Method, Medium and Device for Modal Identification of Offshore Wind Turbines Based on Hybrid Modal Recognition

Abstract

The present invention relates to the technical field of offshore wind energy, and particularly relates to a method, medium and device for modal identification of an offshore wind turbine based on hybrid modal identification, comprising the following steps: using the stochastic subspace-Kalman filtering method to remove the harmonic components in the response; applying the response with the harmonic components removed to the modal identification of the improved natural excitation technique-feature system realization algorithm; applying the response with the harmonic components removed to the frequency identification of the power spectral density transmissibility method; comparing the frequency results of the power spectral density transmissibility method with those of the improved natural excitation technique-feature system realization algorithm to determine the correct structural frequency; and finally extracting the correct structural physical modal information from the improved natural excitation technique-feature system realization algorithm. The present invention solves the technical problem that the existing modal identification methods are limited when applied to offshore wind turbines and cannot accurately identify their modal parameters.

Description

Technical Field

[0001] The present invention relates to the technical field of offshore wind energy, and particularly to an offshore wind turbine modal identification method, medium and device based on hybrid modal identification. Background Art

[0002] Wind energy is one of the fastest-growing areas of renewable energy. As a key facility for offshore wind energy development, offshore wind turbines (OWTs) are vulnerable to damage due to exposure to harsh and complex marine environments. OWTs are continuously affected by environmental loads (such as wind, waves, etc.) during operation and occasionally face extreme loads from events such as typhoons and earthquakes. To monitor the structural safety of OWTs and prevent catastrophic accidents caused by structural damage, vibration-based structural health monitoring (SHM) is generally considered one of the most effective strategies. Wind energy industry companies also hope to ensure the safe operation of OWTs by using SHM strategies. In the SHM of OWTs, modal identification plays a crucial role because the identified modal parameters can provide important information for further structural condition assessment and damage detection. Existing methods are usually developed under the assumption that the structure is excited by Gaussian white noise. However, during the operation of OWTs, the structure is constantly affected by environmental factors such as waves and wind, and the high-speed rotating blades also continuously affect the dynamic behavior of the structure. At this time, the excitation received by OWTs becomes more complex, and in addition to white noise, there is also a high possibility of the influence of colored noise and harmonic excitation. The above excitation conditions clearly violate the Gaussian white noise assumption, which limits the application of traditional modal identification methods to OWTs and makes it challenging to accurately identify their modal parameters. Summary of the Invention

[0003] The purpose of the present invention is to provide an offshore wind turbine modal identification method, medium and device based on hybrid modal identification to solve the technical problem that the existing modal identification methods are limited when applied to offshore wind turbines and cannot accurately identify their modal parameters. The specific technical solutions are as follows:

[0004] The present invention provides an offshore wind turbine modal identification method based on hybrid modal identification, including the following steps:

[0005] S1. Obtain the acceleration vibration response of the target structure through the sensors of the measurement platform;

[0006] S2. Determine the harmonic frequencies of the acceleration vibration response in S1, and then use the stochastic subspace-Kalman filtering method to remove the harmonic components in the response to obtain the acceleration response after removing the harmonic components;

[0007] S3. Apply the acceleration response with harmonic components removed to the modal identification of the improved natural excitation technique - eigensystem realization algorithm to obtain the modal frequency results of the improved natural excitation technique - eigensystem realization algorithm; apply the acceleration vibration response with harmonic components removed to the frequency identification of the power spectral density transmissibility method to obtain the structural frequency after removing colored noise.

[0008] S4. Compare the modal frequency results of the improved natural excitation technique - eigensystem realization algorithm with the structural frequency after removing colored noise to determine the correct structural frequency.

[0009] S5. Extract the correct structural physical modal information from the improved natural excitation technique - eigensystem realization algorithm based on the correct structural frequency obtained in S4 to obtain the correct modal parameters of the wind turbine.

[0010] A further improvement of the modal identification method for offshore wind turbines based on hybrid modal identification in the present invention is that S2 specifically includes:

[0011] S201. Determine the acceleration vibration response and the magnitudes of its harmonic frequencies.

[0012] S202. Based on the acceleration vibration response, obtain the state matrix A, the observation matrix C, and three covariance matrices R, S, and Q through the stochastic subspace method.

[0013] S203. Use the acceleration vibration response, the determined magnitudes of harmonic frequencies, the state matrix, the observation matrix, and the three covariance matrices to obtain the harmonic component response through the Kalman filtering method.

