Method and device for identifying transmission chain state of wind turbine generator set

Abstract

Disclosed is a method and device for identifying the state of a transmission chain of a wind turbine generator set. The method comprises: obtaining actual vibration data of the transmission chain of the wind turbine generator set; obtaining actual energy entropy of the actual vibration data based on the actual vibration data; obtaining actual relative entropy of the actual vibration data based on the actual vibration data and reference vibration data; and identifying the state of the transmission chain based on the actual energy entropy, the actual relative entropy, the reference energy entropy and the reference relative entropy of the reference vibration data. The present disclosure can effectively solve the problem of inaccurate transmission chain fault diagnosis in the prior art.

Description

Technical Field

[0001] The present disclosure generally relates to the field of wind power generation technology, and more specifically, to a method and device for identifying a transmission chain state of a wind power generator set. Background Art

[0002] Generally, wind turbines are installed in areas with harsh natural environments such as deserts, rivers and seas, so the failure rate of wind turbines is relatively high. As the installed capacity and power of wind turbines continue to increase, equipment maintenance and fault diagnosis have become important links to ensure the normal operation of wind turbines. The transmission system is the core component of a wind turbine, and its performance directly affects the power generation efficiency of the wind turbine.

[0003] Each wind turbine generator set is located in a different environment and has different characteristics. Therefore, each piece of vibration data of a wind turbine generator can reflect both the general characteristic information of the equipment and some personalized information of the equipment. Traditional diagnostic methods focus on the application of general rules and rely on sample data for model training. If a case with slightly different characteristics is encountered, the accuracy of model diagnosis is greatly reduced, especially for objects with minor fault information or more data interference information, which are difficult to accurately distinguish. Summary of the invention

[0004] The embodiments of the present disclosure provide a method and device for identifying the state of a transmission chain of a wind turbine generator set, which can effectively solve the problem of inaccurate transmission chain fault diagnosis in the prior art.

[0005] In a general aspect, a method for identifying the state of a transmission chain of a wind turbine generator set is provided, comprising acquiring actual vibration data of the transmission chain of the wind turbine generator set; acquiring actual energy entropy of the actual vibration data based on the actual vibration data; acquiring actual relative entropy of the actual vibration data based on the actual vibration data and reference vibration data; and identifying the state of the transmission chain based on the actual energy entropy, the actual relative entropy, the reference energy entropy and the reference relative entropy of the reference vibration data.

[0006] Optionally, based on the actual energy entropy, the actual relative entropy, the reference energy entropy and the reference relative entropy of the reference vibration data, the state of the transmission chain is identified, including: obtaining an actual fusion feature set of the actual vibration data based on the actual energy entropy and the actual relative entropy; obtaining a reference fusion feature set of the reference vibration data based on the reference energy entropy and the reference relative entropy; and identifying the state of the transmission chain based on the actual fusion feature set and the reference fusion feature set.

[0007] Optionally, when the actual vibration data includes actual vibration data of the transmission chain under N working conditions, and the reference vibration data includes reference vibration data of the transmission chain under N working conditions, the state of the transmission chain is identified based on the actual fusion feature set and the reference fusion feature set, including: identifying the state of the transmission chain based on the fusion features under each working condition in the actual fusion feature set and the fusion features under the corresponding working condition in the reference fusion feature set, wherein the actual fusion feature set includes the fusion features of the actual vibration data of the transmission chain under N working conditions, and the reference fusion feature set includes the fusion features of the reference vibration data of the transmission chain under N working conditions.

[0008] Optionally, based on the fusion features under each working condition in the actual fusion feature set and the fusion features under the corresponding working condition in the reference fusion feature set, the state of the transmission chain is identified, including: dividing the actual fusion feature set and the fusion features in the reference fusion feature set into N fusion feature sets according to N working conditions, wherein the fusion feature set is a merged set of fusion features arranged in time sequence of reference vibration data and fusion features arranged in time sequence of actual vibration data; performing linear fitting processing on the N fusion feature sets respectively; and determining the state of the transmission chain based on each fusion feature set after the linear fitting processing.

[0009] Optionally, the state of the transmission chain is determined based on each fused feature set processed after linear fitting, including: obtaining the slope of each fused feature set after linear fitting processing; when the slope of any fused feature set is greater than a first predetermined value, determining that the state of the transmission chain is a fault state and the fault is gradually deteriorating; when the slope of any fused feature set is greater than a second predetermined value and less than the first predetermined value, determining that the state of the transmission chain is a normal state, wherein the second predetermined value is less than the first predetermined value; when the slope of any fused feature set is less than the second predetermined value, determining that the state of the transmission chain is a fault state and the fault is gradually recovering.

[0010] Optionally, the fusion features in the actual fusion feature set and the reference fusion feature set are divided into N fusion feature sets according to N working conditions, including: smoothing the actual fusion feature set; obtaining the difference between all adjacent feature vectors in the actual fusion feature set after smoothing; when the sum of the differences reaches a preset threshold, sorting the differences in order from large to small; taking the maximum difference as the first difference, and iteratively performing the following operations: adding the first difference and the next difference adjacent to the first difference, and taking the result of the addition as the first difference, until the next difference adjacent to the first difference does not meet a predetermined condition, wherein the predetermined condition is that the current difference and the next difference adjacent to the current difference are both positive, both negative or both zero; when the accumulated value does not meet a preset step condition, the fusion features in the actual fusion feature set and the reference fusion feature set are divided into N fusion feature sets according to N working conditions.

[0011] Optionally, the method further includes: when the accumulated value satisfies a preset step condition, determining that a serious fault occurs in the transmission chain.

[0012] Optionally, before dividing the actual fusion feature set and the fusion features in the reference fusion feature set into N fusion feature sets according to N working conditions, the following steps are further performed: dividing the fusion features in the actual fusion feature set into feature vectors of a predetermined number of segments; obtaining the grid area of ​​each segment of the feature vector, and screening out the grid areas that meet predetermined conditions from the grid areas; and taking the feature vectors corresponding to the screened grid areas as the final actual fusion feature set.

[0013] Optionally, based on the actual vibration data, actual energy entropy of the actual vibration data is obtained, including: obtaining the root mean square of each actual vibration data in the actual vibration data; based on the root mean square, obtaining the average value of the root mean square under each working condition; obtaining the absolute difference between the root mean square of each actual vibration data under each working condition and the average value of the root mean square under the corresponding working condition; eliminating abnormal vibration data from the actual vibration data to obtain valid vibration data, wherein the abnormal vibration data is actual vibration data with an absolute difference greater than a predetermined threshold; performing time-frequency conversion on the valid vibration data to obtain a frequency domain signal of the valid vibration data; and obtaining the actual energy entropy based on the frequency domain signal.

[0014] Optionally, before obtaining the actual energy entropy based on the frequency domain signal, the method further includes: obtaining the electrical frequency of the wind turbine generator set under each working condition; and setting the amplitude corresponding to the electrical frequency in the frequency domain signal to zero.

[0015] Optionally, obtaining the electrical frequency of the wind turbine generator set under each operating condition includes: obtaining the rotational speed of the wind turbine generator set corresponding to each actual vibration data under each operating condition; multiplying the rotational speed corresponding to each actual vibration data by a predetermined coefficient corresponding to the wind turbine generator set to obtain the electrical frequency of the wind turbine generator set under each operating condition.

[0016] In another general aspect, a transmission chain state identification device for a wind turbine generator set is provided, comprising: a vibration data acquisition unit, configured to acquire actual vibration data of the transmission chain of the wind turbine generator set; an energy entropy acquisition unit, configured to acquire actual energy entropy of the actual vibration data based on the actual vibration data; a relative entropy acquisition unit, configured to acquire actual relative entropy of the actual vibration data based on the actual vibration data and reference vibration data; and a state identification unit, configured to identify the state of the transmission chain based on the actual energy entropy, the actual relative entropy, the reference energy entropy of the reference vibration data, and the reference relative entropy.

