Running state monitoring method for electrical equipment of wind power plant

Abstract

The invention discloses an operation state monitoring method for electrical equipment of a wind power plant, which belongs to the technical field of data processing, and comprises the following steps: collecting temperature data of the electrical equipment under different abnormities, performing CEEMDAN decomposition on the data, calculating the number of IMFs, the minimum cross correlation coefficient, the entropy sum and the energy proportion, and calculating the temperature of the electrical equipment under different abnormities; sequencing the four influence factors by using grey correlation analysis, and preliminarily judging whether the electrical equipment is abnormal or not through the first two factors; the method comprises the following steps: performing VMD decomposition on sound data of current electrical equipment, calculating a multi-scale permutation entropy of IMF, dividing the sound data into a noise dominant class, a signal-noise mixed class and a signal dominant class, performing denoising and moving average filtering on the first two classes by using an improved wavelet threshold respectively, reconstructing a denoised sound signal, calculating a distance between a sound signal characteristic frequency and a standard characteristic frequency, and calculating a sound signal characteristic frequency according to the distance between the sound signal characteristic frequency and the standard characteristic frequency. And judging whether the electrical equipment is abnormal through the distance. The method considers the temperature and sound, judges whether the equipment is abnormal in stages, and improves the monitoring accuracy.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing. Specifically, it relates to a method for monitoring the operating status of electrical equipment in a wind farm. Background Art

[0002] Wind power generation is an important part of renewable energy and is of great significance for alleviating the tight power supply and environmental protection. However, with the rapid development of the wind power generation industry, electrical equipment such as wind turbine nacelles and step-up station transformers frequently fail, which has become a key factor restricting the development of the industry. Therefore, it is particularly important to monitor the operating status of electrical equipment in a wind farm.

[0003] Electrical equipment failures will be reflected by changes in physical quantities such as temperature and sound, and there will be a continuous trend before the failure intensifies. Therefore, temperature and sound need to be monitored online. Most traditional methods start from either the temperature or the sound of electrical equipment. In terms of temperature, the temperature value at the current moment is compared with the temperature value when the wind power equipment is normal to judge the abnormality of the electrical equipment. This method does not deeply consider the factors causing the temperature rise of the electrical equipment and is prone to misjudgment. In terms of sound, usually, experienced maintenance personnel need to go to the site to actually listen to the situation of the electrical equipment to judge whether it is abnormal, and simply relying on manual experience cannot guarantee the judgment of whether the electrical equipment is abnormal. To solve the above problems, a judgment method that combines the temperature and sound of electrical equipment is provided. In the first stage, starting from the temperature of the current electrical equipment, it is initially judged whether there is an abnormality. In the second stage, starting from the sound of the electrical equipment collected currently, it is judged again whether the electrical equipment has an abnormality, effectively improving the accuracy of judging whether the electrical equipment has an abnormality. The present invention provides the following technical solutions. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for monitoring the operating status of electrical equipment in a wind farm, so as to solve the problem in the prior art that when monitoring the operation of electrical equipment in a wind farm, only starting from either the temperature or the sound of the wind power equipment, the information of the two cannot be fully combined to achieve a more accurate judgment of the abnormality of the wind power equipment.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] A method for monitoring the operating status of electrical equipment in a wind farm includes the following steps:

[0007] S1. Collect the temperature data of electrical equipment under different abnormalities, perform CEEMDAN decomposition on the data to obtain a set of IMFs, and calculate the number, minimum cross-correlation coefficient, entropy sum, and energy ratio of each IMF;

[0008] S2. Rank the four influencing factors in S1 through grey relational analysis, find the top two factors, and preliminarily determine whether the electrical equipment is abnormal using these two factors;

