Early warning method for wind turbine blade flutter based on low-frequency noise

Abstract

This invention relates to the field of wind power generation technology, and in particular to a method for early warning of wind turbine blade flutter based on low-frequency noise. First, the geometric model information and physical parameters of the wind turbine blade are acquired. Then, the modal frequencies and mode shapes of each order of the blade are calculated. According to the superposition principle of vibration modes, blade flutter will inevitably radiate low-frequency noise at the corresponding modal frequencies. Microphones installed on the wind turbine are used to monitor wind force and low-frequency noise generated during operation in real time. The noise intensity at the blade flutter frequency is obtained and monitored through signal processing, and an early warning intensity threshold is set. When the low-frequency noise signal at the flutter modal frequency exceeds the early warning threshold, a flutter warning is immediately issued. This allows for early warning before significant blade deformation occurs, greatly improving the safety margin.

Description

Technical Field

[0001] This invention relates to the field of wind power generation technology, and in particular to a method for early warning of blade flutter in wind turbines based on low-frequency noise. Background Technology

[0002] In recent years, with the trend towards larger wind turbines, blade sizes have been increasing. Under the premise of unchanged manufacturing processes and materials, this inevitably leads to increased blade flexibility and decreased rigidity, gradually increasing the likelihood of flutter during operation. Currently, flutter is a critical issue that must be considered in the design and manufacturing of large-diameter blades (hundred-meter class). With the increasing installed capacity of large wind turbines, early warning of flutter in operating wind turbine blades is a crucial problem that urgently needs to be solved in engineering.

[0003] With the increasing prominence of flutter problems in engineering applications, scholars both domestically and internationally have conducted extensive research on this issue. They have discovered that flutter is a self-excited vibration phenomenon in which the blade experiences continuous amplitude decay due to the coupling effects of aerodynamic forces, elastic forces, inertial forces, and damping forces in the airflow. When a blade flutters, the amplitude generally increases continuously across three vibration modes: flapping, flaring, and torsion, until the blade fails. Currently, most monitoring of blade flutter utilizes the vibration and torsional characteristics of the blade. One detection method based on the blade tip torsional angle proposes performing Empirical Mode Decomposition (EMD) on the original torsional vibration time series of the blade tip, and then using the first intrinsic mode function (1... st The flutter mode (IMM) is defined as the flutter mode of a blade. The envelope process of the flutter mode is calculated through Hilbert transformation, which ultimately determines the onset time of flutter and enables monitoring and early warning. This method is based on real-time measurement data of the blade tip torsion angle and relies on strain gauge or grating sensors arranged on the blade to monitor the blade torsion angle. When an early warning is issued, the blade has already begun to twist and deform at a large angle. Another method that uses the blade root reaction force to monitor the blade has confirmed that when the blade flutters, the blade root reaction force begins to diverge. The two are correlated, and the blade root reaction force-flutter index can be defined to determine whether the blade has fluttered. Based on this, when an early warning is issued, the blade has already undergone a large-scale vibration deformation.

[0004] Currently, the above methods are applied in blade production and design, and all of them require the placement of sensors on the blades, which is quite complicated in actual operation. It is difficult to monitor and warn of flutter on operating wind turbine blades. Moreover, existing monitoring schemes can only issue warning information after flutter occurs and the blade begins to undergo significant torsion or vibration deformation, leaving little time for subsequent warnings and necessary protective measures, resulting in insufficient safety margin.

[0005] 1) Patent application No. 201310223629.8 discloses a method and system for monitoring the safety status of wind turbines, mainly for early warning of blade tip sweeping across the tower. It detects the noise level of the blades passing over the tower using an acoustic emission device mounted on the tower and correlates it with the safe distance between the blades and the tower. Noise and distance safety thresholds are then set. When the noise exceeds the safety threshold, it means the distance between the blades and the tower is less than the safety threshold, and the system issues an early warning. This patent utilizes the principle that the sound pressure amplitude increases inversely when the distance between the sound source and the sensor decreases. It can monitor the distance from the blade tip to the tower but cannot monitor blade flutter.