[0017] Optionally, the state identification unit is also configured to obtain an actual fusion feature set of actual vibration data based on actual energy entropy and actual relative entropy; obtain a reference fusion feature set of reference vibration data based on reference energy entropy and reference relative entropy; and identify the state of the transmission chain based on the actual fusion feature set and the reference fusion feature set.

[0018] Optionally, when the actual vibration data includes actual vibration data of the transmission chain under N working conditions and the reference vibration data includes reference vibration data of the transmission chain under N working conditions, the state identification unit is further configured to identify the state of the transmission chain based on the fusion features under each working condition in the actual fusion feature set and the fusion features under the corresponding working condition in the reference fusion feature set, wherein the actual fusion feature set includes the fusion features of the actual vibration data of the transmission chain under N working conditions and the reference fusion feature set includes the fusion features of the reference vibration data of the transmission chain under N working conditions.

[0019] Optionally, the state recognition unit is further configured to divide the fused features in the actual fused feature set and the reference fused feature set into N fused feature sets according to N working conditions, wherein the fused feature set is a merged set of fused features arranged in time order of reference vibration data and fused features arranged in time order of actual vibration data; perform linear fitting processing on the N fused feature sets respectively; and determine the state of the transmission chain based on each fused feature set after the linear fitting processing.

[0020] Optionally, the state identification unit is further configured to obtain the slope of each fused feature set after linear fitting processing; when the slope of any fused feature set is greater than a first predetermined value, it is determined that the state of the transmission chain is a fault state and the fault is gradually deteriorating; when the slope of any fused feature set is greater than a second predetermined value and less than the first predetermined value, it is determined that the state of the transmission chain is a normal state, wherein the second predetermined value is less than the first predetermined value; when the slope of any fused feature set is less than the second predetermined value, it is determined that the state of the transmission chain is a fault state and the fault is gradually recovering.

[0021] Optionally, the state recognition unit is also configured to perform smoothing on the actual fusion feature set; obtain the difference between all adjacent feature vectors in the actual fusion feature set after smoothing; when the sum of the differences reaches a preset threshold, sort the differences in order from large to small; take the maximum difference as the first difference, and iteratively perform the following operations: add the first difference and the next difference adjacent to the first difference, and take the result of the addition as the first difference, until the next difference adjacent to the first difference does not meet a predetermined condition, wherein the predetermined condition is that the current difference and the next difference adjacent to the current difference are both positive, both negative or both zero; when the accumulated value does not meet a preset step condition, divide the fusion features in the actual fusion feature set and the reference fusion feature set into N fusion feature sets according to N working conditions.

[0022] Optionally, the device further comprises: a fault determination unit configured to determine that a serious fault occurs in the transmission chain when the accumulated value satisfies a preset step condition.

[0023] Optionally, the state recognition unit is further configured to divide the fusion features in the actual fusion feature set into feature vectors of a predetermined number of segments; obtain the grid area of ​​each segment of the feature vector, and filter out the grid areas that meet predetermined conditions from the grid areas; and use the feature vectors corresponding to the filtered grid areas as the final actual fusion feature set.

[0024] Optionally, the energy entropy acquisition unit is further configured to acquire the root mean square of each actual vibration data in the actual vibration data; based on the root mean square, acquire the average value of the root mean square under each working condition; acquire the absolute difference between the root mean square of each actual vibration data under each working condition and the average value of the root mean square under the corresponding working condition; eliminate abnormal vibration data from the actual vibration data to obtain valid vibration data, wherein the abnormal vibration data is actual vibration data whose absolute difference is greater than a predetermined threshold; perform time-frequency domain conversion on the valid vibration data to obtain a frequency domain signal of the valid vibration data; and acquire the actual energy entropy based on the frequency domain signal.

[0025] Optionally, the energy entropy acquisition unit is further configured to acquire the electrical frequency of the wind turbine generator set under each operating condition before acquiring the actual energy entropy based on the frequency domain signal; and set the amplitude corresponding to the electrical frequency in the frequency domain signal to zero.

[0026] Optionally, the energy entropy acquisition unit is also configured to obtain the rotational speed of the wind turbine generator set corresponding to each actual vibration data under each operating condition; multiply the rotational speed corresponding to each actual vibration data by a predetermined coefficient corresponding to the wind turbine generator set to obtain the electrical frequency of the wind turbine generator set under each operating condition.

[0027] In another general aspect, a computer-readable storage medium storing instructions is provided, wherein when the instructions are executed by at least one computing device, the at least one computing device is prompted to execute any of the above-mentioned methods for identifying the state of a transmission chain of a wind turbine generator set.

[0028] In another general aspect, a system is provided comprising at least one computing device and at least one storage device storing instructions, wherein the instructions, when executed by the at least one computing device, cause the at least one computing device to execute any of the above-described methods for identifying a transmission chain state of a wind turbine generator set.

[0029] According to the method and device for identifying the state of the transmission chain of a wind turbine generator set according to the embodiment of the present disclosure, the state of the transmission chain is identified based on the energy entropy and relative entropy of the actual vibration data and the energy entropy and relative entropy of the reference vibration data, that is, when identifying the state of the transmission chain, the energy entropy and relative entropy of the vibration data are considered at the same time, and various states of the transmission chain, such as the fault state, can be accurately identified. Therefore, through the present disclosure, the problem of inaccurate transmission chain fault diagnosis in the prior art can be effectively solved.

[0030] Additional aspects and / or advantages of the present general inventive concept will be set forth in part in the following description and in part will be apparent from the description or may be learned through practice of the present general inventive concept. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The above and other objects and features of the embodiments of the present disclosure will become more apparent through the following description in conjunction with the accompanying drawings showing the embodiments, in which:

[0032] Figure 1 is a flow chart showing a method for identifying a transmission chain state of a wind turbine generator set according to an embodiment of the present disclosure;

[0033] Figure 2 is a schematic diagram showing the overall flow of a method for identifying a state of a transmission chain of a wind turbine generator set according to an embodiment of the present disclosure;

[0034] Figure 3 is a schematic diagram showing a trend analysis process of an embodiment of the present disclosure;

[0035] Figure 4 is a schematic diagram showing a step analysis process of an embodiment of the present disclosure;

[0036] Figure 5 is a schematic diagram showing effective values ​​of original vibration data according to an embodiment of the present disclosure;

[0037] Figure 6 is a schematic diagram showing effective values ​​of raw vibration data after screening according to an embodiment of the present disclosure;

[0038] Figure 7 is a schematic diagram showing a comparison of frequency spectra of fault vibration data and normal vibration data according to an embodiment of the present disclosure;

[0039] Figure 8 is a schematic diagram showing the comparison of analysis results of various indicators of an embodiment of the present disclosure;

[0040] Fig. 9 is a schematic diagram showing a comparison before and after cleaning of a fusion feature of an embodiment of the present disclosure;

[0041] Fig.10is a schematic diagram showing the fusion feature after smoothing processing of an embodiment of the present disclosure;

[0042] Fig.11 is a schematic diagram showing a spectrum of fault vibration data of a yaw system according to an embodiment of the present disclosure;

[0043] Fig.12 is a schematic diagram showing energy entropy of fault vibration data of a yaw system according to an embodiment of the present disclosure;

[0044] Fig.13 is a schematic diagram showing the fusion characteristics of fault vibration data of a yaw system according to an embodiment of the present disclosure;

[0045] Fig.14 is a schematic diagram showing a spectrum of fault vibration data when a bearing moves according to an embodiment of the present disclosure;

[0046] Fig.15 is a schematic diagram showing effective values ​​of fault vibration data when the bearing moves according to an embodiment of the present disclosure;

[0047] Fig.16 is a schematic diagram showing the fusion characteristics of fault vibration data when the bearing moves according to an embodiment of the present disclosure;

[0048] Fig.17 is a schematic diagram showing data trend analysis when bearings move according to an embodiment of the present disclosure;

[0049] Fig.18 is a block diagram showing a transmission chain state identification device of a wind turbine generator set disclosed in the present invention. DETAILED DESCRIPTION

[0050] The following specific embodiments are provided to help the reader obtain a comprehensive understanding of the methods, devices and / or systems described herein. However, after understanding the disclosure of the present application, various changes, modifications and equivalents of the methods, devices and / or systems described herein will be clear. For example, the order of operations described herein is only an example and is not limited to those orders set forth herein, but can be changed as will be clear after understanding the disclosure of the present application, except for operations that must occur in a specific order. In addition, for greater clarity and simplicity, the description of features known in the art may be omitted.