[0009] S3. If it is preliminarily concluded that the electrical equipment is abnormal based on temperature, perform VMD decomposition on the sound data of the current electrical equipment to obtain a set of IMFs, calculate the multi-scale permutation entropy of each IMF and plot the permutation entropy spectrum, and classify the IMFs into three categories: noise-dominated, signal-noise mixed, and signal-dominated according to the number of extreme points in the permutation entropy spectrum;

[0010] S4. Denoise the noise-dominated IMFs using improved wavelet threshold denoising, filter the signal-noise mixed IMFs using moving average filtering, do not process the signal-dominated IMFs, and reconstruct the denoised sound signal;

[0011] S5. Calculate the distance between the characteristic frequency of the sound signal and the standard characteristic frequency. If the distance is greater than the discrimination threshold, it indicates that the electrical equipment is abnormal; otherwise, it indicates normal.

[0012] As a further solution of the present invention, the method for obtaining the minimum cross-correlation coefficient and energy ratio of IMFs in S1 is as follows:

[0013] Obtain several components {IMF1, IMF2, IMF3,..., IMFN} of the IMF;

[0014] Calculate the cross-correlation coefficient between each IMF according to r(IMFi, IMFj) = Cov(IMFi, IMFj) / sqrt(Var[IMFi]*Var[IMFj]);

[0015] Among them, r represents the cross-correlation coefficient, Cov represents the covariance, sqrt represents taking the square root, and Var represents the variance;

[0016] Find the minimum cross-correlation coefficient rmin among them;

[0017] Calculate the energy ratio of the first M IMFs to the original temperature signal;

[0018] Among them, represents rounding up, and N represents the total number of IMFs;

[0019] Calculate the energy ratio according to H = (IMF1^2 + IMF2^2 +... + IMFM^2) / T^2;

[0020] Among them, T represents the original temperature signal.

[0021] As a further solution of the present invention, the specific steps for ranking the four influencing factors using grey relational analysis in S2 are as follows:

[0022] S21. Determine the evaluation objects, namely the number of IMFs, the minimum cross-correlation coefficient, the sum of entropy values, and the energy ratio; determine the reference series, namely the different anomalies of the wind power equipment, and directly represent mild, moderate, severe, and extreme levels with the numbers 1, 2, 3, and 4 respectively.

[0023] S22. Perform dimensionless processing on the values of the four influencing factors actually collected, mainly by performing mean normalization, that is, each element of the series is divided by its mean.

[0024] S23. Calculate the absolute difference between the different levels of anomalies of the wind power equipment and the four influencing factors according to the formula △oi(k) = |X0(k) - Xi(k)|, where X0(k) is the reference series and Xi(k) is the comparison series.

[0025] S24. Find the maximum and minimum values of the absolute values from S23, that is, calculate according to the formula

[0026] △max = max{△oi(k)}, △min = min{△oi(k)}.

[0027] S25. Calculate the correlation coefficient between the comparison series and the reference series at each moment according to the formula Gi(k) = (△min + ρ * △max) / (△oi(k) + ρ * △max), where ρ is the resolution coefficient.

[0028] S26. Calculate the correlation degree between each influencing factor and the different anomaly levels of the wind power equipment according to the formula ri = sum(Gi(k)) / g, where g is the number of elements in the series.

[0029] S27. Sort the four influencing factors from largest to smallest according to the correlation degree.

[0030] As a further solution of the present invention, in S2, sort the four influencing factors according to the correlation degree, find the top two influencing factors, and actually calculate the values of these two influencing factors according to the current temperature data of the wind power equipment. If one of them exceeds the threshold obtained under normal wind power equipment, it is initially determined that the current wind power equipment has an anomaly.

[0031] As a further solution of the present invention, the method for calculating the multi-scale permutation entropy of each IMF in S3 is as follows:

[0032] S31. Obtain m IMF sequences after VMD decomposition, where IMFm = {xm(i), i = 1, 2,..., N1}, and N1 is the length of the sequence.

[0033] S32. Select a scale factor τ, and according to the formula Perform coarse-graining processing on each IMF sequence, where τ is a positive integer and [] is the rounding symbol.