[0014] S204. Decompose the harmonic component response through the spatial orthogonal projection and matrix decomposition method to obtain the acceleration response with harmonic components removed.

[0015] A further improvement of the modal identification method for offshore wind turbines based on hybrid modal identification in the present invention is that the calculation steps of the state matrix A and the observation matrix C include:

[0016] Let the acceleration vibration response be the observed data y k , based on the observed data y k Construct the Hankel matrix Y of the past time steps past ∈l·i×j and Y future ∈l·i×j, from which the projection matrix O is obtained as:

[0017]

[0018] Among them, O i ∈l·i×l·i, i represents the number of rows of the defined Hankel matrix block,.... +It is indicated that there are l(i + 1) block rows,.- indicates that there are l(i - 1) block rows, j = N - i + 1 is the block column number, N represents the length of the observation sequence, Y future represents the Hankel matrix for future time steps, def indicates that the projection matrix O is obtained from the currently defined equation, and l is the number of sensors;

[0019] Perform singular value decomposition on O i to obtain:

[0020] O i = USV T 2);

[0021] where, T represents the transpose of the matrix, U represents the left matrix after decomposition, V represents the right matrix after decomposition, and S represents the diagonal matrix of singular values;

[0022] Based on formula 2), the extended observability matrix is further obtained:

[0023]

[0024] where, U1 and S1 represent the first n columns of U and S, and n represents the selected order; Γ i-1 represents Γ i the extended observability matrix excluding the last l block rows; then the state sequence is obtained from formula 4):

[0025]

[0026] where, represents the state sequence of the number of block rows of the i-th block matrix, represents the state sequence of the number of block rows of the (i + 1)-th block matrix, represents Γ i the Moore-Penrose pseudoinverse of, represents the Moore-Penrose pseudoinverse excluding the last l rows;

[0027] The matrices A and C are obtained from formula 5):

[0028]

[0029] where, Y i|i represents the Hankel matrix with only one row block.

[0030] A further improvement of the method for modal identification of offshore wind turbines based on hybrid modal identification in the present invention lies in that the covariance matrices Q, S, and R are obtained from the residuals:

[0031]

[0032] where, (.) Tdenotes the transpose of a matrix, E denotes the expectation, and ρ w and ρ v are both Kalman filter residuals.

[0033] A further improvement of the offshore wind turbine modal identification method based on hybrid modal identification in the present invention lies in that S203 specifically includes the following steps:

[0034] Based on Kalman filtering and the system state equation, the state sequence of each order is estimated, and the optimal state estimate is as follows:

[0035]

[0036] where denotes the optimal estimate at the next time step, denotes the state at the current time step, A denotes the state matrix, and K k denotes the Kalman filter, and e is defined k obtained from Equation (8);

[0037] The observation equation is expressed as:

[0038]

[0039] where y k denotes the observed data, and C denotes the observation matrix;

[0040] The update of the Kalman gain K k is expressed as:

[0041] K k =(AP k C T +S)(R + CP k C T ) -1 9);

[0042] where P k denotes the covariance matrix of the Kalman filter;

[0043] The update of the covariance matrix is expressed as:

[0044] P k+1 = AP k A T + Q - K k (AP k C T + S) T 10);

[0045] where P k+1 denotes the covariance matrix of the estimated Kalman filter at the next time step;

[0046] The estimated state and matrix C are linearly transformed based on an invertible matrix V to obtain:

[0047]

[0048] where (.) V represents a linear transformation based on the invertible matrix V, and (.) -1 represents the inverse of the matrix;

[0049] This matrix is defined as a modal basis for distinguishing the modal components and harmonic components of the system, and is constituted as follows:

[0050]

[0051] where represents taking the real part, represents taking the imaginary part, and Ψ represents the eigenvector of matrix A;

[0052] Define a position matrix s for the localization of harmonic components:

[0053]

[0054] where m represents the number of orders of the system mode, h represents the number of harmonic modes, and I represents the identity matrix;

[0055] Based on equations (14) and (16), the harmonic component response is calculated by the following formula:

[0056]

[0057] where l is the number of sensors.