[0051] The features described herein can be implemented in different forms and should not be construed as being limited to the examples described herein. Rather, the examples described herein have been provided to illustrate only some of the many possible ways to implement the methods, devices, and / or systems described herein, which will be clear after understanding the disclosure of the present application.

[0052] As used herein, the term "and / or" includes any one of the associated listed items and any combination of any two or more.

[0053] Although terms such as "first", "second", and "third" may be used herein to describe various members, components, regions, layers, or portions, these members, components, regions, layers, or portions should not be limited by these terms. Instead, these terms are only used to distinguish one member, component, region, layer, or portion from another member, component, region, layer, or portion. Therefore, without departing from the teachings of the examples described herein, the first member, first component, first region, first layer, or first portion referred to in the examples may also be referred to as the second member, second component, second region, second layer, or second portion.

[0054] The terms used herein are only used to describe various examples and are not intended to limit the disclosure. Unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. The terms "comprise", "include" and "have" indicate the presence of the described features, quantities, operations, components, elements and / or combinations thereof, but do not exclude the presence or addition of one or more other features, quantities, operations, components, elements and / or combinations thereof.

[0055] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by a person of ordinary skill in the art to which the present disclosure belongs after understanding the present disclosure. Unless explicitly defined as such herein, terms (such as those defined in a general dictionary) should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and the present disclosure, and should not be interpreted in an idealized or overly formal manner.

[0056] Furthermore, in the description of examples, when it is considered that a detailed description of a well-known related structure or function would cause vague interpretation of the present disclosure, such a detailed description will be omitted.

[0057] The present disclosure provides a method and device for identifying the state of a transmission chain of a wind turbine generator set, which can solve the above-mentioned problem. The method for identifying the state of a transmission chain of a wind turbine generator set disclosed in the present disclosure can be applied to a server, and the server and the wind turbine generator set can be connected wirelessly or by wire, which is not limited here. The above-mentioned server can be a server, or a server cluster composed of several servers, or a cloud computing platform or a virtualization center. The following is an explanation using a server as an example.

[0058] The server obtains actual vibration data of the transmission chain of the wind turbine generator set, and obtains actual energy entropy of the actual vibration data based on the actual vibration data, and then obtains actual relative entropy of the actual vibration data based on the actual vibration data and the reference vibration data, and then identifies the state of the transmission chain based on the actual energy entropy, the actual relative entropy, the reference energy entropy and the reference relative entropy of the reference vibration data, so that when identifying the state of the transmission chain, the energy entropy and relative entropy of the vibration data are taken into consideration at the same time, so that various states of the transmission chain, such as a fault state, can be accurately identified.

[0059] The present disclosure is described in detail below with reference to the accompanying drawings.

[0060] The present invention proposes a method for identifying the state of a transmission chain of a wind turbine generator set. Figure 1 1 is a flow chart showing a method for identifying a transmission chain state of a wind turbine generator set according to an embodiment of the present disclosure. Figure 1 The method for identifying the state of the transmission chain of a wind turbine generator set comprises the following steps:

[0061] In step S101, actual vibration data of a transmission chain of a wind turbine generator set is obtained.

[0062] The actual vibration data obtained may be vibration data of the wind turbine generator set during a predetermined operation time, for example, vibration data of the wind turbine generator set during half a month of operation, or vibration data of the wind turbine generator set during one month of operation, which is not limited in the embodiments of the present disclosure.

[0063] The above-mentioned actual vibration data refers to the vibration data of the wind turbine generator set during its actual operation.

[0064] In step S102 , actual energy entropy of the actual vibration data is acquired based on the actual vibration data.

[0065] Before obtaining the actual energy entropy of the actual vibration data, the actual vibration data may be first converted into a time-frequency domain to obtain a frequency domain signal of the actual vibration data, and then the energy entropy of the actual vibration data may be obtained based on the frequency domain signal of the actual vibration data.

[0066] Furthermore, due to the complex and changeable on-site environment of wind turbines, the actual vibration data obtained often contains some invalid data, which affects the analysis and feature extraction of the actual vibration data. Therefore, the actual vibration data can be cleaned before obtaining the actual energy entropy. For example, the actual vibration data can be processed by the Laida criterion (also known as the "3σ criterion") to effectively eliminate the gross errors in the actual vibration data; based on the inherent electrical frequency of the wind turbine, some actual vibration data can be eliminated to effectively avoid the interference caused by the electrical frequency, laying the foundation for the analysis of the actual vibration data.

[0067] According to an embodiment of the present disclosure, before obtaining the actual energy entropy of the actual vibration data, the actual vibration data can be processed by the Laida criterion (also known as the "3σ criterion"), and then the time-frequency domain conversion is performed, such as: obtaining the root mean square of each actual vibration data in the actual vibration data, based on the root mean square, obtaining the average value of the root mean square under each working condition, obtaining the absolute difference between the root mean square of each actual vibration data under each working condition and the average value of the root mean square under the corresponding working condition, eliminating abnormal vibration data from the actual vibration data to obtain effective vibration data, performing time-frequency domain conversion on the effective vibration data to obtain the frequency domain signal of the effective vibration data, and obtaining the actual energy entropy based on the frequency domain signal.

[0068] Here, the abnormal vibration data is the actual vibration data whose absolute difference is greater than a predetermined threshold. According to this embodiment, invalid data in the actual vibration data is eliminated by RMS, and energy entropy is obtained based on the eliminated actual vibration data, which can improve the accuracy of subsequent identification of the transmission chain state.

[0069] For example, taking the actual vibration data under a certain working condition i as an example to remove invalid data, the root mean square set of each actual vibration data is RMS i = {rms i1 ,rms i2... ,rms in ,…rms ir}. is the average value of all RMS values ​​under this condition, and σ is the standard deviation of all RMS values ​​under this condition. The vibration data is qualified if it meets the following conditions:

[0070]

[0071] In the formula, i = 1, 2, ... k, represents the i-th working condition; j = 1, 2, ... r, represents the number of actual vibration data. It should be noted that in general, the "3σ criterion" is applicable to objects with a large sample size, so the 3σ criterion can be used to eliminate outliers in the root mean square rms under the same working condition.

[0072] At present, each wind turbine has its own unique frequency characteristics. Taking a direct-drive wind turbine as an example, the actual vibration data spectrum of a normal wind turbine contains a large number of electrical frequencies and its frequency amplitude is unstable, which causes great interference to the energy comparison of the signal power spectrum before and after the fault. When the speed is known, these electrical frequencies can be calculated by the parameters of the wind turbine. Therefore, the electrical frequency can be obtained by the parameters of the wind turbine and the electrical frequency can be eliminated, thereby effectively avoiding the interference caused by the electrical frequency and laying the foundation for vibration data analysis. For example, before obtaining the actual energy entropy based on the frequency domain signal, the electrical frequency of the wind turbine under each working condition can be obtained, and the amplitude corresponding to the electrical frequency in the frequency domain signal is set to zero. Here, obtaining the electrical frequency of the wind turbine under each working condition can be achieved by the following method: obtaining the speed of the wind turbine corresponding to each actual vibration data under each working condition, multiplying the speed corresponding to each actual vibration data by the predetermined coefficient corresponding to the wind turbine, and obtaining the electrical frequency of the wind turbine under each working condition.

[0073] For example, the specific method of obtaining the electrical frequency can be as follows: when the speed S of the wind turbine generator set is known, the electrical frequency corresponding to the actual vibration data under the same working condition is F ij :

[0074]

[0075]

[0076] f ij =a*s i (4)

[0077] Wherein, i=1,2,…k represents k kinds of working conditions, j=1,2,…v represents the number of electrical frequencies corresponding to each working condition, a represents electrical parameters, different electrical frequencies correspond to different parameters, and different wind turbines have their own electrical parameters.

