Rapid evaluation method and system for generator vibration data of wind turbine generator set

Abstract

The present application provides a rapid evaluation method and system for generator vibration data of a wind turbine generator set. The method comprises: step 1, acquiring current-day operation data of a wind turbine generator set under test, wherein the current-day operation data comprises a time label, an operation power and a generator vibration signal; step 2, constructing a two-dimensional plane scatter plot on the basis of the current-day operation data; and step 3, combining the obtained two-dimensional plane scatter plot with a preset regression function, and evaluating the vibration state of said wind turbine generator set. The present application utilizes a lossless compression method to process data, so that the current-day vibration state of a generator can be accurately evaluated.

Description

A wind turbine generator vibration data rapid evaluation method and system

[0001] The present application claims priority to the Chinese patent application No. 202410210347.2, filed on February 26, 2024, and entitled "A wind turbine generator vibration data rapid evaluation method and system", the whole content of which is incorporated herein by reference. TECHNICAL FIELD

[0002] The present application belongs to the field of new energy wind power generation, and particularly relates to a wind turbine generator vibration data rapid evaluation method and system. BACKGROUND

[0003] During the operation of a wind turbine generator, due to the continuous impact of varying loads, the blades, hub, and components in the transmission chain rigidly connected to the hub, including the main shaft, gearbox, high-speed shaft, and generator, are prone to failure. According to statistics, the related faults of the transmission chain of the wind turbine generator cause the longest downtime and have the greatest impact on power production. Among them, the generator in the transmission chain is an important rotating mechanical-electrical component, and its failure may lead to extreme serious accidents such as cabin fire, runaway, and tower collapse. Therefore, the existing method is to monitor the vibration data of the generator in real time during the operation of the wind turbine generator, use vibration monitoring and analysis technology for fault diagnosis and early warning, and use time domain analysis, frequency domain analysis, and time-frequency domain superposition analysis to predict potential problems and future faults of the wind turbine generator, so as to take timely measures and improve the reliability of the wind turbine generator operation. However, the existing generator vibration data has a high sampling frequency and a large amount of data, and real-time data analysis requires a large amount of calculation. If offline analysis is used, there is a problem of poor timeliness. Therefore, there is an urgent need for a wind turbine generator vibration data rapid evaluation method to reduce the data analysis workload and ensure rapid and accurate evaluation of the generator state.

[0004] SUMMARY

[0005] The present application aims to provide a wind turbine generator vibration data rapid evaluation method and system, which solves the defects of large real-time data analysis and poor timeliness of offline analysis in the existing wind turbine fault early warning method. The present application can quickly process massive high-frequency generator vibration data, detect potential operation state deviation of the wind turbine generator early, and avoid the adverse consequences of extreme events caused by faults.

[0006] In order to achieve the above-mentioned purpose, the technical solution adopted by the present application is as follows:

[0007] The wind turbine generator vibration data rapid evaluation method provided by the present application comprises the following steps:

[0008] Step 1, obtaining the operation data of the wind turbine to be tested on the day, the operation data including time label, operation power and generator vibration signal;

[0009] Step 2, constructing a two-dimensional plane scatter plot according to the operation data on the day;

[0010] Step 3, combining the obtained two-dimensional plane scatter plot with the preset regression function to evaluate the vibration state of the wind turbine to be tested;

[0011] The method for constructing the preset regression function is as follows:

[0012] S31, obtaining a matrix composed of the operation data of the wind turbine to be tested in the last year, the operation data including time label, operation power and generator vibration signal;

[0013] S32, dividing the matrix according to the time label to obtain a plurality of sub-matrices;

[0014] S33, non-distortionally compressing the data of each sub-matrix to obtain non-distortionally compressed data corresponding to each sub-matrix;

[0015] S34, extracting features from the non-distortionally compressed data to obtain the preset regression function.

[0016] Optionally, in step 2, the two-dimensional plane scatter plot is constructed according to the operation data on the day, and the specific method is as follows:

[0017] The operation power in the operation data on the day is taken as the x-axis, and the generator vibration signal in the operation data on the day is taken as the y-axis to construct the two-dimensional plane scatter plot.