[0006] 2) Patent application No. 202311416928.3 discloses a method, system and medium for determining the critical speed of aeroelastic flutter of flexible blades. By obtaining the parameters of wind turbine blades, the aerodynamic load is calculated. Then, the blade is equivalent to a cantilever beam structure. Based on the aerodynamic load of the blade and the finite element spatial beam model, the flutter speed and flutter frequency are obtained, which provides a guarantee for the aeroelastic design of flexible blades.

[0007] 3) Patent application No. 201810258984.1 discloses a method and device for calculating the critical flutter wind speed of a wind turbine blade airfoil. The method establishes the aeroelastic motion equation of the wind turbine airfoil by obtaining the airfoil parameters, and then further establishes the energy balance equation of the blade airfoil. After analyzing the equation, the flutter wind speed is obtained.

[0008] The above technical approach considers how to obtain flutter wind speed, flutter frequency and flutter rotation speed during the design phase of large flexible blades, and how to use the magnitude of acoustic signals to warn of blade sweep risk, but it does not solve the problem of early warning of possible blade flutter risk in actual wind turbine operation. Summary of the Invention

[0009] The purpose of this invention is to overcome the shortcomings of the prior art and provide a wind turbine blade flutter early warning method based on low-frequency noise. This method can issue an early warning as soon as blade flutter occurs during actual wind turbine operation, thereby improving the safety of wind turbine operation.

[0010] To achieve the above objectives, the present invention adopts the following technical solution:

[0011] A method for early warning of wind turbine blade flutter based on low-frequency noise, comprising:

[0012] 1) Obtain the blade parameters of the wind turbine generator set;

[0013] 2) Based on the wind turbine blade parameters obtained in step 1), establish a finite element model of the blade and calculate the mode shape and modal frequency of the blade;

[0014] 2.1 Establish a three-dimensional geometric model of the blades based on the obtained wind turbine blade parameters;

[0015] 2.2 The three-dimensional geometric model is equivalent to an idealized cantilever beam structure;

[0016] 2.3 Establish a set of nonlinear coupled partial differential equations (PDEs); the set of nonlinear coupled partial differential equations includes the set of control differential equations for the aerodynamic force L and pitch M acting on the blade and the set of control differential equations for the pitch M and aerodynamic force L acting on the 2D airfoil under aerodynamic loads.

[0017] Hamilton's principle formula (1) is:

[0018] (1)

[0019] In equation (1), t1 is the starting time of the blade motion, t2 is the ending time of the blade motion, δU is the variation of strain energy, δT is the variation of kinetic energy, and δW is the virtual work done.

[0020] In the modeling of the rotating blade, the motion in both flapping and torsional directions is considered. Using Hamilton's principle formula (1), the control differential equations (2) and (3) for the aerodynamic force L and the variable pitch M acting on the blade are derived.

[0021] (2)

[0022] (3)

[0023] In formulas (2) and (3), w and Let w' be the first-order partial derivative of the waving displacement with respect to the spatial coordinate x, and w' be the waving displacement with respect to the spatial coordinate x. w" Let x be the second-order partial derivative of the wave displacement with respect to the spatial coordinate x. For the second-order partial derivative of the wave displacement with respect to time, This is the first-order partial derivative of the torsional displacement with respect to the torsional direction. G is the angular acceleration, G is the torsional stiffness, J is the torsional moment of inertia at the blade interface, GJ is the torsional stiffness, E is the elastic modulus, I is the moment of inertia in the flapping direction, EI is the flapping bending stiffness, Km is the polar radius of gyration about the elastic axis, and K... m1 and K m2 , respectively, are the axial mass gyration radii about the main neutral axis and perpendicular to the elastic axis, m is the mass per unit length of the blade, ρ is the density of the blade, A is the cross-sectional area of ​​the blade, e is the centroid offset, Ω is the rotor speed, the lower limit of integration x is the distance from the blade's spanwise calculated section to the rotor center, R is the blade's spanwise length, L is the aerodynamic force, and M is the pitch.