[0078] After obtaining the electrical frequency, select the actual vibration data under a working condition and convert the actual vibration data from the time domain signal to the frequency domain signal:

[0079] Y={Y 11 Y 12 … Y 1r}

[0080] Y ij is the frequency domain amplitude sequence of each frequency domain signal, and the frequency domain signal Y ij The amplitude corresponding to the identified electrical frequency is replaced with 0 to obtain Y after removing the electrical frequency. ij ={y 1, y2,…yu}, where i = 1, 2, ... k; j = 1, 2, ... r; u is the number of spectrum data points with non-zero amplitude in the frequency domain signal.

[0081] After cleaning the actual vibration data, the energy entropy of the actual vibration data can be obtained by the following method: ij Divide into m segments and calculate the energy entropy H of each frequency domain signal ij m , the energy entropy matrix H corresponding to the frequency domain matrix Y is:

[0082]

[0083]

[0084]

[0085] Where [a, b] is the range of points corresponding to the mth frequency segment, 0 <a<b≤u,y p Represents the amplitude of the p-th point of the frequency domain signal.

[0086] return Figure 1 In step S103, the actual relative entropy of the actual vibration data is obtained based on the actual vibration data and the reference vibration data. When obtaining the actual relative entropy, reference vibration data can be selected from the standard library according to the actual needs of identifying the state of the transmission chain, and then the relative entropy index of each actual vibration data and the reference vibration data is analyzed. Here, the higher the similarity between the actual vibration data and the reference vibration data, the smaller the relative entropy.

[0087] For example, taking the frequency domain matrix Y under the same working condition as the research object, the relative entropy KL of the actual vibration data and the reference vibration data is calculated. ij , forming the relative entropy matrix KL:

[0088]

[0089]

[0090]

[0091] in, kl represents the relative entropy mean of the mth frequency band of the jth data in the actual vibration data and the 1st to nth data in the reference vibration data. j1 m It represents the relative entropy between the jth data in the actual vibration data and the mth frequency band of the first reference vibration data. represents the amplitude sequence of the mth frequency band of the i-th reference vibration data, Represents the amplitude sequence of the mth frequency band of the jth data in the actual vibration data.

[0092] return Figure 1 In step S104, the state of the transmission chain is identified based on the actual energy entropy, the actual relative entropy, the reference energy entropy and the reference relative entropy of the reference vibration data.

[0093] In this step, the energy entropy and relative entropy of the reference vibration data can be obtained in advance. The specific acquisition process is similar to the energy entropy and relative entropy of the actual vibration data, which will not be discussed here. It should be noted that when obtaining the relative entropy of the reference vibration data, the reference vibration data can be compared with itself to obtain the relative entropy of the reference vibration data.

[0094] According to an embodiment of the present disclosure, based on actual energy entropy, actual relative entropy, reference energy entropy and reference relative entropy of reference vibration data, identifying the state of the transmission chain can be achieved in the following manner: based on actual energy entropy and actual relative entropy, obtaining an actual fusion feature set of actual vibration data; based on reference energy entropy and reference relative entropy, obtaining a reference fusion feature set of reference vibration data; based on the actual fusion feature set and the reference fusion feature set, identifying the state of the transmission chain. According to this embodiment, a fusion feature including energy entropy and relative entropy is obtained, which can facilitate the subsequent identification of the state of the transmission chain.

[0095] It should be noted that energy entropy reflects the amount of information in the data within the corresponding frequency band, which is an absolute indicator, while relative entropy reflects the similarity between the current data and the reference data, which is a relative indicator. According to the vibration characteristics of the transmission chain, when a device fails, a fault frequency will appear in the spectrum, and the energy entropy will increase accordingly. It is particularly effective for data with obvious changes in the amplitude of the fault frequency. When the corresponding frequency band of the fault vibration data is compared with the reference vibration data, it will be found that the data similarity decreases. Therefore, it is more accurate to identify the state of the transmission chain by combining energy entropy and relative entropy.

[0096] Specifically, the fusion matrix can be obtained as follows:

[0097] T=aH m×r ×b(1-KL m×r ) (11)

[0098] Where a and b are the weight ratios of the two indicators. The real-time vibration data collected in real time is compared with the reference vibration data of the same working condition to calculate the fusion feature T. Based on the change trend of T over time, the state of the transmission chain can be identified. For example, when T gradually increases over time, it means that the equipment may have a fault and the fault is deteriorating.

[0099] According to an embodiment of the present disclosure, actual vibration data may include actual vibration data of the transmission chain under N working conditions, and reference vibration data may also include reference vibration data of the transmission chain under N working conditions. At this time, based on the actual fusion feature set and the reference fusion feature set, identification of the state of the transmission chain can be performed according to N working conditions, such as: based on the fusion features under each working condition in the actual fusion feature set and the fusion features under the corresponding working condition in the reference fusion feature set, identifying the state of the transmission chain, wherein the actual fusion feature set includes the fusion features of the actual vibration data of the transmission chain under N working conditions, and the reference fusion feature set includes the fusion features of the reference vibration data of the transmission chain under N working conditions.

[0100] According to an embodiment of the present disclosure, the acquired fusion features can be subjected to linear fitting processing to simplify the difficulty of identifying the state of the transmission chain. For example, based on the fusion features under each working condition in the actual fusion feature set and the fusion features under the corresponding working condition in the reference fusion feature set, the specific process of identifying the state of the transmission chain can be achieved by, but not limited to, the following methods: dividing the fusion features in the actual fusion feature set and the reference fusion feature set into N fusion feature sets according to N working conditions, wherein the fusion feature set is a combined set of fusion features arranged in the time sequence of the reference vibration data and fusion features arranged in the time sequence of the actual vibration data, performing linear fitting processing on each of the N fusion feature sets, and determining the state of the transmission chain based on each fusion feature set after the linear fitting processing. According to this embodiment, the difficulty of identification can be simplified by performing linear fitting processing on the fusion feature set.

[0101] According to an embodiment of the present disclosure, the state of the transmission chain is determined based on each fused feature set after linear fitting processing. The state of the transmission chain can be determined by the slope after linear fitting. For example, the slope of each fused feature set after linear fitting processing is first obtained. When the slope of any fused feature set is greater than a first predetermined value, the state of the transmission chain is determined to be a fault state and the fault is gradually deteriorating; when the slope of any fused feature set is greater than a second predetermined value and less than the first predetermined value, the state of the transmission chain is determined to be a normal state, wherein the second predetermined value is less than the first predetermined value; when the slope of any fused feature set is less than the second predetermined value, the state of the transmission chain is determined to be a fault state and the fault is gradually recovering. By determining the changing trend of the fused feature through the slope of the fused feature set after linear fitting, the state of the transmission chain can be determined quickly and conveniently. It should be noted that the above-mentioned predetermined values ​​are all set according to actual needs.

[0102] Specifically, after the linear fitting process, different levels of thresholds can be set in combination with the slope of the fused feature set after linear fitting to judge the changing trend of the fused features. For example, when the slope of any fused feature set is greater than a first predetermined value, the state of the transmission chain is determined to be a fault state and the fault is gradually deteriorating; when the slope of any fused feature set is greater than a second predetermined value and less than the first predetermined value, the state of the transmission chain is determined to be a normal state, wherein the second predetermined value is less than the first predetermined value; when the slope of any fused feature set is less than the second predetermined value, the state of the transmission chain is determined to be a fault state and the fault is gradually recovering.

[0103] Optionally, after determining the state of the transmission chain, the corresponding flag value can be set to display to the user, such as setting flag = 0, indicating that the fusion feature trend is stable; flag = 1, indicating that the fusion feature trend is slowly rising; flag = 2, indicating that the fusion feature trend has a step increase; flag = -2, indicating that the fusion feature trend has a step decrease; flag = -1, indicating that the fusion feature trend is slowly decreasing.