[0058] A further improvement of the modal identification method for offshore wind turbines based on hybrid modal identification in the present invention lies in that the improved natural excitation technique - eigensystem realization algorithm method in S3 is specifically as follows:

[0059] First, the acceleration response after removing the harmonic components is used as the input of the improved natural excitation technique - eigensystem realization algorithm method;

[0060] Then, the key parameter values of the natural excitation technique - eigensystem realization algorithm are obtained according to the double reference point Monte Carlo stabilization diagram;

[0061] And then, a result set is preliminarily obtained based on the double reference point Monte Carlo stabilization diagram and the fuzzy clustering method;

[0062] Finally, the modal parameter results of the eigensystem realization algorithm are obtained.

[0063] The further improvement of the modal identification method for offshore wind turbines based on hybrid modal identification lies in that the power spectral density transmissibility method in S3 specifically includes the following steps:

[0064] The transmissibility function is the ratio of two dynamic responses x p (t) and x q (t). Then the power spectral transmissibility is the ratio of the cross-power spectral density with respect to the reference responses x z (i) and x p (t), x q (t), as shown below:

[0065]

[0066] Among them, represents the cross-correlation power spectral density of the responses x p (t) and x z (t), represents the cross-correlation power spectral density of the responses x q (t) and x z (t). The ratio of the cross-power spectral density eliminates the forced vibration term at the natural frequency ω m and finally converges to the ratio of the vibration mode amplitudes:

[0067]

[0068] Therefore, the power spectral density transmissibility method (at the natural frequency) is independent of the applied excitation and the reference response x z (i), represents that when the frequency is approximately equal to the natural frequency, φ pm and φ qm both represent the vibration mode amplitudes;

[0069] When the power spectral density transmissibility method is measured at L positions to obtain the responses, a power spectral density transmissibility method matrix is constructed:

[0070]

[0071] Among them, represents the cross-correlation power spectral density of the responses x L (t) and x L (t), represents the cross-correlation power spectral density of the responses x q (t) and x L (t); The power spectral density transmissibility method matrix has a unique property at the natural frequency, that is, the rank is 1, and its columns are linearly correlated. Therefore, the modal frequency is identified by performing singular value decomposition on this matrix.

[0072] The further improvement of the modal identification method for offshore wind turbines based on hybrid modal identification in the present invention lies in that, in order to avoid the appearance of false modes, the matrix of the power spectral density transmissibility method is reconstructed, and the reconstructed matrix of the power spectral density transmissibility method is as follows:

[0073]

[0074] Wherein, [T q (ω)] ++ represents the reconstructed matrix of the power spectral density transmissibility method, represents the matrix T q (ω) of the p-th singular value, represents the matrix T q (ω) of the p-th right singular vector, represents the matrix T q (ω) of the p-th left singular vector, o represents the number of singular values used for summation; by constructing L reconstructed matrices of the power spectral density transmissibility method, performing singular value decomposition, taking the first singular value, and finally performing weighted averaging to obtain the weighted average function π(ω):

[0075]

[0076] Wherein, represents represents the matrix T q (ω) of the 1st singular value;

[0077] Finally, the structural physical modal information of the system is screened out from the results of the eigensystem realization algorithm according to the weighted average function of formula (20).

[0078] The present invention further includes a readable storage medium storing a computer program, and the computer program is adapted to be loaded and executed by a processor to perform the above-mentioned modal identification method for offshore wind turbines based on hybrid modal identification.

[0079] The present invention further includes a computer device, the computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, it runs the above-mentioned modal identification method for offshore wind turbines based on hybrid modal identification.

[0080] Applying the technical solution of the present invention has the following beneficial effects:

[0081] The modal identification method for an offshore wind turbine based on hybrid modal identification of the present invention obtains the measured acceleration response; determines the harmonic frequency and extracts the harmonic component response by the Stochastic Subspace - Kalman Filter method; obtains the acceleration response after removing the harmonic component through orthogonal projection and LQ decomposition; then applies it to the improved Natural Excitation Technique - Eigensystem Realization Algorithm to obtain the modal parameter results; the acceleration response after removing the harmonic component is also applied to the Power Spectral Density Transfer Rate method to eliminate the influence of colored noise; finally, the correct structural physical mode is selected from the results of the improved Natural Excitation Technique - Eigensystem Realization Algorithm according to the modal frequency results of the Power Spectral Density Transfer Rate method. The method of the present invention first removes the influence of harmonic components on modal identification by using the Stochastic Subspace - Kalman Filter method, and then eliminates the influence of colored noise on modal identification according to the Power Spectral Density Transfer Rate method, and finally screens out the correct modal parameters from the results of the improved Natural Excitation Technique - Eigensystem Realization Algorithm. This hybrid modal identification framework has strong adaptability and effectiveness for the modal identification of offshore wind turbines under complex excitation compared with single methods, and solves the technical problem that the existing modal identification methods are limited when applied to offshore wind turbines and cannot accurately identify their modal parameters.