[0024] Based on Theodosen theory, the governing differential equations for the pitch moment M and aerodynamic force L under aerodynamic loads on a 2D airfoil are derived, as shown in equations (4) and (5):

[0025] (4)

[0026] (5)

[0027] In formulas (4) and (5), a is the distance between the elastic axis and the midpoint of the chord, b is the half-chord length, u is the incoming flow velocity, and C is the distance between the elastic axis and the midpoint of the chord. La ρ is the lift coefficient. ∞ Let C(k) = F(k) + iG(k) be the air density, and let C(k) = F(k) + iG(k) be the Theodosian function, where... , f f The natural frequency of the airfoil, i Let F(k) be the imaginary unit and G(k) be the real and imaginary parts of the Theodosian function, respectively. and These are the first derivatives of the waving displacement and torsional displacement with respect to time, respectively. The meanings of the other symbols are the same as those in formulas (2) and (3).

[0028] 2.4 The idealized beam structure is spatially discretized using the finite element method (FEA) and converted into linearized equations;

[0029] Substituting equations (4) and (5) into equations (2) and (3), and combining this with the finite element method, the blade is considered as a variable cross-section linear Euler-Bernoulli beam, with three degrees of freedom at each node, namely w, , Discretize the structural region into 1D 2-node elements, and the shape function of each node's degree of freedom. w(x) is assumed to be a linear function and a cubic polynomial, respectively, as shown in formulas (6) and (7);

[0030] (6)

[0031] (7)

[0032] In formulas (6) and (7) N i e For the shape function of the waving direction, d i e For freedom of movement, P i e For the twist direction shape function, To indicate the degree of freedom, the superscript 'e' signifies 'element' in the finite element method. Let be the degree of freedom of the i-th node in swinging. Let be the slope of the swing freedom of the i-th node. Let represent the torsional degree of freedom of the i-th node.

[0033] 2.5 Perform linear analysis to obtain the mode shapes and modal frequencies of the blade.

[0034] Substituting equations (6) and (7) into equations (2) and (3) and performing weighted integration over the unit cell, we obtain the differential equation shown in equation (8).

[0035] (8)

[0036] Where, M s and M a These are the structural and aerodynamic mass matrices, respectively, where C is the aerodynamic damping matrix, and K is the aerodynamic mass matrix. s and K a For structural and aerodynamic stiffness matrices, and These are the second derivatives of the waving displacement and torsional displacement with respect to time, respectively. and The first derivatives of the waving displacement and torsional displacement with respect to time are given by w and , respectively. The eigenvalues ​​and eigenvectors are obtained by solving equation (8), which represents the flapping displacement and torsional displacement, respectively. The modal frequencies and mode shapes are then obtained. The modal frequencies are denoted as f1 in ascending order. <f2<......<f n The 4th to 6th order modal frequencies are selected as the monitoring objects.

[0037] Based on equation (8), we can further solve for the eigenvalues ​​and eigenvectors of the undamped system, as shown in equation (9):

[0038] Based on equation (8), we can further solve for the eigenvalues ​​and eigenvectors of the undamped system, as shown in equation (9):

[0039] (9)

[0040] Formula (9) K t In the middle is the overall stiffness matrix, M. t The total mass matrix, Given the natural angular frequency, solve the differential equation (9) to obtain the mode shapes and modal frequencies of the blade.

[0041] 3) Determine the measurement area for low-frequency noise, set up a noise sensor group, and acquire the noise signal when the wind turbine is working;

[0042] Based on the blade parameters of the wind turbine generator, the aerodynamic load distribution in the span direction is calculated using the modified blade element momentum theory. The range of the tower swept by the section with the largest aerodynamic load is set as the noise detection sensor placement area, and a low-frequency noise measurement sensor group is placed in this area.