[0104] According to the embodiments of the present disclosure, step analysis can be performed before linear fitting to avoid the influence of step changes on the recognition results. This is because, under the current ideal situation, as the fault of the wind turbine generator set gradually deteriorates, its fusion feature gradually increases over time, and the use of linear fitting can accurately analyze the changing trend of the health status of the wind turbine generator set. However, the operating environment of the wind turbine generator set is harsh and the working conditions are complex and changeable. Therefore, for some sudden or periodic abnormal vibrations, it is difficult to accurately analyze them by directly using linear fitting, and it may even cause system misjudgment. This situation is reflected in the fusion feature T, that is, a step change occurs, which needs to be identified separately.

[0105] Specifically, the above-mentioned division of the fusion features in the actual fusion feature set and the reference fusion feature set into N fusion feature sets according to N working conditions can be achieved in the following manner: smoothing the actual fusion feature set; obtaining the difference between all adjacent feature vectors in the actual fusion feature set after smoothing; when the sum of the differences reaches a preset threshold, sorting the differences in order from large to small; taking the maximum difference as the first difference, iteratively performing the following operations: adding the first difference and the next difference adjacent to the first difference, and taking the result of the addition as the first difference, until the next difference adjacent to the first difference does not meet a predetermined condition, wherein the predetermined condition is that the current difference and the next difference adjacent to the current difference are both positive, both negative or both zero; when the accumulated value does not meet a preset step condition, dividing the fusion features in the actual fusion feature set and the reference fusion feature set into N fusion feature sets according to N working conditions; when the accumulated value meets the preset step condition, determining that a serious fault occurs in the transmission chain.

[0106] According to an embodiment of the present disclosure, a step analysis is performed before linear fitting. Only when there is no step change, linear fitting and subsequent identification processing are carried out. When there is a step change, a fault is directly reported, which can improve the accuracy of identification.

[0107] For example, the step analysis can be implemented in the following way. The fused feature is equally divided into new_num segments, and the moving average method is used to smooth the fused feature T′:

[0108]

[0109] where i = 1, 2, … new_num, representing the i-th smoothing point. It should be noted that the fused feature for smoothing can be the fused feature after removing some data based on the network area, or the fused feature without removing some data based on the network area. In the above formula, T′ takes the fused feature after removing some data based on the network area. The specific removal of some data based on the network area will be discussed in detail later and will not be elaborated here for the time being.

[0110] The characteristic of a step change is that the value changes suddenly, that is, the difference between two adjacent values suddenly becomes large. Therefore, the difference method is used to calculate the difference between two adjacent feature vectors in the actually smoothed fused feature set, that is, the relative change value delt of Smooth and the sum total of its change amounts:

[0111] delt(i) = Smooth(i + 1) - Smooth(i) (13)

[0112] total = ∑delt (14)

[0113] where i = 1, 2, … new_num - 1. total can reflect the cumulative change amount of the fused feature vector. A threshold thre_total is set. When total meets the threshold condition, delt is sorted from large to small, and the larger value of delt is found. It is judged whether the adjacent delt(i) has the same sign. If the signs are the same, it indicates that the change trend of the previous group of points is the same as that of the current point, and the accumulation is carried out until the condition is not met, that is, the adjacent delt(i) has different signs.

[0114]

[0115] where 1 < i ≤ m < new_num - 1. When i = 1 or m = new_num - 1, condition (3) or (4) is discarded.

[0116] According to the calculated Accu delt , judge whether it meets the following conditions:

[0117]

[0118] If the above conditions are met, it is determined that the data for this period of time has a step change.

[0119] According to an embodiment of the present disclosure, before dividing the actual fusion feature set and the fusion features in the reference fusion feature set into N fusion feature sets according to N working conditions, the fusion features in the actual fusion feature set can be equally divided into feature vectors of a predetermined number of segments, the grid area of ​​each segment of the feature vector is obtained, and the grid areas that meet predetermined conditions are screened out from the grid areas, and the feature vectors corresponding to the screened grid areas are used as the final actual fusion feature set.

[0120] Here, the grid area is the product of the eigenvalue difference and the number of actual vibration data in each segment of the eigenvector, and the eigenvalue difference is the difference between the maximum eigenvalue and the minimum eigenvalue in each segment of the eigenvector. According to this embodiment, some features in the partial fusion features are eliminated based on the grid area to further ensure the accuracy of recognition.

[0121] For example, some data can be removed based on the grid area in the following way: Analyze the change trend of the fusion feature T over time for a single wind turbine. Assume that a wind turbine has N data in a month, calculate the corresponding fusion feature T, divide the fusion feature into num segments, and each segment fusion feature vector is T i , calculate the grid area S of each segment of fusion features i _area, that is:

[0122]

[0123] Where i = 1, 2, ... num, S i _area represents the area of ​​the i-th grid. Area S i The smaller the _area is, the denser the fusion feature points are, which indicates that the stability of the vibration data is better. i _area is sorted from large to small, assuming S i _area follows a normal distribution, and the t-test criteria are used to calculate from the larger area in turn, and the grid areas that do not meet the conditions are screened out. If the current grid area meets the conditions, the remaining smaller grid areas will not be screened. The specific judgment conditions are as follows:

[0124]

[0125]

[0126]

[0127] When the area S i_area does not meet the t-test criteria, indicating that the interval fusion feature vector T i The discreteness is large, and there may be individual abnormal points, so it is necessary to i Perform a secondary test and set T i ={T i1 ,T i2 …,T iG}. is the mean, G is the number of fused feature medians, σ iT is the standard deviation of the group of fused feature vectors, and the data is qualified if it meets the following conditions:

[0128]

[0129]

[0130]

[0131] After two t-test criterion processes, the outliers in the data have been eliminated, and a new feature vector T' can be obtained with a length of N'. Smoothing is then performed based on the obtained new feature vector T'.

[0132] The following is based on Figures 2 to 4 The above embodiments are systematically described as follows: Figure 2 As shown in the figure, after obtaining the real-time vibration data, the data can be preprocessed first, such as eliminating some invalid vibration data through the "3σ criterion", and then the data can be binned according to the working conditions to extract the natural frequency of the wind turbine generator set. For example, the amplitude corresponding to the electrical frequency of the wind turbine generator set in the frequency domain signal of the pre-processed vibration data is set to 0, and then the frequency band is divided and analyzed, the energy entropy is calculated, and the relative entropy is calculated, so as to obtain the corresponding fusion matrix based on the calculated energy entropy and relative entropy, and the trend analysis is performed based on the fusion matrix.

[0133] Here, the trend analysis is as follows Figure 3 As shown in , the grid area of ​​the features in the actual fusion feature set can be checked first, and then when the grid area is large, outliers can be removed, that is, features with large grid areas can be removed, and then data smoothing and step analysis can be performed (the specific process of step analysis is shown in Figure 4 As described above, since the above embodiment has been discussed in detail, it will not be discussed here in detail), based on the step analysis result, linear analysis is performed, and then fault diagnosis is performed based on the linear analysis result.

[0134] In order to verify the advantages of using fusion features for fault analysis in the embodiment of the present disclosure, taking a certain wind farm as an example, Wasserstein distance W, energy entropy H, cosine similarity C, fusion feature T, etc. are selected as indicators to analyze and compare the vibration data of the faulty wind turbine before and after the fault.

[0135] In order to improve the power generation efficiency of the wind turbines in the wind farm, the major components of the wind turbines are inspected and replaced. The vibration data before and after the replacement are collected. The first 58 items are the vibration data before the replacement, of which items 49 to 57 are serious fault degradation data, and the remaining data are minor fault data; items 59 to 185 are the vibration data after the replacement (the following are all represented by normal vibration data, which is equivalent to the reference vibration data in the above embodiment). At this time, the rotation speed is about 1200rpm, and the effective value of the original vibration data (that is, the root mean square RMS of the original vibration data) is as follows Figure 5 As shown in the figure, we can see that there are some outliers, which need to be eliminated using the 3σ criterion. The effective value of the original vibration data after elimination is as follows: Figure 6 shown.