[0082] In addition to the purposes, features and advantages described above, the present invention has other purposes, features and advantages. The following will refer to the drawings for a further detailed description of the present invention. Brief Description of the Drawings

[0083] The drawings forming a part of this application are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0084] Figure 1 is the flowchart of the modal identification method for an offshore wind turbine based on hybrid modal identification of the present invention;

[0085] Figure 2 is the power spectrum diagram before removing the harmonic component of the modal identification method for an offshore wind turbine based on hybrid modal identification of the present invention;

[0086] Figure 3 is the power spectrum diagram after removing the harmonic component of the modal identification method for an offshore wind turbine based on hybrid modal identification of the present invention;

[0087] Figure 4 is the schematic diagram of the results of TMC - NExT - ERA of the modal identification method for an offshore wind turbine based on hybrid modal identification of the present invention;

[0088] Figure 5 is the schematic diagram of the results after passing through PSDT of the modal identification method for an offshore wind turbine based on hybrid modal identification of the present invention;

[0089] Figure 6 It is a schematic diagram of the TMC-NExT-ERA result after PSDT screening for the modal identification method of an offshore wind turbine based on hybrid modal identification of the present invention. Detailed implementation manners

[0090] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0091] See Figure 1 As shown, a modal identification method of an offshore wind turbine based on hybrid modal identification includes the following steps:

[0092] S1. Obtain the acceleration vibration response of the target structure through the sensors of the actual measurement platform;

[0093] S2. Determine the harmonic frequency of the acceleration vibration response in S1, and then use SI-KF (Stochastic Subspace - Kalman Filtering Method) to remove the harmonic components in the response to obtain the acceleration response after removing the harmonic components, as Figure 2 and Figure 3 shown;

[0094] S3. Apply the acceleration response after removing the harmonic components to the modal identification of TMC-NExT-ERA (Improved Natural Excitation Technique - Eigensystem Realization Algorithm) to obtain the modal frequency result of TMC-NExT-ERA, as Figure 4 shown; Apply the acceleration vibration response after removing the harmonic components to the frequency identification of PSDT (Power Spectral Density Transfer Rate Method) to obtain the structural frequency after removing the colored noise, as Figure 5 shown;

[0095] S4. Compare the modal frequency result of TMC-NExT-ERA and the structural frequency after removing the colored noise to determine the correct structural frequency;

[0096] S5. Extract the correct structural physical modal information from TMC-NExT-ERA based on the correct structural frequency obtained in S4 to obtain the correct modal parameters of the wind turbine, as Figure 6 shown.

[0097] As Figure 1 shown, the specific process is: perform power spectral analysis and determine the harmonic frequency of the acceleration vibration response of the structure, then remove the harmonic components through SI-KF and retain the system response. Through the modal identification of TMC-NExT-ERA, obtain the dual-reference-point Monte Carlo stability diagram, and at the same time, through the frequency identification of PSDT, obtain the average weight function. Finally, obtain the modal parameters through the dual-reference-point Monte Carlo stability diagram and the average weight function.

[0098] Figure 2 andFigure 3 wherein, the PSD (Power Spectral Density) is the vertical coordinate, the frequency is the horizontal coordinate, and the blue curve and the red curve are the power spectral densities of two acceleration responses in mutually orthogonal directions in the fifth degree of freedom. Figure 2 The sky-blue area in [Figure] is the harmonic component, and the light orange area is the colored noise component. Figure 2 The sky-blue area in [Figure] is the harmonic component, Figure 3 The red vertical line in [Figure] is the state where the harmonics have been removed. Figure 4 In [Figure], the number of calculations is the left vertical coordinate, the PSD (Power Spectral Density) is the right vertical coordinate, the frequency is the horizontal coordinate, the blue curve and the red curve are the power spectral densities of two acceleration responses in mutually orthogonal directions in the fifth degree of freedom in the fan, the red dots are stable results, the light pink dots are unstable false results, the light orange vertical area is the colored noise component, the orange horizontal area is the area where the number of calculations is less than 50, and the light blue part is the area where the number of calculations is greater than 50. Figure 5 In [Figure], the four blue dots are the determined first four natural frequencies, the red vertical line is the state where the harmonics have been removed, and at this time, the colored noise component in the light orange area is eliminated. Figure 6 In [Figure], the damping ratio is the vertical coordinate, the frequency is the horizontal coordinate, the blue circles are the ERA first-order modal clustering sets, the orange circles are the ERA second-order modal clustering sets, the yellow circles are the ERA third-order modal clustering sets, the purple circles are the ERA fourth-order modal clustering sets, and the green circles are the ERA fifth-order modal clustering sets.