[0018] Optionally, in S32, the matrix is divided according to the time label to obtain a plurality of sub-matrices, and the specific method is as follows:

[0019] The data corresponding to the same value in the time label is taken as a group to obtain a plurality of sub-matrices.

[0020] Optionally, before the data of each sub-matrix is non-distortionally compressed, the data of each sub-matrix is preprocessed to obtain a preprocessed sub-matrix.

[0021] Optionally, the specific method for preprocessing the data of each sub-matrix is as follows:

[0022] The row corresponding to the operation power less than or equal to zero in each sub-matrix is deleted to obtain a removed sub-matrix;

[0023] The operation power of each removed sub-matrix is rearranged in ascending order, and is equally divided into a plurality of small sub-matrices according to a preset power window;

[0024] The meaningless data points in each small sub-matrix are removed to obtain a processed small sub-matrix;

[0025] All the processed small sub-matrices are combined and rearranged according to the order of time labels to obtain a preprocessed sub-matrix.

[0026] Optionally, in S33, a Midimax compression algorithm is used to compress the data of each sub-matrix without distortion to obtain non-distortion compression data corresponding to each sub-matrix.

[0027] Optionally, in S34, feature extraction is performed on the non-distortion compression data, and the specific method is:

[0028] The obtained non-distortion compression data is sorted and reconstructed according to the order of running power from small to large to obtain reconstructed data.

[0029] The running power in the reconstructed data is taken as the x-axis, and the generator vibration signal in the reconstructed data is taken as the y-axis to construct a two-dimensional plane scatter plot.

[0030] An upper boundary point in the two-dimensional plane scatter plot is identified by using a K-means clustering method to obtain a point set.

[0031] A regression function corresponding to the point set is constructed by using a nonlinear regression method.

[0032] Optionally, in step 3, the running state of the to-be-tested wind turbine generator is evaluated, and the specific method is:

[0033] An upper boundary point of the two-dimensional plane scatter plot is identified by using a K-means clustering method to obtain a point set.

[0034] If ω1 points in the point set are located on the upper side of the regression function, the to-be-tested generator has abnormal vibration on the day; and is recorded as class2 (i.e., the second state classification).

[0035] If ω2 points in the point set are located on the upper side of the regression function, and there are no ω1 points located on the upper side of the regression function, the to-be-tested generator has a slight state deviation on the day; and is recorded as class1 (i.e., the first state classification).

[0036] If the number of points located on the upper side of the regression function in the point set is less than ω2, the to-be-tested generator has normal vibration on the day; and is recorded as class0 (i.e., the 0th state classification).

[0037] Wherein, ω1 is a dynamic abnormal threshold; ω2 is a dynamic deviation threshold; r is the number of the point set, and has:

[0038] ω2<ω1<r.

[0039] Step 4, record the result obtained after evaluation of the data compressed by non-distortion in step 3 on the day, and assign a value for each state classification, class2 (i.e. the second state classification) is assigned a value of 2, class1 (i.e. the first state classification) is assigned a value of 1, and class0 (i.e. the 0th state classification) is assigned a value of 0;

[0040] Calculate the state value accumulation Score of the previous 30 days every 30 days,

[0041] If Score≤10, no action is taken for the generator of the unit;

[0042] If 10<Score≤20, arrange a tour at an appropriate time in the form of opportunity maintenance;

[0043] If 20<Score≤30, arrange a tour as soon as possible;

[0044] If Score>30, the generator has obvious long-term state deviation, and maintenance is arranged;

[0045] By combining the vibration data with the preset regression function, evaluation is performed every day, and the evaluation results are integrated every 30 days, and corresponding tour and maintenance measures are taken according to different integrated score values; the state deviation of the generator is identified to the greatest extent, and the safety of the unit is ensured.

[0046] A wind turbine generator vibration data rapid evaluation system, comprising:

[0047] A current data acquisition unit configured to acquire current operation data of a wind turbine to be tested, the current operation data comprising a time label, an operation power and a generator vibration signal;

[0048] A scatter plot construction unit configured to construct a two-dimensional plane scatter plot according to the current operation data;

[0049] A vibration state evaluation unit configured to combine the obtained two-dimensional plane scatter plot with a preset regression function, and evaluate the vibration state of the wind turbine to be tested.