[0043] 4) The noise signal measured by the sensor group is processed to obtain the low-frequency noise signal;

[0044] The low-frequency noise sensor group contains at least one sensor, and may also contain multiple sensors to form a test array. Sensor No. 1 is set as the reference microphone in the array. Cross-correlation analysis is performed on the noise signal measured by the microphone group to obtain the time delay of different sensors in the sensor group relative to the reference sensor. The low-frequency noise generated by the blade is enhanced by using delay accumulation technology. Finally, a low-pass filter is set to filter the enhanced sound signal to obtain the low-frequency noise. The upper limit frequency of the low-frequency noise can be set to 10 to 100 times the passing frequency of the blade.

[0045] 5) Set a low-frequency noise warning threshold at the blade flutter frequency;

[0046] Low-frequency noise monitoring is performed on wind turbine generators under normal operating conditions, and a database is established. A flutter warning threshold is set based on the noise intensity at the corresponding flutter frequency in the database. The size of the threshold value indicates the probability of triggering a flutter warning. A larger threshold value results in a lower probability of triggering a warning, and a smaller threshold value results in a higher probability of triggering a warning. If there is no noise corresponding to the flutter frequency in the low-frequency noise database established when the wind turbine generator is operating normally, the flutter warning threshold can be set to the logical value "none", which means that there is no low-frequency noise corresponding to the flutter frequency in the low-frequency noise database when the wind turbine is operating normally.

[0047] 6) Perform low-frequency noise monitoring on the wind turbine generator set in operation. When the noise amplitude at the flutter frequency exceeds the warning threshold, a flutter alarm will be issued.

[0048] The system performs time-frequency analysis on the detected low-frequency noise, monitors the noise amplitude at the flutter frequency in real time, and issues a flutter alarm when the noise amplitude at the flutter frequency exceeds the warning threshold; when set to logic "none", an alarm will be triggered if the flutter frequency occurs.

[0049] This invention addresses the challenges of early warning in operating wind turbine generators and the frequent blade flutter that leads to blade breakage and damage. First, it acquires the geometric model information and physical parameters of the blade. Then, it calculates the modal frequencies and mode shapes of the blade. Based on the superposition principle of vibration modes, blade flutter inevitably radiates low-frequency noise at the corresponding modal frequencies. Microphones mounted on the wind turbine tower are used to monitor the low-frequency noise generated during normal blade rotation in real time. Signal processing is used to obtain and monitor the noise intensity at the blade flutter frequency, and a warning threshold is set. When the low-frequency noise signal at the flutter modal frequency exceeds the warning threshold, a flutter warning is immediately issued. This allows for early warning before significant blade deformation occurs, greatly improving the safety margin. Attached Figure Description

[0050] Figure 1 A schematic diagram of a wind turbine blade flutter early warning system provided by the present invention;

[0051] Figure 2 A side view of a wind turbine blade flutter early warning system provided by the present invention;

[0052] Figure 3 A component analysis diagram of low-frequency aerodynamic noise generated by a wind turbine generator set during operation, provided by the present invention;

[0053] Figure 4 The flowchart of the blade flutter early warning method provided by the present invention;

[0054] Figure 5 This is a flowchart illustrating the practical application of low-frequency noise monitoring for blade flutter in this invention.

[0055] Figure 6 This is a diagram showing the distribution of lift and drag loads along the wingspan at a certain moment in this invention.

[0056] Figure 7 This is a flowchart of the sensor group testing low-frequency noise in this invention;

[0057] Figure 8 This is a three-dimensional blade model diagram established based on the obtained blade information in this invention;

[0058] Figure 9 This is a diagram showing the variation of the inflow velocity as defined in this invention;

[0059] Figure 10 This is a graph showing the monitoring results of the low-frequency noise time-domain signal calculated in this invention;

[0060] Figure 11 This is a graph showing the monitoring results of the low-frequency signal time-frequency diagram calculated in this invention;

[0061] Figure 12The graph shows the changes in blade load and rotor speed over time.