[0099] Preferably, S2 specifically includes:

[0100] S201. Determine the acceleration vibration response and the magnitudes of its harmonic frequencies;

[0101] S202. Based on the acceleration vibration response, obtain the state matrix A, the observation matrix C, and the three covariance matrices R, S, and Q through the stochastic subspace method;

[0102] S203. Use the acceleration vibration response, the determined magnitudes of the harmonic frequencies, the state matrix, the observation matrix, and the three covariance matrices to obtain the harmonic component response through the Kalman filtering method;

[0103] S204. Decompose the harmonic component response through the spatial orthogonal projection and matrix decomposition method to obtain the acceleration response with the harmonic components removed.

[0104] Preferably, the calculation steps of the state matrix A and the observation matrix C include:

[0105] Let the acceleration vibration response be the observed data y k , based on the observed data y k a Hankel matrix Y of past time steps can be constructed past ∈l·i×j and Yfuture ∈ l·i × j, from which the projection matrix O can be obtained as follows:

[0106]

[0107] where O i ∈ l·i × l·i, i represents the number of rows of the defined Hankel matrix block, and... + represents that there are l(i + 1) rows, and... - represents that there are l(i - 1) rows, j = N - i + 1 is the number of block columns, N represents the length of the observation sequence, and Y future represents the Hankel matrix of future time steps, def indicates that the projection matrix O is obtained from the currently defined equation, and l is the number of sensors;

[0108] Performing singular value decomposition on O i gives:

[0109] O i = USV T 2);

[0110] where T represents the transpose of the matrix, U represents the left matrix after decomposition, V represents the right matrix after decomposition, and S represents the diagonal matrix of singular values;

[0111] Based on formula 2), the extended observability matrix is further obtained:

[0112]

[0113] where U1 and S1 are the first n columns of U and S, and n represents the selected order; Γ i-1 is the extended observability matrix of Γ i excluding the last l rows; then the state sequence can be obtained from formula 4):

[0114]

[0115] where represents the state sequence of the number of rows of the i-th matrix block, represents the state sequence of the number of rows of the (i + 1)-th matrix block, represents the Moore-Penrose pseudoinverse of Γ i , represents the Moore-Penrose pseudoinverse excluding the last l rows;

[0116] Matrices A and C can be obtained from formula 5):

[0117]

[0118] where Y i|iDenote a Hankel matrix with only one row block.

[0119] Preferably, the covariance matrices Q, S, and R can be obtained from the residuals:

[0120]

[0121] where (.) T denotes the transpose of a matrix, E denotes the expectation, and ρ w and ρ v are both Kalman filter residuals.

[0122] Preferably, S203 specifically includes the following steps:

[0123] Estimate the state sequence of each order based on the Kalman filter and the system state equation, and the optimal state estimate is as follows:

[0124]

[0125] where denotes the optimal estimate at the next time step, denotes the state at the current time step, A denotes the state matrix, and K k denotes the Kalman filter, and define e k is obtained from Equation (8);

[0126] The observation equation can be expressed as:

[0127]

[0128] where y k denotes the observed data, and C denotes the observation matrix;

[0129] The update of the Kalman gain K k can be expressed as:

[0130] K k =(AP k C T +S)(R + CP k C T ) -1 (9);

[0131] where P k denotes the covariance matrix of the Kalman filter;

[0132] The update of the covariance matrix can be expressed as:

[0133] P k+1 =AP k A T +Q - K k (AP k C T+S) T 10);

[0134] Wherein, P k+1 represents the covariance matrix of the estimated next time step Kalman filter;

[0135] The estimated state and matrix C are linearly transformed based on an invertible matrix V to obtain:

[0136]