[0050] Compared with the prior art, the wind turbine generator vibration data rapid evaluation system has the following beneficial effects:

[0051] The wind turbine generator vibration data rapid evaluation method provided by the application adopts non-distortion compressed data to process the operation data of the wind turbine to be tested in the previous year, retains the data characteristics, and can quickly and with small calculation amount, does not introduce non-actual data, and only returns a subset of the original data. The feature extraction method used can update the regression function every day, and the regression function is combined with the operation data of the generator on the day to obtain the vibration state of the generator on the day, thereby realizing accurate evaluation of the vibration of the generator. BRIEF DESCRIPTION OF DRAWINGS

[0052] FIG. 1 is a whole technical diagram;

[0053] FIG. 2 is a data acquisition and preprocessing method and system flow chart;

[0054] FIG. 3 is a data non-distortion compression method and system flow chart;

[0055] FIG. 4 is a feature extraction method and system flow chart;

[0056] FIG. 5 is an evaluation method and system flow chart. DETAILED DESCRIPTION

[0057] In the following description, for the purpose of explanation and not limitation, specific details are set forth, such as particular system structures, techniques, etc., in order to provide a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application can be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.

[0058] Embodiment 1

[0059] As shown in FIGS. 1-5, the present embodiment provides a wind turbine generator vibration data rapid evaluation method and system, which includes data acquisition and preprocessing, data non-distortion compression, feature extraction, and final evaluation, and specifically includes the following steps:

[0060] Step 1, the present application takes a target wind turbine #1 in a wind farm as an example. Data acquisition includes the running time label T, running power P, and vibration signal V collected by the generator vibration acceleration sensor of the #1 wind turbine up to the day before analysis, where the sampling frequency is 10 kHz, and a data matrix A can be constructed:

[0061] where A is an n-row 3-column matrix; n represents the number of data collection points of the #1 wind turbine in the previous year; T1, T2, …, T i ,…,T n are time labels in the collected data, which are in the format of 20XX-MM-DD:HH-FF-SS-UUUUUU, where XX represents the "year" in the corresponding time label, MM corresponds to "month", DD corresponds to "day", HH corresponds to "hour", FF corresponds to "minute", SS corresponds to "second", and UUUUUU corresponds to "ten microseconds"; P1, P2, …, P i ,…,P n are running powers corresponding to the time labels; V1, V2, …, V i ,…,V nThe vibration signal is collected by the generator vibration acceleration sensor corresponding to the time label.

[0062] Step 2: Divide matrix A into twelve submatrices, as follows:

[0063] Dividing matrix A according to the time labels in the collected data, where time labels T with the same "month" value are grouped into the same matrix, we can obtain:

[0064] Where A1 represents the operating data acquired by wind turbine #1 in the same month of the previous year; A2 represents the operating data acquired by wind turbine #1 in the following month of the previous year, and so on. 12 This is the operating data obtained for wind turbine #1 in the first month of this year.

[0065] Step 3, for A1, A2, ..., A 12 Data preprocessing is performed, taking matrix A1 as an example. The specific method is as follows:

[0066] First, remove the rows in the second column of the power data that are less than or equal to 0 from matrix A1 to form the removed submatrix A1'.

[0067] Secondly, the operating power in the second column of matrix A1' is rearranged in ascending order and further divided into 10 smaller submatrices: A′1=[A′ 11 ,A′ 12 …A 110 ']

[0068] Among them, A' 11 This refers to the data obtained for Unit #1 under the condition of 0-10% rated power; A' 12 This refers to the data obtained when the power of Unit #1 is at 10-20% of its rated power; and so on, A' 110 This is the data obtained when the power of Unit #1 is at 90-100% of its rated power.