[0034] S33. Calculate the permutation entropy of the sequence .

[0035] S34. Substitute different scale factors τ respectively, repeat steps S32 and S33, and calculate the permutation entropy at different scales.

[0036] As a further solution of the present invention, in S3, the IMF categories are divided according to the number of extreme points in the permutation entropy spectrum. If the number of extreme points in the IMF permutation entropy spectrum is not less than 2, it is classified as the noise-dominated category. If the number of extreme points in the IMF permutation entropy spectrum is 1, it is classified as the signal-noise mixed category. If the IMF permutation entropy spectrum shows a monotonic trend and no extreme points are found, it is classified as the signal-dominated category.

[0037] As a further solution of the present invention, the method for using the improved wavelet threshold denoising for the noise-dominated IMF in S4 is as follows:

[0038] S41. Perform wavelet transform on the noise-dominated IMF according to the formula , where W(a, b) represents the wavelet coefficient after transformation, represents the complex conjugate of the wavelet basis function;

[0039] S42. Perform threshold processing on the wavelet coefficients. The specific improved threshold function is , where W is the original wavelet coefficient, W' is the processed wavelet coefficient, λ is the set threshold, and a is the preset adjustment factor;

[0040] S43. Reconstruct the original signal using the processed wavelet coefficients to obtain the denoised IMF sequence.

[0041] As a further solution of the present invention, the selection of the threshold λ is closely related to the denoising effect of the IMF. By setting an adaptive threshold, the threshold can be adaptively changed with the change of the decomposition scale, making it consistent with the propagation characteristics of the noise at each scale of the wavelet transform, achieving a better denoising effect. The specific adaptive threshold formula is where represents the variance of the wavelet coefficients.

[0042] As a further solution of the present invention, the method for calculating the distance between the characteristic frequency of the sound signal and the standard characteristic frequency in S5 is as follows:

[0043] S51. After the sound signal is Fourier-transformed, it can be expressed as P = (p 1 , p 2 ,..., p i ,..., p N2 ), where p iis the i - th harmonic amplitude of the sound signal, and N2 represents the number of harmonics;

[0044] S52. Sort the sound signal according to the frequency amplitude. The frequency corresponding to the j - th in the sorting is f(j). According to the formula calculate the proportion of the first n frequencies;

[0045] S53. When H(n) ≥ 0.8, regard the first n sorted frequencies as the characteristic frequencies of the sound signal;

[0046] S54. Record that the number of standard characteristic frequencies of the sound signal obtained when the wind power equipment operates normally is n1, and the amplitudes of the first j frequencies of its standard characteristic frequencies are p n1j and calculate the distance between the characteristic frequency and the standard characteristic frequency according to the formula The beneficial effects of the present invention are as follows:

[0047] 1. The present invention fully considers the real - time temperature and real - time sound data of electrical equipment in an actual wind farm. In the first stage, starting from the real - time collected temperature data, it preliminarily judges whether the electrical equipment is abnormal. In the second stage, starting from the real - time collected sound data, it judges again whether the electrical equipment is abnormal. By combining the current temperature and sound of the electrical equipment in the wind farm and considering them comprehensively, it fully excavates the useful information of the electrical equipment in the current situation and improves the accuracy of the abnormal judgment of the electrical equipment.

[0048] 2. The present invention fully considers several factors affecting the temperature of electrical equipment, ranks the importance of these influencing factors through grey relational analysis, finds the top two factors, excludes some unimportant factors, and improves the reliability of the abnormal monitoring of electrical equipment in the first stage.