[0137] Wherein, (.) V represents a linear transformation based on the invertible matrix V, and (.) -1 represents the inverse of the matrix;

[0138] This matrix is defined as a modal basis for distinguishing the modal components and harmonic components of the system and is constituted as follows:

[0139]

[0140] Wherein, represents taking the real part, represents taking the imaginary part, and Ψ represents the eigenvector of matrix A;

[0141] Define a position matrix s for the localization of harmonic components:

[0142]

[0143] Wherein, m represents the order number of the system mode, h represents the number of harmonic modes, and I represents the identity matrix;

[0144] Based on formula 11) and formula 13), the harmonic component response is calculated by the following formula:

[0145]

[0146] Wherein, l is the number of sensors.

[0147] Preferably, the TMC-NExT-ERA method in S3 is specifically as follows: First, the acceleration response after removing the harmonic components is used as the input of the TMC-NExT-ERA method. Then, the key parameter values of NExT-ERA (Natural Excitation Technique - Eigensystem Realization Algorithm) are obtained according to the double reference point Monte Carlo stabilization diagram. And subsequently, a result set is preliminarily obtained based on the double reference point Monte Carlo stabilization diagram and the fuzzy C-means clustering method (fuzzy clustering method). Finally, the modal parameter results of ERA (Eigensystem Realization Algorithm) are obtained.

[0148] Preferably, the PSDT method in S3 specifically includes the following steps:

[0149] The transfer rate function is the ratio of two dynamic responses x p (t) and x q (t), and the power spectral transfer rate is the ratio of the cross-power spectral density with respect to the reference responses x z (i) and x p (t), x q (t), as shown below:

[0150]

[0151] where, denotes the cross-correlation power spectral density of the responses x p (t) and x z (t), denotes the cross-correlation power spectral density of the responses x q (t) and x z (t). The ratio of the cross-power spectral densities eliminates the forced vibration term at the natural frequency ω m and finally converges to the ratio of the vibration mode amplitudes:

[0152]

[0153] Therefore, PSDT (at the natural frequency) is independent of the applied excitation and the reference response x z (i), denotes that when the frequency is approximately equal to the natural frequency, φ pm and φ qm both represent the vibration mode amplitudes;

[0154] When the PSDT of the responses is measured at L positions, a PSDT matrix is constructed:

[0155]

[0156] where, denotes the cross-correlation power spectral density of the responses x L (t) and x L (t), denotes the cross-correlation power spectral density of the responses x q (t) and x L (t); the PSDT matrix has a unique property at the natural frequency, i.e., the rank is 1 and its columns are linearly dependent. Therefore, the modal frequencies are identified by performing singular value decomposition on this matrix.

[0157] Preferably, to avoid false modes, the PSDT matrix is reconstructed, and the reconstructed PSDT matrix is as follows:

[0158]

[0159] Among them, [T q (ω)] ++ represents the matrix of the reconstructed power spectral density transfer rate method, represents the matrix T q (ω)'s p-th singular value, represents the matrix T q (ω)'s p-th right singular vector, represents the matrix T q (ω)'s p-th left singular vector, o represents the number of singular values used for summation; by constructing L reconstructed PSDT matrices, performing singular value decomposition, taking the first singular value, and finally performing weighted averaging to obtain the weighted average function π(ω):

[0160]

[0161] Among them, represents the matrix T q (ω)'s first singular value;

[0162] Finally, the structural physical modal information of the system is screened from the ERA results according to the weighted average function of formula (20), and the correct modal parameters of the wind turbine are obtained.

[0163] The present invention also includes a readable storage medium storing a computer program, and the computer program is adapted to be loaded and executed by a processor to perform the above-mentioned method for modal identification of an offshore wind turbine based on hybrid modal identification.

[0164] The present invention also includes a computer device, the computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, it runs the above-mentioned method for modal identification of an offshore wind turbine based on hybrid modal identification.