[0069] Using the 6σ principle to apply to A' 11 ,A' 12 …A 110 'Perform data processing, remove meaningless points, and use A' 11 For example, the specific method is as follows:

[0070] Calculate A' 11 The mathematical expectation ε' of the third column of vibration data 11 With variance σ' 11 ;

[0071] By A' 11 The first element of the third column, V'11 Start to determine,

[0072] If V' 11 > ε' 11 + 6 x σ' 11 Or V' 11 < ε' 11 - 6 x σ' 11 , then reject V' 11 The corresponding row data;

[0073] Otherwise, keep the row data;

[0074] Then A' 11 After processing, the matrix A" is obtained 11 ;

[0075] Similarly, A" processed by the 6σ principle can be obtained 12 …A 110 ";

[0076] Finally, for A" 11 ,A" 12 …A 110 ", to maximize the authenticity of the data and subsequent compression, no data rejection processing is performed, and the ten matrices A" 11 ,A" 12 …A 110 " are sequentially combined and reordered according to the order of the first column time label to obtain the preprocessed sub-matrix A"1.

[0077] Similarly, A2,…, A 12 are processed in the above manner, and finally A"2,…A" 12 are obtained.

[0078] Then we have:

[0079] Where A" is the preprocessed running data of #1 wind turbine in the previous year, A"1, A"2,…A" 12 are the preprocessed running data of #1 wind turbine in the previous year from the previous month to the previous month.

[0080] Step 4, the sampling frequency of the preprocessed running data of #1 wind turbine in the previous year in step 3 is 10 kHz, and the data volume is huge, so direct analysis and calculation is computationally intensive, and data non-distortion compression is needed, the specific method is as follows:

[0081] Herein, taking the non-distortion compression of the operation data A"1 of the #1 wind turbine in the previous year and the month as an example, the operation data A"1 is sorted according to the order of the time tags, and is a general high-frequency acquisition time domain sequence. According to step 1-3, A"1 is an m-row 3-column matrix, wherein the first column is the time tag, the second column is the operation power corresponding to the time tag, and the third column is the vibration signal collected by the vibration acceleration sensor corresponding to the time tag, and m≤100000×3600×24×31=2.6784×10 11

[0082] The Midimax compression algorithm is adopted to carry out non-distortion compression on A"1, and the specific process is as follows:

[0083] S41, inputting the time sequence data, i.e. A"1 and the compression coefficient λ, into the Midimax algorithm. In this example, the compression coefficient λ is much smaller than m, i.e. λ<<m

[0084] S42, splitting the time sequence data A"1 into λ non-overlapping windows of equal size, wherein the size L of each non-overlapping window is calculated as follows:

[0085] Then, the original A"1 is divided into λ non-overlapping window corresponding matrices, i.e.

[0086] S43, sorting the λ non-overlapping window corresponding matrices in S42 in ascending order according to the operation power (i.e. the second column of the corresponding matrix), to obtain B1, B2…, B λ ;

[0087] S44, for B1, B2…, B λ , taking B1 as an example, selecting the rows corresponding to the k largest elements, the rows corresponding to the k smallest elements, and the rows corresponding to the k elements closest to e1 in the third column (i.e. the vibration signal collected by the generator vibration acceleration sensor) of each matrix to form a matrix C1, wherein e1 represents the mathematical expectation of the third column (i.e. the vibration signal collected by the generator vibration acceleration sensor) of the matrix B1. Similarly, C2, C3…C λ , C1, C2…C λ are all 3k-row 3-column matrices;

[0088] S45, reordering the matrices C1, C2…C λ obtained in S44 according to the operation power in the second column from small to large to form a matrix D1, and D1 is a 3×k×λ-row 3-column matrix. Then, the original m-row 3-column matrix A"1 is compressed into a 3×k×λ-row 3-column matrix D1;

[0089] Similarly, A"2, A"3, …A" 12non-distortion compression, and finally obtain D2, D3…D 12 .