[0062] In the diagram: 1. Blade, 2. Tower, 3. Noise sensor placement area, 4. Low-frequency noise sensor. Detailed Implementation

[0063] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0064] The structures, proportions, and sizes illustrated in the accompanying drawings are merely for illustrative purposes and to aid those skilled in the art in understanding and reading the invention. They are not intended to limit the scope of the invention and therefore have no substantial technical significance. Any modifications to the structure, changes in proportions, or adjustments to size, provided they do not affect the effectiveness or purpose of the invention, should still fall within the scope of the technical content disclosed herein. Furthermore, the terms "upper," "lower," "left," "right," "middle," and "one" used in this specification are merely for clarity and not intended to limit the scope of the invention. Changes or adjustments to their relative relationships, without substantially altering the technical content, should also be considered within the scope of the invention's implementation.

[0065] Figure 3 When airflow acts on the blade, it generates aerodynamic loads on the blade surface. The aerodynamic load fluctuations directly radiate a portion of low-frequency aerodynamic noise. At the same time, the blade vibrates after the aerodynamic loads act on it. At this time, the blade can be equivalent to a cantilever beam structure. According to the principle of vibration mode superposition, the vibration of the blade can be decomposed into the superposition of various vibration modes, where k1, k2, ..., kn are the coefficients of the corresponding modes. Each vibration mode will also radiate low-frequency aerodynamic noise of the corresponding mode frequency f1, f2, ..., fn. Therefore, the low-frequency aerodynamic noise radiated by the wind turbine blade contains both the aerodynamic noise caused by load fluctuations and the aerodynamic noise caused by blade vibration.

[0066] Figure 4 In the process, when the blades flutter under aerodynamic loads, the blades move in a superposition of flutter mode 1, flutter mode 2, ..., flutter mode n. The mode coefficient kn represents the proportion of contribution of each mode in the flutter. At this time, the low-frequency noise must contain the frequency corresponding to the flutter mode with a large contribution.

[0067] Figure 5 In practical applications, the low-frequency noise signal radiated by the wind turbine is directly obtained by using a microphone. Time-frequency analysis of the low-frequency noise can be performed to obtain the low-frequency aerodynamic noise frequency components of the blade at a certain moment. Using the modal frequencies calculated in the early stage, the frequency of each flutter mode is monitored in real time to see if it appears or if there is a sudden change. If it does, an alarm is immediately triggered.

[0068] A method for early warning of wind turbine blade flutter based on low-frequency noise, comprising:

[0069] 1. For example Figure 1 , Figure 2 The blade parameters of the three-bladed wind turbine generator shown are obtained.

[0070] 2. Based on the obtained wind turbine blade parameters, obtain the mode shapes and modal frequencies of each blade according to the following steps;

[0071] 2-1- Taking the NREL 5MW wind turbine as an example, the blade parameters are obtained to establish a three-dimensional model of the blade, as follows: Figure 8 As shown;

[0072] 2-2- The three-dimensional geometric model is equivalent to an idealized cantilever beam structure; the blade is discretized into 19 blade element units in the span direction, and the cross-sectional characteristic parameters of each blade element unit are obtained through publicly available data, including the chord length, torsion angle, torsional moment of inertia J, flapping moment of inertia I, mass per unit length m, Young's modulus E, shear modulus G and material density ρ of each cross section.

[0073] 2-3-Establish a set of nonlinear coupled partial differential equations; the set of nonlinear coupled partial differential equations includes the control differential equations of the aerodynamic force L and pitch M acting on the blade and the control differential equations of the pitch M and aerodynamic force L acting on the 2D airfoil under aerodynamic loads.

[0074] Hamilton formula:

[0075] (1)

[0076] In equation (1), t1 is the starting time of the blade motion, t2 is the ending time of the blade motion, δU is the variation of strain energy, δT is the variation of kinetic energy, and δW is the virtual work done.

[0077] In the modeling of the rotating blade, the motion in both flapping and torsional directions is considered. The control differential equations (2) and (3) for the aerodynamic force L and the pitch M acting on the blade are derived using Hamilton's principle formula (1).