[0049] 3. The present invention takes into account that the actually collected sound information will be affected by surrounding noise. Therefore, by decomposing and classifying the original sound signal, a more detailed noise reduction effect is achieved, which prepares for the subsequent abnormal judgment of electrical equipment through sound information. Description of the Drawings

[0050] Figure 1 is the flow schematic diagram of the present invention. Detailed Embodiments

[0051] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0052] As Figure 1, An operating state monitoring method for electrical equipment in a wind farm, including:

[0053] S1. Collect the temperature data of electrical equipment under different abnormalities, perform CEEMDAN decomposition on the data to obtain a set of IMFs, and calculate the number, minimum cross-correlation coefficient, sum of entropy values, and energy ratio of each IMF;

[0054] The calculation method of the minimum cross-correlation coefficient of each IMF is as follows:

[0055] Obtain several components of the IMF {IMF1, IMF2, IMF3,...};

[0056] Calculate the cross-correlation coefficient between each IMF according to r(IMFi, IMFj) = Cov(IMFi, IMFj) / sqrt(Var[IMFi]*Var[IMFj]);

[0057] Among them, r represents the cross-correlation coefficient, Cov represents the covariance, sqrt represents taking the square root, and Var represents the variance;

[0058] Find out the minimum cross-correlation coefficient rmin among them;

[0059] The calculation method of the sum of entropy values is as follows:

[0060] For each IMF, represent it as a series of discrete numerical points, denoted as {a1, a2,..., aw}, where w is the number of samples;

[0061] Determine the value range of the IMF, denoted as [p, q], and divide the value range into G equal-width intervals, denoted as {I1, I2,..., IG};

[0062] For the discrete numerical points of each IMF, count which interval they fall into. If ni is the number of numerical points falling into the interval Ii, then record the frequency as fi = ni / w. Here, the frequency and probability are equivalent;

[0063] According to the formula Calculate the entropy value of each IMF, and then according to the formula H 总 = ∑H(IMF) to calculate the sum of entropy values of all IMFs;

[0064] The calculation method of the energy ratio is as follows:

[0065] Obtain several components of the IMF {IMF1, IMF2, IMF3,..., IMFN};

[0066] Calculate the cross-correlation coefficient between each IMF according to r(IMFi, IMFj) = Cov(IMFi, IMFj) / sqrt(Var[IMFi]*Var[IMFj]);

[0067] Among them, r represents the cross - correlation coefficient, Cov represents the covariance, sqrt represents taking the square root, and Var represents the variance;

[0068] Find the minimum cross - correlation coefficient rmin among them;

[0069] Calculate the energy proportion of the first M IMFs to the original temperature signal;

[0070] Among them, represents rounding up, and N represents the total number of IMFs;

[0071] Calculate the energy proportion according to H=(IMF1^2 + IMF2^2+...+IMFM^2) / T^2;

[0072] Among them, T represents the original temperature signal;

[0073] S2. Rank the four influencing factors in S1 through grey relational analysis, find the top two factors, and preliminarily judge whether the electrical equipment is abnormal by using these two factors;

[0074] The specific steps for ranking the four influencing factors by using grey relational analysis are as follows:

[0075] S21. Determine the evaluation objects, namely the number of IMFs, the minimum cross - correlation coefficient, the sum of entropy values, and the energy proportion; determine the reference sequence, namely the different abnormalities of the wind power equipment. Represent mild, moderate, severe, and extreme directly with numbers 1, 2, 3, and 4;

[0076] S22. Perform dimensionless processing on the values of the four influencing factors actually collected. Since the physical meanings of the influencing factors are different, the dimensions of the data are not necessarily the same, which is not convenient for comparison. Therefore, when performing grey analysis, dimensionless processing is generally required, and there are various methods for dimensionless processing, including initial value method, mean value method, etc. Which specific method to choose depends on the specific situation;

[0077] In an embodiment of the present invention, the selected dimensionless method is mean value processing, that is, each element of the sequence is divided by its mean value;

[0078] S23. Calculate the absolute difference between the different degrees of abnormality of the wind power equipment and the four influencing factors according to the formula △oi(k)=|X0(k)-Xi(k)|, where X0(k) is the reference sequence and Xi(k) is the comparison sequence;