[0165] The modal identification method of an offshore wind turbine based on hybrid modal identification in the present invention includes obtaining measured acceleration responses, determining harmonic frequencies and extracting harmonic component responses through the SI-KF method, obtaining the acceleration responses with harmonic components removed through orthogonal projection and LQ decomposition, then applying them to TMC-NExT-ERA to obtain modal parameter results, applying the acceleration responses with harmonic components removed to the PSDT method simultaneously to eliminate the influence of colored noise, and finally screening and obtaining the correct structural physical modes from TMC-NExT-ERA according to the frequency results of PSDT. The method of the present invention first removes the influence of harmonic components on modal identification through the SI-KF method, then eliminates the influence of colored noise on modal identification according to the PSDT method, and finally screens out the correct modal parameters from the results of TMC-NExT-ERA. This hybrid modal identification framework has strong adaptability and effectiveness for the modal identification of offshore wind turbines under complex excitation compared with single methods, and solves the technical problem that the existing modal identification methods are limited when applied to offshore wind turbines and cannot accurately identify their modal parameters.

[0166] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for modal identification of offshore wind turbines based on hybrid modal identification, characterized in that: The steps include: S1. Obtain the acceleration vibration response of the target structure through the sensor of the measurement platform; S2, determining the harmonic frequency of the acceleration vibration response in S1, and then using the random subspace-Kalman filter method to remove the harmonic components in the response to obtain the acceleration response with the harmonic components removed; S3. The acceleration response with harmonic components removed is used for modal identification of the improved natural excitation technology-characteristic system implementation algorithm to obtain the modal frequency results of the improved natural excitation technology-characteristic system implementation algorithm; the acceleration vibration response with harmonic components removed is used for frequency identification of the power spectral density transfer rate method to obtain the structural frequency after colored noise is removed; S4. Compare the modal frequency results of the improved natural excitation technology-characteristic system implementation algorithm with the structural frequency after removing the colored noise to determine the correct structural frequency; S5. Based on the correct structural frequency obtained in S4, the correct structural physical modal information is extracted from the improved natural excitation technology-characteristic system implementation algorithm to obtain the correct modal parameters of the fan.

2. The offshore wind turbine modal identification method based on hybrid modal identification according to claim 1 is characterized in that: S2 specifically includes: S201, determining the acceleration vibration response and its harmonic frequency magnitude; S202, obtaining a state matrix A, an observation matrix C, and three covariance matrices R, S, and Q through a random subspace method based on the acceleration vibration response; S203, using the acceleration vibration response, the determined harmonic frequency magnitude, the state matrix, the observation matrix and the three covariance matrices to obtain the harmonic component response through the Kalman filtering method; S204, decomposing the harmonic component response by using a spatial orthogonal projection and matrix decomposition method to obtain an acceleration response with the harmonic component removed.

3. The offshore wind turbine modal identification method based on hybrid modal identification according to claim 2 is characterized in that: The calculation steps of the state matrix A and the observation matrix C include: Assume the acceleration vibration response is the observed data y k , based on the observed data y k Construct the Hank matrix Y of the past time steps past ∈l·i×j and the Hank matrix Y of the future time step future ∈l·i×j, the projection matrix O is obtained as follows: Among them, O i ∈l·i×l·i, i represents the number of rows of the defined Hank matrix block. + Indicates that there are l(i+1) blocks of rows. - indicates that there are l(i-1) blocks with rows, j=N-i+1 indicates the number of blocks with columns, N indicates the length of the observation sequence, def indicates that the projection matrix O is obtained by the currently defined equation, and l is the number of sensors; O i Perform singular value decomposition to obtain: A i =USV T 2); Where T represents the transpose of the matrix, U represents the decomposed left matrix, V represents the decomposed right matrix, and S represents the singular value diagonal matrix; Based on formula 2), the extended observability matrix is ​​further obtained: Among them, U1 and S1 represent the first n columns of U and S, and n represents the order of selection; Γ i-1 Represents Γ i Remove the last l rows of the extended observability matrix; then the state sequence From formula 4), we get: in, The state sequence representing the number of rows of the i-th matrix block, is the state sequence representing the number of rows of the i+1th matrix block, Represents Γ i The Moore-Penrose pseudoinverse, express Remove the Moore-Penrose pseudoinverse of the last l rows; The matrices A and C are obtained by formula 5): Among them, Y i|i Represents a Hank matrix with only one row of blocks.

4. The offshore wind turbine modal identification method based on hybrid modal identification according to claim 3 is characterized in that: The covariance matrices Q, S, and R are obtained from the residuals: Where E represents the expectation, ρ w and ρ v Both represent Kalman filter residuals.