[0136] The fault vibration data after cleaning is selected for feature comparison with the normal vibration data. The spectrum comparison is as follows: Figure 7 As shown in the figure, the gearbox meshing frequency of normal vibration data is the dominant frequency of the spectrum, but when the rear bearing of the low-speed shaft is worn, the meshing frequency energy decreases, the fault frequency and the frequency multiple amplitude increase, which will mask the meshing frequency. At this time, if only the energy entropy value is used for distinction, the change may not be obvious, because the gearbox meshing frequency and the fault frequency energy tend to increase and decrease. At this time, the fusion feature can be used to compare the similarity of the vibration data, which has a good effect. Calculate the fusion feature indicators of multiple normal vibration data and fault vibration data, and some results are shown in Table 1:

[0137] Table 1 Comparison of fusion indicators

[0138] Fault data Normal data 0.493257 0.112967 0.448658 0.030813 0.489174 0.021394 0.487662 0.100153 0.560322 0.054345 0.549329 0.045879 0.481405 0.089338 0.485379 0.129294 0.471101 0.055647

[0139] Note: The bold data are Figure 3 Two data feature values.

[0140] The fusion matrix of the filtered vibration data is calculated. In order to unify the measurement standard, each index is normalized and the relative entropy of the actual vibration data is calculated based on the normal vibration data. It should be noted that if the above index values ​​are relatively small, it means that the data is normal, while if the index values ​​are relatively large, it means that the unit data may be faulty. The specific analysis results are as follows: Figure 8 As shown, Figure 8 (A) is a schematic diagram of the analysis results of the vibration data of the faulty wind turbine using the Wasserstein distance W. Figure 8(B) is a schematic diagram of the analysis results of the vibration data of the faulty wind turbine generator set using energy entropy H. Figure 8 (C) is a schematic diagram of the analysis results of the vibration data of the faulty wind turbine generator set using cosine similarity C. Figure 8 (D) is a schematic diagram of the analysis results of the vibration data of a faulty wind turbine using the fusion feature T.

[0141] Combination Figure 8 As shown in Table 2, the cosine similarity C indicator is completely unable to distinguish between the normal or faulty state of the unit. The W distance or energy entropy H indicator can identify serious faults of the unit, but the indicator value under normal conditions fluctuates too much and cannot distinguish between minor faults and normal conditions. The fusion feature T indicator has good discrimination for different states of the unit. Table 2 lists the standard deviation of the analysis results of each indicator of normal vibration data. It can be seen that the fusion feature T indicator has a small judgment deviation for vibration data under the same state, and the indicator has good stability.

[0142] Table 2 Standard deviation of characteristic values ​​of normal vibration data

[0143] Performance Indicators H W C T Standard Deviation 0.1431 0.2131 0.2203 0.0916

[0144] In addition, the present disclosure also verifies the trend analysis in the above embodiment. In actual industrial applications, the centralized monitoring system of wind power plants needs to have the function of automatic fault trend analysis and alarm. By observing the changes in the above fusion feature T, it is found that there are still individual outliers in the normal state, which affects the accuracy of data trend analysis. Therefore, the grid area analysis method can be used for cleaning. The fusion feature of vibration data before and after cleaning is compared. Fig. 9 As shown in the figure, we can see that the outliers in the vibration data are successfully removed after cleaning. Then, the vibration data after cleaning is smoothed, and the effect of the processed vibration data is shown in the figure below. Fig.10 According to the smoothed curve of the vibration data, it can be seen that there is a sudden change in this section of vibration data, which first rises and then falls. The step analysis method is used to obtain flag = (2, -2), which is consistent with Figure 7 The situations are consistent, indicating that the trend analysis algorithm of the above embodiment is accurate.

[0145] To further verify the trend analysis algorithm, vibration data of 5 faulty units were collected, and fusion feature calculation was performed to obtain flag bits, as shown in Table 3 below, which also shows that the trend analysis algorithm of the above embodiment is accurate.

[0146] Table 3 Analysis results of the trend of different units

[0147] Serial Number Changing Trends flag 1 Step up 2 2 smooth 0 3 Two-step rise (2，2) 4 Rise first then fall (2，-2) 5 Slight rise 1

[0148] Therefore, when the wind turbine generator set is officially put into operation, the present disclosure can monitor and calculate the fusion feature T of the vibration data at each position of the transmission chain in real time, and analyze the change trend of T according to the working conditions; when wear occurs at a certain point of the transmission chain, the vibration data of the corresponding measuring point will change. Using the method disclosed in the present disclosure, if it is found that the fusion feature T trend of the vibration data of a certain measuring point rises slightly, the system will issue an early warning at this time. If the fusion feature T trend of the vibration data continues to rise, an alarm message will be issued to facilitate on-site maintenance personnel to investigate and deal with the problem as soon as possible to ensure the normal and stable operation of the wind turbine.

[0149] The above-mentioned embodiment is verified again from the following aspects, respectively, of specific fault feature identification and bearing movement problem diagnosis.

[0150] (1) Specific fault feature identification

[0151] During the inspection of a wind turbine generator in a wind farm, it was found that there was a problem with the yaw system brake. At this time, the frequency spectrum of the vibration data at the bearing measurement point was as follows: Fig.11 As shown. Select the following under different working conditions when the fan is operating normally: Fig.11 The characteristic vibration data shown is used as the standard vibration data. At this time, the standard vibration data can be used to replace the normal vibration data above to calculate the energy entropy of the actual vibration data. The real-time vibration data of the collected wind turbine is analyzed and calculated. It is found that the vibration data of the wind farm unit is mixed with similar characteristic data. A section of data is selected for analysis, among which the 14th, 25th, 27th, 75th to 91st data have the characteristics of yaw fault. The energy entropy calculation results are shown in Fig.12 As shown, the above data is analyzed and found that the frequency domain entropy value of normal vibration data is low. The data with entropy value>5 is verified and it is found that most yaw fault features can be identified, but the 29th, 64th and 66th data do not have yaw system fault features. Using the energy entropy index will cause misjudgment. Therefore, the fusion feature T proposed in this disclosure is used for calculation, and the results are as follows Fig.13 As shown:

[0152] contrast Fig.12 and Fig.13 It is found that the fusion feature T can accurately distinguish the fault characteristics of the yaw system and is not interfered by other abnormal data. In practical applications, data can be diagnosed in real time and an alarm can be issued.

[0153] (2) Diagnosis of bearing movement problems

[0154] The bearings of a certain type of unit often move, causing interference between the inner ring of the bearing and the rolling element, and eventually the inner ring of the bearing is crushed, which in turn causes the generator to fail and get stuck, causing a series of secondary damage to the wind turbine. By observing the existing fault samples, it is found that the fault has no obvious impact on the time domain signal in both the early and late stages. In the frequency domain, there is a characteristic frequency of bearing fault in the low frequency band in the early stage, and a "haystack" phenomenon appears in the late stage, and the frequency band distribution range gradually increases. The frequency spectrum characteristics of the vibration data at this time are as follows: Fig.14 shown.

[0155] In order to detect such faults as early as possible, the fusion characteristic index T disclosed in this disclosure is used to analyze the unit spectrum [50Hz~200Hz] data. First, the vibration data of the unit is selected, which includes the data of the unit bearing from normal, early movement to complete failure. Fig.15 is the effective value corresponding to the fault vibration data when the bearing moves (i.e., the root mean square of the fault vibration data). It can be seen that the change of the bearing cannot be clearly monitored from the effective value. Therefore, the weights of the energy entropy and relative entropy are set to a = 0.8 and b = 0.2 respectively, and the fusion characteristic index T of the vibration data during this period is calculated, as follows: Fig.16 As shown. Using trend analysis method to do analysis, such as Fig.17 As shown in the figure, the trend analysis algorithm can accurately identify that there is an upward trend in the vibration data of this section, and automatically issue an alarm to remind on-site personnel to perform predictive maintenance as soon as possible, which can effectively avoid the unit bearing from getting stuck and maintain the normal operation of the unit.