[0079] S24. Find the maximum and minimum values of the absolute values from S23, that is, according to the formula

[0080] △max = max{△oi(k)}, and △min = min{△oi(k)} are calculated;

[0081] S25. Calculate the correlation coefficients between the comparison sequence and the reference sequence at each moment according to the formula Gi(k) = (△min + ρ * △max) / (△oi(k) + ρ * △max), where ρ is the discrimination coefficient, usually taken as 0.5;

[0082] S26. Calculate the correlation degree between each influencing factor and different abnormal degrees of the wind power equipment according to the formula ri = sum(Gi(k)) / g, where g is the number of elements in the sequence;

[0083] S27. Sort the four influencing factors from large to small according to the correlation degree;

[0084] Sort the four influencing factors according to the correlation degree, find the top two influencing factors, calculate the values of these two influencing factors according to the actual temperature data of the current wind power equipment, and if one of them exceeds the threshold obtained under normal wind power equipment, it is initially judged that the current wind power equipment is abnormal;

[0085] In this step, the four influencing factors of the electrical equipment temperature are sorted through grey relational analysis, the top two influencing factors are found, and the influence of some unimportant factors on the electrical equipment temperature is excluded, improving the accuracy of the initial abnormal judgment of the electrical equipment;

[0086] S3. Perform VMD decomposition on the sound data of the current electrical equipment to obtain a set of IMFs, calculate the multi-scale permutation entropy of each IMF and draw the permutation entropy spectrum, and classify the IMFs into three categories: noise-dominated, signal-noise mixed, and signal-dominated according to the number of extreme points of the permutation entropy spectrum;

[0087] The method for calculating the multi-scale permutation entropy of each IMF is as follows:

[0088] S31. Obtain m IMF sequences after VMD decomposition, where IMFm = {xm(i), i = 1, 2,..., N1}, and N1 is the length of the sequence;

[0089] S32. Select a scale factor τ, and according to the formula perform coarse-graining processing on each IMF sequence, where τ is a positive integer, and [] is the rounding symbol;

[0090] Coarse-graining is a scale transformation of the original time series. The original data points are merged into fewer data points by averaging or summarizing, so as to obtain a new sequence that describes the characteristics of the time series on a larger scale. The time series after coarse-graining retains the key characteristics of the original time series while simplifying the relative size relationship between elements, making it easier to calculate the permutation entropy.

[0091] S33, sequence Calculate its permutation entropy;

[0092] S34, respectively substitute different scale factors τ, repeat steps S32 and S33, and calculate the permutation entropy at different scales;

[0093] The IMF category is divided according to the number of extreme value points of the permutation entropy spectrum. If the number of extreme value points of the IMF permutation entropy spectrum is not less than 2, it is classified as the noise-dominated category. If the number of extreme value points of the IMF permutation entropy spectrum is 1, it is classified as the signal-noise mixed category. If the IMF permutation entropy spectrum shows a monotonic trend and no extreme value points are found, it is classified as the signal-dominated category.

[0094] For noise-dominated signals, their multi-scale permutation entropy spectra often show more complex and changeable characteristics. The introduction of noise will increase the uncertainty of the signal. The permutation entropy values ​​of noise signals at different scales are usually larger than those of noise-free signals, and the change of permutation entropy may be more drastic as the scale increases. For noise-free signals, they often have more regular waveforms and frequencies, so the permutation entropy values ​​at different scales do not change much. Specifically, as the scale increases, the permutation entropy of the signal may show a linear increase or a stable change pattern, and this change is usually smooth and monotonous. For mixed signal-noise signals, their change pattern is often between the above two signals, and has some similarities with both.