5. The offshore wind turbine modal identification method based on hybrid modal identification according to claim 2 is characterized in that: S203 specifically includes the following steps: Based on Kalman filtering and system state equation estimation, the state sequence of each order is obtained, and the optimal state estimation is as follows: in, represents the optimal estimate for the next time step, represents the state of the current time step, A represents the state matrix, K k represents Kalman filtering, and defines e k From formula 8) we get: The observation equation is expressed as: Where C represents the observation matrix; Kalman gain K k The update is expressed as: K k =(AP k C T +S)(R+CP k C T ) -1 9); Among them, P k represents the covariance matrix of the Kalman filter; The update of the covariance matrix is ​​expressed as: P k+1 =AP k A T +QK k (AP k C T +S) T 10); Among them, P k+1 Represents the estimated covariance matrix of the Kalman filter for the next time step; The estimated state and matrix C are linearly transformed based on a reversible matrix V to obtain: in,(.) V Represents a linear change based on the reversible matrix V, (.) -1 represents the inverse of a matrix; This matrix is ​​defined as a modal basis to distinguish the modal and harmonic components of the system and is constructed as follows: in, represents the real part, represents the imaginary part, Ψ represents the eigenvector of matrix A; Define a position matrix s for the location of harmonic components: Where m represents the order of the system mode, h represents the number of harmonic modes, and I represents the identity matrix; Based on equation 11) and equation 13), the harmonic component response Calculated by the following formula:

6. The offshore wind turbine modal identification method based on hybrid modal identification according to claim 1, characterized in that: The improved natural excitation technology in S3 - the characteristic system implementation algorithm is specifically as follows: Firstly, the acceleration response with harmonic components removed is used as the input of the improved natural excitation technology-characteristic system implementation algorithm method; Then, the key parameter values ​​of the natural excitation technology-characteristic system implementation algorithm are obtained based on the dual reference point Monte Carlo stability diagram; Then, a preliminary result set was obtained based on the double reference point Monte Carlo stability diagram and fuzzy clustering algorithm; Finally, the modal parameter results of the characteristic system implementation algorithm are obtained.

7. The offshore wind turbine modal identification method based on hybrid modal identification according to claim 1, characterized in that: The power spectral density transfer rate method in S3 specifically includes the following steps: The transmissibility function is the two dynamic responses x p (t) and x q (t), then the power spectrum transfer rate is relative to the reference response x z (i) and x p (t), x q The ratio of the cross power spectral density of (t) is as follows: in, Represents the response x p (t) and x z The cross-correlation power spectral density of (t) is Represents the response x q (t) and x z (t), the ratio of the cross-correlation power spectrum density makes the natural frequency ω m At , the forced vibration term is eliminated, and finally converges to the ratio of the vibration mode amplitudes: So the power spectral density transfer rate method is for the applied excitation and the reference response x z (i) is independent, When the frequency is approximately equal to the natural frequency, φ pm and φ qm All represent vibration mode amplitudes; When the power spectrum density transfer rate method of the response is measured at L positions, the power spectrum density transfer rate method matrix is ​​constructed: in, Represents the response x L (t) and x L The cross-correlation power spectral density of (t) is Represents the response x q (t) and x L The cross-correlation power spectral density of (t), the power spectral density transfer rate method matrix has a unique property at the natural frequency, that is, the rank is 1, and its columns are linearly correlated, so the modal frequency is identified by performing singular value decomposition on the matrix.

8. The offshore wind turbine modal identification method based on hybrid modal identification according to claim 7 is characterized in that: In order to avoid the occurrence of false modes, the power spectrum density transfer rate method matrix is ​​reconstructed. The reconstructed power spectrum density transfer rate method matrix is ​​shown as follows: Among them, [T q (ω)] ++ represents the reconstructed power spectral density transfer rate matrix, Represents the matrix T q The pth singular value of (ω), Represents the matrix T q The pth right singular vector of (ω), Represents the matrix T q The pth left singular vector of (ω), o represents the number of singular values ​​used for summation; by constructing L reconstructed power spectral density transfer rate matrices, performing singular value decomposition, taking the first singular value, and finally performing weighted averaging to obtain the weighted average function π(ω): in, Represents the matrix T q The first singular value of (ω); Finally, the structural physical modal information of the system is screened out from the characteristic system implementation algorithm results according to the weighted average function of formula 20).

9. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor and executing the offshore wind turbine modal identification method based on hybrid modal identification according to any one of claims 1 to 8.

10. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the offshore wind turbine modal identification method based on hybrid modal identification according to any one of claims 1 to 8 is run.

Images