[0156] The present invention fuses indicators such as energy entropy and relative entropy to obtain fusion features, so that the fusion features take into account both the absolute measurement standard and the relative measurement standard of the vibration data, so that a comprehensive analysis can be performed on each frequency band of the vibration data, which is suitable for transmission chain equipment under various complex working conditions, avoiding the limitation of single indicator evaluation, and the recognition method of the present invention does not require a large number of data samples for model training, and can accurately judge the characteristic change trend and fault degradation degree of each component of the transmission chain and issue alarm information in a timely manner, and is not affected by factors such as the geographical location and environment of the wind turbine. In addition, the recognition method of the present invention also has a high recognition rate for data (faults) with special characteristics. Furthermore, the recognition method of the present invention is simple and efficient in calculation, has a wide range of applications, and is very practical for distributed power plants with few personnel on duty.

[0157] It should be noted that the above-mentioned fusion indicators can also be replaced by approximate indicators, such as JS distance, Euclidean distance method, etc., and other algorithms can also be used to remove outliers in the trend analysis algorithm, such as k-means, DBSCAN, etc.

[0158] Fig.18 is a block diagram showing a transmission chain state identification device of a wind turbine generator set disclosed in the present invention, such as Fig.18 As shown, the device includes a vibration data acquisition unit 180 , an energy entropy acquisition unit 182 , a relative entropy acquisition unit 184 and a state identification unit 186 .

[0159] A vibration data acquisition unit 180 is configured to acquire actual vibration data of a transmission chain of a wind turbine generator set;

[0160] An energy entropy acquisition unit 182 is configured to acquire actual energy entropy of the actual vibration data based on the actual vibration data;

[0161] The relative entropy acquisition unit 184 is configured to acquire an actual relative entropy of the actual vibration data based on the actual vibration data and the reference vibration data;

[0162] The state identification unit 186 is configured to identify the state of the transmission chain based on the actual energy entropy, the actual relative entropy, the reference energy entropy and the reference relative entropy of the reference vibration data.

[0163] According to an embodiment of the present disclosure, the state identification unit 186 is also configured to obtain an actual fusion feature set of actual vibration data based on actual energy entropy and actual relative entropy; to obtain a reference fusion feature set of reference vibration data based on reference energy entropy and reference relative entropy; and to identify the state of the transmission chain based on the actual fusion feature set and the reference fusion feature set.

[0164] According to an embodiment of the present disclosure, when the actual vibration data includes actual vibration data of the transmission chain under N working conditions, and the reference vibration data includes reference vibration data of the transmission chain under N working conditions, the state identification unit 186 is further configured to identify the state of the transmission chain based on the fusion features under each working condition in the actual fusion feature set and the fusion features under the corresponding working condition in the reference fusion feature set, wherein the actual fusion feature set includes the fusion features of the actual vibration data of the transmission chain under N working conditions, and the reference fusion feature set includes the fusion features of the reference vibration data of the transmission chain under N working conditions.

[0165] According to an embodiment of the present disclosure, the state identification unit 186 is further configured to divide the fusion features in the actual fusion feature set and the reference fusion feature set into N fusion feature sets according to N working conditions, wherein the fusion feature set is a merged set of fusion features arranged in time sequence of reference vibration data and fusion features arranged in time sequence of actual vibration data; perform linear fitting processing on the N fusion feature sets respectively; and determine the state of the transmission chain based on each fusion feature set after the linear fitting processing.

[0166] According to an embodiment of the present disclosure, the state identification unit 186 is also configured to obtain the slope of each fused feature set after linear fitting processing; when the slope of any fused feature set is greater than a first predetermined value, it is determined that the state of the transmission chain is a fault state and the fault is gradually deteriorating; when the slope of any fused feature set is greater than a second predetermined value and less than the first predetermined value, it is determined that the state of the transmission chain is a normal state, wherein the second predetermined value is less than the first predetermined value; when the slope of any fused feature set is less than the second predetermined value, it is determined that the state of the transmission chain is a fault state and the fault is gradually recovering.

[0167] According to an embodiment of the present disclosure, the state identification unit 186 is further configured to perform smoothing on the actual fusion feature set; obtain the difference between all adjacent feature vectors in the actual fusion feature set after smoothing; when the sum of the difference values ​​reaches a preset threshold, sort the difference values ​​in order from large to small; take the maximum difference value as the first difference value, and iteratively perform the following operations: add the first difference value and the next difference value adjacent to the first difference value, and take the result of the addition as the first difference value, until the next difference value adjacent to the first difference value does not meet a predetermined condition, wherein the predetermined condition is that the current difference value and the next difference value adjacent to the current difference value are both positive, both negative, or both zero; when the accumulated value does not meet a preset step condition, the fusion features in the actual fusion feature set and the reference fusion feature set are divided into N fusion feature sets according to N working conditions.

[0168] According to an embodiment of the present disclosure, the device further includes: a fault determination unit 188 configured to determine that a serious fault occurs in the transmission chain when the accumulated value meets a preset step condition.

[0169] According to an embodiment of the present disclosure, the state recognition unit 186 is also configured to divide the fusion features in the actual fusion feature set into feature vectors of a predetermined number of segments; obtain the grid area of ​​each segment of the feature vector, and filter out the grid areas that meet predetermined conditions from the grid areas; and use the feature vectors corresponding to the filtered grid areas as the final actual fusion feature set.

[0170] According to an embodiment of the present disclosure, the energy entropy acquisition unit 182 is also configured to acquire the root mean square of each actual vibration data in the actual vibration data; based on the root mean square, acquire the average value of the root mean square under each working condition; acquire the absolute difference between the root mean square of each actual vibration data under each working condition and the average value of the root mean square under the corresponding working condition; eliminate abnormal vibration data from the actual vibration data to obtain valid vibration data, wherein the abnormal vibration data is actual vibration data whose absolute difference is greater than a predetermined threshold; perform time-frequency domain conversion on the valid vibration data to obtain a frequency domain signal of the valid vibration data; and acquire the actual energy entropy based on the frequency domain signal.

[0171] According to an embodiment of the present disclosure, the energy entropy acquisition unit 182 is also configured to acquire the electrical frequency of the wind turbine generator set under each operating condition before acquiring the actual energy entropy based on the frequency domain signal; and set the amplitude corresponding to the electrical frequency in the frequency domain signal to zero.

[0172] According to an embodiment of the present disclosure, the energy entropy acquisition unit 182 is also configured to obtain the rotational speed of the wind turbine generator set corresponding to each actual vibration data under each operating condition; multiply the rotational speed corresponding to each actual vibration data by a predetermined coefficient corresponding to the wind turbine generator set to obtain the electrical frequency of the wind turbine generator set under each operating condition.

[0173] According to an embodiment of the present disclosure, a computer-readable storage medium storing instructions is provided, wherein when the instructions are executed by at least one computing device, the at least one computing device is prompted to execute a transmission chain state identification method for a wind turbine generator set as described in any of the above embodiments.

[0174] According to an embodiment of the present disclosure, a system is provided comprising at least one computing device and at least one storage device storing instructions, wherein the instructions, when executed by the at least one computing device, prompt the at least one computing device to execute a method for identifying a transmission chain state of a wind turbine generator set as described in any of the above embodiments.