[0090] Step 5, feature extraction is performed on the non-distortion compressed data

[0091] S51, D1, D2…D 12 are combined into a matrix D, and the matrix D is a 36xkxl matrix with 3 columns, wherein the first column is a time label, the second column is a running power corresponding to the time label, and the third column is a vibration signal collected by a vibration acceleration sensor corresponding to the time label;

[0092] S52, the matrix D is reordered according to the running power in the second column from small to large to form a matrix O;

[0093] S53, a two-dimensional plane scatter plot is constructed using the matrix O in S52, wherein the x-axis is the running power in the second column of the matrix O, and the y-axis is the vibration signal collected by the vibration acceleration sensor in the third column of the matrix O;

[0094] S54, K-means clustering method is used to identify the upper boundary points in the scatter plot in S53 to form a point set, and the upper boundary points in this set are v, denoted as edge O =[dot O1 ,dot O2 ,…,dot Ov ]: dot O1 =(x O1 ,y O1 ) dot O2 =(x O2 ,y O2 ) dot Ov =(x Ov ,y Ov )

[0095] wherein dot O1 ,dot O2 ,…,dot Ov are the coordinates of each point;

[0096] S55, a nonlinear regression method is used to construct a regression function f O of the upper boundary point set edge O in S54, and the independent variable of the regression function f O is the running power, and the dependent variable is the vibration signal collected by the vibration acceleration sensor.

[0097] Step 6, the wind turbine to be tested is evaluated according to the features extracted in step 5, and the target wind turbine #1 in the wind farm is taken as an example

[0098] S61, at 00:00 every night, extracting the running time label T, the running power P and the vibration signal V collected by the generator vibration acceleration sensor of the day (00:00 to 23:59), where the sampling frequency is 10 kHz, to form a data matrix H';

[0099] S62, removing the rows corresponding to the power data less than or equal to 0 in the second column of the matrix H' to form a matrix H;

[0100] S63, constructing a two-dimensional plane scatter plot using the matrix H in S62, where the x-axis is the running power in the second column of the matrix H and the y-axis is the vibration signal collected by the generator vibration acceleration sensor in the third column of the matrix H;

[0101] S64, identifying the upper boundary points in the scatter plot in S63 using the K-means clustering method to form a point set, where the number of the upper boundary points is r, denoted as edge H =[dot H1 ,dot H2 ,…,dot Hr ], dot H1 =(x H1 ,y H1 ) dot H2 =(x H2 ,y H2 ) dot Hr =(x Hr ,y Hr )

[0102] where dot H1 ,dot H2 ,…,dot Hr are the coordinates of the points;

[0103] S65, analyzing and judging the data obtained in S64,

[0104] if ω1 points in dot H1 ,dot H2 ,…,dot Hr are located on the upper side of the regression function in S55, it is determined that the generator vibration of the day is abnormal;

[0105] if ω2 points in dot H1 ,dot H2 ,…,dot Hr are located on the upper side of the regression function in S55, and there are no ω1 points located on the upper side of the regression function in S55, it is determined that the generator vibration of the day is slightly state deviated;

[0106] if dot H1 ,dot H2 ,…,dotHr If the number of points above the regression function in S55 is less than ω2, it is assessed that the generator vibration is normal on the day;

[0107] Wherein ω1 and ω2 are dynamic anomaly threshold and dynamic offset threshold respectively, and have: ω2<ω1<r.

[0108] Embodiment 2

[0109] On the basis of embodiment 1, the wind turbine generator vibration data rapid evaluation system provided by the embodiment comprises:

[0110] A day data acquisition unit is configured to acquire the running data of the wind turbine to be tested on the day, wherein the running data comprises a time label, running power and generator vibration signal;

[0111] A scatter plot construction unit is configured to construct a two-dimensional plane scatter plot according to the running data on the day;

[0112] A vibration state evaluation unit is configured to combine the obtained two-dimensional plane scatter plot with a preset regression function, and evaluate the vibration state of the wind turbine to be tested.

[0113] The application adopts a non-distortion compression method to process data, which can retain data characteristics, quickly and with small amount of calculation, does not introduce non-actual data, only returns a subset of original data, and the feature extraction method can update the regression curve every day to realize accurate evaluation of the generator vibration.