[0078] (2)

[0079] (3)

[0080] The boundary condition is as follows: at the leaf root x=0: At the tip of the leaf, x=R: .

[0081] In formulas (2) and (3), w and Let w' be the first-order partial derivative of the waving displacement with respect to the spatial coordinate x, and w' be the waving displacement with respect to the spatial coordinate x. w" Let x be the second-order partial derivative of the wave displacement with respect to the spatial coordinate x. For the second-order partial derivative of the wave displacement with respect to time, This is the first-order partial derivative of the torsional displacement with respect to the torsional direction. G is the angular acceleration, G is the torsional stiffness, J is the torsional moment of inertia at the blade interface, GJ is the torsional stiffness, E is the elastic modulus, I is the moment of inertia in the flapping direction, EI is the flapping bending stiffness, Km is the polar radius of gyration about the elastic axis, and K... m1 and K m2 , respectively, are the axial mass gyration radii about the main neutral axis and perpendicular to the elastic axis, m is the mass per unit length of the blade, ρ is the density of the blade, A is the cross-sectional area of ​​the blade, e is the centroid offset, Ω is the rotor speed, the lower limit of integration x is the distance from the blade's spanwise calculated section to the rotor center, R is the blade's spanwise length, L is the aerodynamic force, and M is the pitch.

[0082] Based on Theodosen theory, the governing differential equations for the pitch moment M and aerodynamic force L under aerodynamic loads on a 2D airfoil are derived, as shown in equations (4) and (5):

[0083] (4)

[0084] (5)

[0085] In formulas (4) and (5), a is the distance between the elastic axis and the midpoint of the chord, b is the half-chord length, u is the incoming flow velocity, and C is the distance between the elastic axis and the midpoint of the chord. La ρ is the lift coefficient. ∞ Let C(k) = F(k) + iG(k) be the air density, and let C(k) = F(k) + iG(k) be the Theodosian function, where... , f f The natural frequency of the airfoil, i Let F(k) be the imaginary unit and G(k) be the real and imaginary parts of the Theodosian function, respectively. and These are the first derivatives of the waving displacement and torsional displacement with respect to time, respectively. The meanings of the other symbols are the same as those in formulas (2) and (3).

[0086] 2-4- The idealized beam structure is spatially discretized using the finite element method (FEA) and converted into linearized equations;

[0087] Substituting equations (4) and (5) into equations (2) and (3), and combining this with the finite element method, the blade is considered as a variable cross-section linear Euler-Bernoulli beam, with three degrees of freedom at each node, namely w, , Discretize the structural region into 1D 2-node elements, and the shape function of each node's degree of freedom. w(x) is assumed to be a linear function and a cubic polynomial, respectively, as shown in formulas (6) and (7);

[0088] (6)

[0089] (7)

[0090] In formulas (6) and (7) N i e For the shape function of the waving direction, d i e For freedom of movement, P i e For the twist direction shape function, To indicate the degree of freedom, the superscript 'e' signifies 'element' in the finite element method. Let be the degree of freedom of the i-th node in swinging. Let be the slope of the swing freedom of the i-th node. Let represent the torsional degree of freedom of the i-th node.

[0091] 2-5- Perform linear analysis to obtain the mode shapes and modal frequencies of the blade:

[0092] Substituting equations (6) and (7) into equations (2) and (3) and performing weighted integration over the unit cell, we obtain the differential equation shown in equation (8). (8)

[0093] Where, M s and M a These are the structural and aerodynamic mass matrices, respectively, where C is the aerodynamic damping matrix, and K is the aerodynamic mass matrix. s and K a For structural and aerodynamic stiffness matrices, and These are the second derivatives of the waving displacement and torsional displacement with respect to time, respectively. and The first derivatives of the waving displacement and torsional displacement with respect to time are given by w and , respectively. The eigenvalues ​​and eigenvectors are obtained by solving equation (8), which represents the flapping displacement and torsional displacement, respectively. The modal frequencies and mode shapes are then obtained. The modal frequencies are denoted as f1 in ascending order. <f2<......<f n The 4th to 6th order modal frequencies are selected as the monitoring objects.