[0095] S4, use improved wavelet threshold to denoise the IMF dominated by noise, use sliding average filtering to denoise the IMF mixed with signal and noise, and do not process the IMF dominated by signal, and reconstruct the denoised sound signal;

[0096] The steps of improving wavelet threshold denoising are:

[0097] S41, according to the formula The noise-dominated IMF is subjected to wavelet transform, where W(a,b) represents the transformed wavelet coefficients. represents the complex conjugate of the wavelet basis function;

[0098] S42, threshold processing is performed on the wavelet coefficients. The specific improved threshold function is: Among them, W is the original wavelet coefficient, W' is the processed wavelet coefficient, λ is the set threshold, and a is a preset adjustment factor that can adjust the smoothness of the threshold function;

[0099] In an embodiment of the present invention, db5 is selected as the wavelet basis function, and the decomposition level is set to 4;

[0100] S43. Reconstruct the original signal using the processed wavelet coefficients to obtain the denoised IMF sequence;

[0101] The selection of the threshold λ is closely related to the denoising effect of the IMF. By setting an adaptive threshold, the threshold can adaptively change with the decomposition scale, making it consistent with the propagation characteristics of the noise at each scale of the wavelet transform, achieving a better denoising effect. The specific adaptive threshold formula is Among them, represents the variance of the wavelet coefficients;

[0102] When processing signals using the traditional soft threshold function, since all wavelet coefficients smaller than the threshold are set to zero, some important features of the signal will be lost; while when processing signals using the traditional hard threshold function, since all wavelet coefficients larger than the threshold are directly retained, some noise in the signal is not completely removed in the end, resulting in a poor denoising effect. The present invention introduces an adjustment factor for these two problems to obtain an improved adaptive threshold function, and also performs adaptive processing on the thresholds at different scales, enabling it to adaptively change with the scale. While ensuring an improved denoising effect, the flexibility and applicability of the denoising algorithm are enhanced;

[0103] S5. Calculate the distance between the characteristic frequency of the sound signal and the standard characteristic frequency. If this distance is greater than the discrimination threshold, it indicates that the electrical equipment is abnormal; otherwise, it indicates normal;

[0104] When the electrical equipment is in an abnormal state, its internal mechanical movement, current flow, etc. may all be different from the normal state. These differences are usually reflected in the generated sound. The characteristic frequency is an important attribute of the sound signal, which reflects the relative intensity or energy distribution of each frequency component in the sound. Calculate the Euclidean distance between the characteristic frequency of the current state of the electrical equipment and the normal state, and compare this distance with the judgment threshold to effectively determine whether the current electrical equipment is abnormal;

[0105] The method for calculating the distance between the characteristic frequency of the sound signal and the standard characteristic frequency is:

[0106] S51. After the sound signal is Fourier-transformed, it is expressed as P = (p 1 , p 2 ,..., p i ,..., p N2 ), where pi is the amplitude of the i-th harmonic of the sound signal, and N2 represents the number of harmonics;

[0107] S52. Sort the sound signal according to the frequency amplitude. The frequency corresponding to the j-th in the sorting is f(j). According to the formula Calculate the proportion of the first n frequencies;

[0108] S53. When H(n) ≥ 0.8, then regard the first n sorted frequencies as the characteristic frequencies of the sound signal;

[0109] S54. Record that the number of standard characteristic frequencies of the sound signal obtained when the wind power equipment operates normally is n1, and the amplitudes of the first j frequencies of its standard characteristic frequencies are p n1j , and according to the formula Calculate the distance between the characteristic frequency and the standard characteristic frequency;

[0110] In the description of the specification, the descriptions referring to terms such as "one embodiment", "example", "specific example", etc. mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0111] The above content is only an example and explanation of the present invention. Those skilled in the art of the present technology make various modifications or supplements to the described specific embodiments or use similar ways to replace them. As long as they do not deviate from the invention or exceed the scope defined by the claims of the present invention, they should all belong to the protection scope of the present invention.