[0114] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A wind turbine generator vibration data quick evaluation method, characterized in that, The method comprises the following steps: Step 1, obtaining the operation data of the wind turbine on the day, the operation data comprising a time label, an operation power and a generator vibration signal; Step 2, constructing a two-dimensional plane scatter plot according to the operation data on the day; Step 3, combining the obtained two-dimensional plane scatter plot with a preset regression function to evaluate the vibration state of the wind turbine to be tested; The method for constructing the preset regression function comprises the following steps: S31, obtaining a matrix composed of the operation data of the wind turbine in the previous year, the operation data comprising a time label, an operation power and a generator vibration signal; S32, dividing the matrix according to the time label to obtain a plurality of sub-matrices; S33, performing non-distortion compression on the data of each sub-matrix to obtain non-distortion compression data corresponding to each sub-matrix; S34, performing feature extraction on the non-distortion compression data to obtain the preset regression function.

2. A wind turbine generator vibration data quick evaluation method according to claim 1, characterized in that, In step 2, the two-dimensional plane scatter plot is constructed according to the operation data on the day, and the specific method is as follows: The operation power in the operation data on the day is taken as the x-axis, and the generator vibration signal in the operation data on the day is taken as the y-axis to construct the two-dimensional plane scatter plot.

3. A wind turbine generator vibration data quick evaluation method according to claim 1, wherein In S32, the matrix is divided according to the time label to obtain a plurality of sub-matrices, and the specific method is as follows: The data corresponding to the same value in the time label is taken as a group to obtain a plurality of sub-matrices.

4. A wind turbine generator vibration data quick evaluation method according to claim 1, wherein Before performing non-distortion compression on the data of each sub-matrix, the data of each sub-matrix is preprocessed to obtain a preprocessed sub-matrix.

5. A wind turbine generator vibration data quick evaluation method according to claim 4, wherein, The specific method for preprocessing the data of each sub-matrix is as follows: The row corresponding to the operation power less than or equal to zero in each sub-matrix is deleted to obtain a removed sub-matrix; The operation power of each removed sub-matrix is rearranged in ascending order, and is equally divided into a plurality of small sub-matrices according to a preset power window; The meaningless data points in each small sub-matrix are removed to obtain a processed small sub-matrix; All the processed small sub-matrices are combined and rearranged in the order of the time label to obtain a preprocessed sub-matrix.

6. A wind turbine generator vibration data quick evaluation method according to claim 1, wherein In S33, the Midimax compression algorithm is used to perform non-distortion compression on the data of each sub-matrix to obtain non-distortion compression data corresponding to each sub-matrix.

7. A wind turbine generator vibration data quick evaluation method according to claim 1, wherein In S34, the specific method for performing feature extraction on the non-distortion compression data is as follows: The obtained non-distortion compression data is sorted and reconstructed in the order of the operation power from small to large to obtain reconstructed data; The operation power in the reconstructed data is taken as the x-axis, and the generator vibration signal in the reconstructed data is taken as the y-axis to construct a two-dimensional plane scatter plot; The K-means clustering method is used to identify the upper boundary points in the two-dimensional plane scatter plot to obtain a point set; The nonlinear regression method is used to construct a regression function corresponding to the point set. In step 3, the specific method for evaluating the operation state of the wind turbine to be tested is as follows:

8. A wind turbine generator vibration data quick evaluation method according to claim 1, wherein, The K-means clustering method is used to identify the upper boundary points of the two-dimensional plane scatter plot to obtain a point set; If ω1 points in the point set are located on the upper side of the regression function, the vibration of the wind turbine to be tested on the day is abnormal. ​ If ω2 points in the point set are located on the upper side of the regression function, and there are no ω1 points located on the upper side of the regression function, then the vibration of the to-be-tested generator set on the day exists a slight state deviation; If the number of points located on the upper side of the regression function in the point set is less than ω2, then the vibration of the to-be-tested generator set on the day is normal; Wherein, ω1 is a dynamic anomaly threshold; ω2 is a dynamic deviation threshold; r is the number of the point set, and has: ω2 < ω1 < r.

9. A wind turbine generator vibration data quick evaluation system, characterized by, It comprises: A day data acquisition unit configured to acquire day operation data of a to-be-tested wind turbine, the day operation data comprising a time label, operation power and generator vibration signals; A scatter plot construction unit configured to construct a two-dimensional plane scatter plot according to the day operation data; A vibration state evaluation unit configured to combine the obtained two-dimensional plane scatter plot with a preset regression function, and evaluate the vibration state of the to-be-tested wind turbine.

Images