[0094] Based on equation (8), we can further solve for the eigenvalues ​​and eigenvectors of the undamped system, as shown in equation (9):

[0095] (9)

[0096] Formula (9) K t In the middle is the overall stiffness matrix, M. t The total mass matrix, Given the natural angular frequency, solve the differential equation (9) to obtain the mode shapes and modal frequencies of the blade. The results are shown in Table 1.

[0097] Table 1. Mode shapes and modal frequencies of the blades

[0098]

[0099] 3. Determine the measurement area for low-frequency noise, set up a noise sensor group, and acquire the noise signal when the fan is working;

[0100] Calculate the lift and drag load distribution along the wingspan, such as Figure 6 The measurement area is set in the region corresponding to the blade element unit with the greatest load, such as... Figure 1 2. Place a low-frequency noise sensor in the measurement area. The coordinates of the low-frequency noise sensor are (0,0,30) m.

[0101] 4. The noise signal measured by the sensor is processed to obtain a low-frequency noise signal;

[0102] The noise signal measured by the sensor was simulated using a computer in the laboratory. For a given NREL 5MW on-road wind turbine, the initial pitch angle and yaw angle are 0, the blades are in the attached flow state, and the pitch and yaw controller is disabled to allow the rotor to rotate freely. The inflow wind speed is set as follows: 0-400 s: the wind speed remains constant at 4.5 m / s; 400-700 s: the wind speed linearly increases from 4.5 m / s to 20 m / s; 700-999 s: the wind speed remains constant at 20 m / s. The changes in inflow wind speed are as follows: Figure 9 As shown.

[0103] Figure 1 , Figure 2 In the measurement area, the time-domain low-frequency noise signal measured by the low-frequency noise sensor at the coordinates (0,0,30) m is as follows: Figure 10 As shown.

[0104] 5. Set a low-frequency noise warning threshold at the blade flutter frequency;

[0105] The low-frequency noise of the wind turbine generator obtained from simulation calculations is monitored, and a database is established. A flutter warning threshold is set according to the noise intensity at the corresponding modal frequency in the database. The size of the threshold value represents the probability of triggering a flutter warning; a larger threshold value results in a lower probability of triggering a warning, and a smaller threshold value results in a higher probability of triggering a warning. If no noise corresponding to the flutter frequency appears in the low-frequency noise database established when the wind turbine generator is operating normally, the flutter warning threshold can be set to the logical value "none," indicating that there is no low-frequency noise corresponding to the flutter frequency in the low-frequency noise database when the wind turbine is operating normally. In this example, based on the low-frequency noise measured by the low-frequency noise sensors arranged in the measurement area, since low-order modes are easily triggered, the low-frequency noise warning threshold at the first-order flapwise mode frequency of 0.68 Hz is set to the logical value "none."

[0106] 6. Perform time-frequency monitoring on the operating wind turbine generator set, and issue a flutter alarm when the noise amplitude at the modal frequency exceeds the warning threshold.

[0107] Time-frequency analysis is performed on the detected low-frequency noise, and the noise amplitude at the modal frequency is monitored in real time. When the noise amplitude at the modal frequency exceeds the warning threshold, a flutter alarm is issued; when set to the logic value "none", an alarm is triggered if flutter frequency occurs. In this example, the time-frequency monitoring results are as follows: Figure 11 As shown, the low-frequency noise intensity at the first-order flapwise frequency begins to increase at approximately 650 s, and the characteristic of this frequency begins to appear in the time-frequency graph and gradually increases, triggering the warning condition.

[0108] This is compared with the conventional method of placing sensors on the blades to detect changes in blade load to determine flutter. Figure 12 To simulate the changes in lift load and rotor speed over time as measured by sensors placed at the blade tip, it can be seen that at 723s, the rotor speed no longer increases, but the blade load continues to increase and shows a diverging trend. Therefore, it is determined that the blade flutters at this moment and an early warning is issued.

