modal components corresponding to the current vibration signal.

3. The method according to claim 2, wherein the performing multiple decomposition on the current vibration signal by using an empirical wavelet transform decomposition method to obtain a plurality of first target modal components corresponding to the current vibration signal includes:

performing multiple decomposition on the current vibration signal by adopting an empirical wavelet transformation decomposition method to obtain a plurality of first initial modal components;

Determining the number of the first target modal components according to a preset maximum value threshold and a first correlation coefficient threshold;

the plurality of first target modal components is determined from the plurality of first initial modal components according to the number of first target modal components.

4. A method according to claim 3, wherein determining the number of first target modal components according to a preset maximum threshold and a first correlation coefficient threshold comprises:

counting the number of the first initial modal components with local maxima greater than the maximum threshold as a first number;

counting the number of the first initial modal components with a first correlation coefficient larger than the first correlation coefficient threshold as a second number; the first correlation coefficient represents a degree of correlation between the first initial modal component and a corresponding vibration peak;

the minimum of the first number and the second number is determined as the number of first target modal components.

5. The method according to claim 2, wherein the performing multiple decomposition on the current vibration signal by using a variational mode decomposition method to obtain a plurality of target second mode components corresponding to the current vibration signal includes:

Multiple decomposition is carried out on the current vibration signal by adopting a variable-decomposition mode decomposition method, so as to obtain a plurality of second initial mode components;

determining the number of the second target modal components according to a preset center frequency threshold and a second correlation number threshold;

and determining the second target modal components from the second initial modal components according to the number of the second target modal components.

6. The method of claim 5, wherein determining the number of second target modal components according to a preset center frequency threshold and a second phase relationship threshold comprises:

calculating the difference ratio of the center frequencies of the adjacent second initial modal components, and counting the number of the second initial modal components with the difference ratio of the center frequencies larger than the center frequency threshold as a third number;

counting the number of the second initial modal components with a second correlation number greater than the second correlation number threshold as a fourth number; the second correlation coefficient represents a degree of correlation between the second initial modal component and a corresponding vibration peak;

the minimum of the third number and the fourth number is determined as the number of second target modal components.

7. The method according to any one of claims 2-6, wherein predicting each of the modal components by using a plurality of prediction methods to obtain a prediction result corresponding to each of the prediction methods includes:

for each prediction method, predicting each first target modal component by adopting the prediction method to obtain a component prediction result of each first target modal component; adding component prediction results of all first target modal components to obtain a first prediction result corresponding to the prediction method;

for each prediction method, predicting each second target modal component by adopting the prediction method to obtain a component prediction result of each second target modal component; and adding the component prediction results of all the second target modal components to obtain a second prediction result corresponding to the prediction method.

8. The method according to any one of claims 1-6, wherein assigning weights to each of the predicted outcomes according to a preset weight assignment rule comprises:

generating a prediction evaluation index matrix according to a prediction evaluation standard, wherein the prediction evaluation standard comprises an average absolute error, a mean square error, a root mean square error, an average absolute percentage error and a symmetrical average absolute percentage error, and the prediction evaluation index matrix is a matrix formed by the average absolute error, the mean square error, the root mean square error, the average absolute percentage error and the symmetrical average absolute percentage error;

And a fuzzy analytic hierarchy process and an entropy weight process are adopted, and weights are distributed to the prediction results according to the prediction evaluation index matrix.

9. A prediction device for vibration signals of a converter transformer, the device comprising:

the prediction module is used for predicting each modal component by adopting a plurality of prediction methods to obtain a prediction result corresponding to each prediction method; the prediction method is used for predicting the peak value of the vibration signal at the next moment;

the distribution module is used for distributing weights for the prediction results according to preset weight distribution rules;

10. A computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, implements the steps of the method of any of claims 1 to 8.

11. A computer-readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 8.