Claims

1. A method for monitoring the operating status of electrical equipment in a wind farm, characterized in that: It includes the following steps: S1. Collect the temperature data of electrical equipment under different abnormalities, decompose the data by CEEMDAN to obtain a set of IMFs, and calculate the number, minimum cross-correlation coefficient, sum of entropy values, and energy ratio of each IMF; S2. Rank the four influencing factors in S1 through grey relational analysis, find the top two factors, and preliminarily judge whether the electrical equipment is abnormal by using these two factors; S3. If it is initially concluded that the electrical equipment is abnormal based on the temperature, perform VMD decomposition on the sound data of the current electrical equipment to obtain a set of IMFs, calculate the multi-scale permutation entropy of each IMF and draw a permutation entropy spectrum, and classify the IMFs into three categories: dominated by noise, mixed signal and noise, and dominated by signal according to the number of extreme points of the permutation entropy spectrum; S4. Denoise the IMFs dominated by noise by using improved wavelet threshold denoising, filter the IMFs with mixed signal and noise by using moving average filtering, do not process the IMFs dominated by signal, and reconstruct the denoised sound signal; S5. Calculate the distance between the characteristic frequency of the sound signal and the standard characteristic frequency. If the distance is greater than the discrimination threshold, it indicates that the electrical equipment is abnormal, otherwise, it indicates normal.

2. A method for monitoring the operating status of wind farm electrical equipment according to claim 1, characterized in that: The confirmation method of different abnormal electrical equipment in S1 is: Collect the maintenance records after the historical electrical equipment fails, including the name of the electrical equipment, the start time of maintenance, and the end time of maintenance; Conduct statistical analysis on the maintenance time data, and calculate the mean value Tmean and standard deviation Tstd of the maintenance time of each electrical equipment; If t ≤ Tmean - Tstd, it indicates mild abnormality; if Tmean - Tstd < t ≤ Tmean, it indicates moderate abnormality; Tmean < t ≤ Tmean + Tstd, it indicates severe abnormality; if t > Tmean + Tstd, it indicates extremely severe abnormality; Among them, t represents the current maintenance time.

3. A method for monitoring the operating status of wind farm electrical equipment according to claim 1, characterized in that: The method for obtaining the minimum cross-correlation coefficient and energy ratio of IMFs in S1 is: Obtain several components {IMF1, IMF2, IMF3,..., IMFN} of the IMF; Calculate the cross-correlation coefficient between each IMF according to r(IMFi, IMFj) = Cov(IMFi, IMFj) / sqrt(Var[IMFi]*Var[IMFj]); Among them, r represents the cross-correlation coefficient, Cov represents the covariance, sqrt represents taking the square root, and Var represents the variance; Find the minimum cross-correlation coefficient rmin among them; Calculate the energy ratio of the first M IMFs to the original temperature signal; in, represents rounding up, and N represents the total number of IMFs; Calculate the energy ratio according to H = (IMF1^2 + IMF2^2 +... + IMFM^2) / T^2; Among them, T represents the original temperature signal.

4. A method for monitoring the operating status of wind farm electrical equipment according to claim 1, characterized in that: The specific steps for ranking the four influencing factors by using grey relational analysis in S2 are: S21. Determine the evaluation objects, namely the number of IMFs, the minimum cross-correlation coefficient, the sum of entropy values, and the energy ratio; determine the reference sequence, namely the different abnormalities of the wind power equipment, and directly represent mild, moderate, severe, and extremely severe with numbers 1, 2, 3, and 4; S22, non-dimensionalizing the values ​​of the four influencing factors actually collected, mainly performing mean processing, that is, dividing the elements of each series by its mean; S23. Calculate the absolute difference between the different degrees of abnormality of the wind power equipment and the four influencing factors according to the formula △oi(k)=|X0(k)-Xi(k)|, where X0(k) is the reference sequence and Xi(k) is the comparison sequence; S24, find the maximum and minimum absolute values ​​from S23, that is, according to the formula △max=max{△oi(k)}, △min=min{△oi(k)} is calculated; S25, calculating the correlation coefficient of the comparison sequence and the reference sequence at each time according to the formula Gi(k)=(△min+ρ*△max) / (△oi(k)+ρ*△max), where ρ is the resolution coefficient; S26. Calculate the correlation between each influencing factor and different abnormality levels of the wind power equipment according to the formula ri=sum(Gi(k)) / g, where g is the number of elements in the sequence; S27. Sort the four influencing factors from large to small according to the degree of correlation.