[0175] Although some embodiments of the present disclosure have been shown and described, it will be appreciated by those skilled in the art that modifications may be made to the embodiments without departing from the principles and spirit of the present disclosure, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for identifying the state of a transmission chain of a wind turbine generator set, characterized in that: include: Obtain actual vibration data of the transmission chain of the wind turbine; Based on the actual vibration data, obtaining actual energy entropy of the actual vibration data; Based on the actual vibration data and the reference vibration data, obtaining an actual relative entropy of the actual vibration data; identifying the state of the transmission chain based on the actual energy entropy, the actual relative entropy, the reference energy entropy and the reference relative entropy of the reference vibration data, Wherein, the identifying the state of the transmission chain based on the actual energy entropy, the actual relative entropy, the reference energy entropy and the reference relative entropy of the reference vibration data comprises: Based on the actual energy entropy and the actual relative entropy, obtaining an actual fusion feature set of the actual vibration data; Based on the reference energy entropy and the reference relative entropy, obtaining a reference fusion feature set of the reference vibration data; Based on the actual fused feature set and the reference fused feature set, identifying the state of the transmission chain, Wherein, when the actual vibration data includes actual vibration data of the transmission chain under N working conditions, and the reference vibration data includes reference vibration data of the transmission chain under the N working conditions, identifying the state of the transmission chain based on the actual fusion feature set and the reference fusion feature set includes: Based on the fusion features under each working condition in the actual fusion feature set and the fusion features under the corresponding working condition in the reference fusion feature set, the state of the transmission chain is identified, wherein the actual fusion feature set includes fusion features of actual vibration data of the transmission chain under N working conditions, and the reference fusion feature set includes fusion features of reference vibration data of the transmission chain under N working conditions, The step of identifying the state of the transmission chain based on the fusion features under each working condition in the actual fusion feature set and the fusion features under the corresponding working condition in the reference fusion feature set includes: Dividing the actual fusion feature set and the fusion features in the reference fusion feature set into N fusion feature sets according to N working conditions, wherein the fusion feature set is a combined set of the fusion features arranged in the time sequence of the reference vibration data and the fusion features arranged in the time sequence of the actual vibration data; Performing linear fitting processing on the N fusion feature sets respectively; Determine the state of the transmission chain based on each fused feature set after linear fitting processing, Wherein, determining the state of the transmission chain based on each fused feature set after linear fitting processing includes: Get the slope of each fusion feature set after linear fitting processing; When the slope of any fused feature set is greater than a first predetermined value, determining that the state of the transmission chain is a fault state and the fault is gradually deteriorating; When the slope of any fused feature set is greater than a second predetermined value and less than the first predetermined value, determining that the state of the transmission chain is a normal state, wherein the second predetermined value is less than the first predetermined value; When the slope of any fused feature set is less than the second predetermined value, it is determined that the state of the transmission chain is a fault state and the fault is gradually recovering.

2. The transmission chain state identification method according to claim 1, characterized in that: The actual fusion feature set and the fusion features in the reference fusion feature set are divided into N fusion feature sets according to N working conditions, including: Performing smoothing on the actual fusion feature set; Obtain the difference between all adjacent feature vectors in the actual fusion feature set after smoothing; When the sum of the difference values ​​reaches a preset threshold, sorting the difference values ​​in descending order; Taking the maximum difference as the first difference, iteratively performing the following operations: adding the first difference and the next difference adjacent to the first difference, and taking the addition result as the first difference, until the next difference adjacent to the first difference does not meet a predetermined condition, wherein the predetermined condition is that the current difference and the next difference adjacent to the current difference are both positive, both negative, or both zero; When the accumulated value does not satisfy the preset step condition, the fusion features in the actual fusion feature set and the reference fusion feature set are divided into N fusion feature sets according to N working conditions.

3. The transmission chain state identification method according to claim 2, characterized in that: The method further comprises: When the accumulated value meets a preset step condition, it is determined that a serious fault occurs in the transmission chain.

4. The transmission chain state identification method according to claim 1, characterized in that: Before dividing the actual fusion feature set and the fusion features in the reference fusion feature set into N fusion feature sets according to N working conditions, the method further includes: Divide the fused features in the actual fused feature set into feature vectors of a predetermined number of segments; Obtaining the grid area of ​​each segment of the feature vector, and screening out the grid area that meets the predetermined condition from the grid areas; The feature vector corresponding to the filtered grid area is used as the final actual fusion feature set.

5. The transmission chain state identification method according to claim 1, characterized in that: The acquiring actual energy entropy of the actual vibration data based on the actual vibration data includes: Obtaining a root mean square of each piece of actual vibration data; Based on the root mean square, obtaining an average value of the root mean square under each working condition; Obtain the absolute difference between the root mean square of each actual vibration data under each working condition and the average value of the root mean square under the corresponding working condition; Eliminating abnormal vibration data from the actual vibration data to obtain valid vibration data, wherein the abnormal vibration data is the actual vibration data whose absolute difference is greater than a predetermined threshold; Convert the effective vibration data into time-frequency domain to obtain a frequency domain signal of the effective vibration data; The actual energy entropy is obtained based on the frequency domain signal.

6. The transmission chain state identification method according to claim 5, characterized in that: Before acquiring the actual energy entropy based on the frequency domain signal, the method further includes: Obtain the electrical frequency of the wind turbine generator set under each operating condition; The amplitude of the frequency domain signal corresponding to the electrical frequency is set to zero.

7. The transmission chain state identification method according to claim 6, characterized in that: The obtaining of the electrical frequency of the wind turbine generator set under each working condition comprises: Obtain the speed of the wind turbine generator set corresponding to each piece of actual vibration data under each working condition; The rotation speed corresponding to each piece of actual vibration data is multiplied by a predetermined coefficient corresponding to the wind generator set to obtain the electrical frequency of the wind generator set under each working condition.

8. A transmission chain state identification device for a wind turbine generator set, characterized in that: include: A vibration data acquisition unit, configured to acquire actual vibration data of a transmission chain of a wind turbine generator set; an energy entropy acquisition unit, configured to acquire actual energy entropy of the actual vibration data based on the actual vibration data; a relative entropy acquisition unit, configured to acquire an actual relative entropy of the actual vibration data based on the actual vibration data and the reference vibration data; a state identification unit configured to identify the state of the transmission chain based on the actual energy entropy, the actual relative entropy, the reference energy entropy and the reference relative entropy of the reference vibration data, The state identification unit is further configured to: obtain an actual fusion feature set of the actual vibration data based on the actual energy entropy and the actual relative entropy; obtain a reference fusion feature set of the reference vibration data based on the reference energy entropy and the reference relative entropy; and identify the state of the transmission chain based on the actual fusion feature set and the reference fusion feature set. Wherein, in the case where the actual vibration data includes actual vibration data of the transmission chain under N working conditions, and the reference vibration data includes reference vibration data of the transmission chain under the N working conditions, the state identification unit is further configured to: identify the state of the transmission chain based on the fusion features under each working condition in the actual fusion feature set and the fusion features under the corresponding working condition in the reference fusion feature set, wherein the actual fusion feature set includes the fusion features of the actual vibration data of the transmission chain under N working conditions, and the reference fusion feature set includes the fusion features of the reference vibration data of the transmission chain under N working conditions, The state recognition unit is further configured to: divide the actual fusion feature set and the fusion features in the reference fusion feature set into N fusion feature sets according to N working conditions, wherein the fusion feature set is a combined set of fusion features arranged in the time sequence of the reference vibration data and fusion features arranged in the time sequence of the actual vibration data; perform linear fitting processing on the N fusion feature sets respectively; determine the state of the transmission chain based on each fusion feature set after the linear fitting processing, Wherein, the state identification unit is further configured to: obtain the slope of each fused feature set after linear fitting processing; when the slope of any fused feature set is greater than a first predetermined value, determine that the state of the transmission chain is a fault state and the fault is gradually deteriorating; when the slope of any fused feature set is greater than a second predetermined value and less than the first predetermined value, determine that the state of the transmission chain is a normal state, wherein the second predetermined value is less than the first predetermined value; when the slope of any fused feature set is less than the second predetermined value, determine that the state of the transmission chain is a fault state and the fault is gradually recovering.

9. A computer-readable storage medium storing instructions, wherein: When the instruction is executed by at least one computing device, the at least one computing device is prompted to execute the method for identifying the state of a transmission chain of a wind turbine generator set according to any one of claims 1 to 7.

10. A system comprising at least one computing device and at least one storage device storing instructions, wherein: When the instructions are executed by the at least one computing device, the at least one computing device is prompted to execute the method for identifying the state of a transmission chain of a wind turbine generator set according to any one of claims 1 to 7.

Images