5. The method for monitoring the operating status of wind farm electrical equipment according to claim 1, characterized in that: In S2, the four influencing factors are sorted according to the correlation, the top two influencing factors are found, and the values ​​of the two influencing factors are actually calculated based on the current temperature data of the wind power equipment. If one of them exceeds the threshold value obtained under normal wind power equipment, it is preliminarily judged that the current wind power equipment is abnormal.

6. A method for monitoring the operating status of wind farm electrical equipment according to claim 1, characterized in that: The method for calculating the multi-scale permutation entropy of each IMF in S3 is: S31, obtaining m IMF sequences after VMD decomposition, where IMFm={xm(i), i=1,2,...,N1}, and N1 is the length of the sequence; S32. Select a scale factor τ according to the formula Each IMF sequence is coarse-grained, where τ is a positive integer and [] is a rounding symbol; S33, sequence Calculate its permutation entropy; S34. Substitute different scale factors τ respectively, repeat steps S32 and S33, and calculate the permutation entropy at different scales.

7. A method for monitoring the operating status of wind farm electrical equipment according to claim 1, characterized in that: In S3, IMF categories are divided according to the number of extreme points in the permutation entropy spectrum. If the number of extreme points in the IMF permutation entropy spectrum is not less than 2, it is classified as noise-dominated. If the number of extreme points in the IMF permutation entropy spectrum is 1, it is classified as a signal-noise mixed class. If the IMF permutation entropy spectrum shows a monotonic trend and no extreme points are found, it is classified as a signal-dominated class.

8. A method for monitoring the operating status of wind farm electrical equipment according to claim 1, characterized in that: The method of denoising the IMF of the noise-dominated class in S4 using improved wavelet threshold is: S41, according to the formula The noise-dominated IMF is subjected to wavelet transform, where W(a,b) represents the transformed wavelet coefficients. represents the complex conjugate of the wavelet basis function; S42, threshold processing is performed on the wavelet coefficients. The specific improved threshold function is: Wherein, W is the original wavelet coefficient, W' is the processed wavelet coefficient, λ is the set threshold, and a is the preset adjustment factor; S43. Reconstruct the original signal using the processed wavelet coefficients to obtain a denoised IMF sequence.

9. A method for monitoring the operating status of electrical equipment in a wind farm according to claim 8, characterized in that: The specific adaptive threshold formula is: in, represents the variance of the wavelet coefficients.

10. A method for monitoring the operating status of wind farm electrical equipment according to claim 1, characterized in that: The method for calculating the distance between the characteristic frequency of the sound signal and the standard characteristic frequency in S5 is: S51, the sound signal is expressed as P = (p1, p2, ..., p i ,...,p N2 ), where p i is the i-th frequency multiple amplitude of the sound signal, and N2 represents the frequency multiple number; S52. Sort the sound signals by frequency amplitude. The frequency corresponding to the jth frequency is f(j). According to the formula Calculate the proportion of the first n frequencies; S53, when H(n)>=0.8, the first n frequencies in the sorting are regarded as the characteristic frequencies of the sound signal; S54. The standard characteristic frequency number of the sound signal obtained when the wind power equipment is operating normally is n1, and the amplitude of the first j frequencies of the standard characteristic frequency is p n1j , according to the formula Calculate the distance between the characteristic frequency and the standard characteristic frequency.

Images