[0109] Comparing the two, it can be found that the method proposed in this invention advances the flutter warning time from 723 seconds to 650 seconds, which is 73 seconds earlier and significantly improves the time margin of flutter warning.

[0110] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A low frequency noise based wind turbine blade flutter warning method, characterized in that, The method comprises the following steps: (a) obtaining wind turbine blade parameters; (b) obtaining wind turbine blade parameters according to step (a), establishing a finite element model of the blade, and calculating the modal shape and modal frequency of the blade; (c) determining a measurement area of low-frequency noise, arranging a noise sensor group, and obtaining noise signals of the wind turbine during operation; according to the wind turbine blade parameters, the maximum position section of the aerodynamic load distribution in the spanwise direction is calculated by using the corrected blade element momentum theory, and the range of the tower drum swept by the position section is set as the noise detection sensor arrangement area, and a low-frequency noise measurement sensor group is arranged in the area; (d) processing the noise signals measured by the sensors to obtain low-frequency noise signals; The low-frequency noise sensor group comprises at least one sensor or a plurality of test arrays; The noise signals measured by the microphone group are processed to obtain low-frequency noise, and the upper limit frequency of the low-frequency noise is set to 10-100 times the passing frequency of the blade; (e) setting a low-frequency noise early warning threshold at the blade flutter frequency; The wind turbine in the normal working state is monitored for low-frequency noise, a database is established, the flutter early warning threshold is set according to the noise intensity corresponding to the flutter frequency in the database, the size of the threshold value represents the size of the flutter triggering early warning probability, the threshold value is large, the triggering early warning probability is small, the threshold value is small, the triggering probability is large, if there is no noise corresponding to the flutter frequency in the low-frequency noise database established when the wind turbine is working normally, the flutter early warning threshold is set to a logical value "none", which represents that there is no low-frequency noise corresponding to the flutter frequency in the low-frequency noise database when the wind turbine is working normally; (f) performing time-frequency monitoring on the low-frequency noise of the working wind turbine, and issuing a flutter alarm when the noise amplitude at the flutter frequency exceeds the early warning threshold.

2. The low frequency noise based wind turbine blade flutter warning method of claim 1, wherein, In step (b), the modal shape and modal frequency of the blade are obtained according to the following steps: 2.1 establishing a three-dimensional geometric model of the blade according to the obtained wind turbine blade parameters; 2.2 equivalent the three-dimensional geometric model into an idealized cantilever beam structure; 2.3 establishing a nonlinear coupled partial differential equation set; 2.4 using the finite element method to discretize the idealized beam structure in space and converting it into a linearized equation; 2.5 performing linear analysis to obtain the modal shape and modal frequency of the blade.

3. The low frequency noise based wind turbine blade flutter warning method of claim 2, wherein, In step 2.3, the nonlinear coupled partial differential equation set includes the control differential equation set of the aerodynamic force L and the variable pitch moment M acting on the blade, and the control differential equation set of the variable pitch moment M and the aerodynamic force L acting on the 2D airfoil.

4. The low frequency noise based wind turbine blade flutter warning method of claim 3, wherein, In the modeling of the rotating blade, the waving and torsional motions in two directions are considered, and the control differential equation set (2) and (3) of the aerodynamic force L and the variable pitch moment M acting on the blade are derived by using the Hamilton principle formula: (2) (3) In the formulas (2) and (3), w and are respectively the flapwise displacement and the torsional displacement, w' is the first order partial derivative of the flapwise displacement with respect to the spatial coordinate x, w" is the second order partial derivative of the flapwise displacement with respect to the spatial coordinate x, is the second order partial derivative of the flapwise displacement with respect to time, is the first order partial derivative of the torsional displacement with respect to the torsional direction, is the angular acceleration, G is the torsional stiffness, J is the torsional moment of inertia of the blade interface, GJ is the torsional stiffness, E is the elastic modulus, I is the flapwise moment of inertia, EI is the flapwise bending stiffness, Km is the polar radius of gyration about the elastic axis, K m1 and K m2 are respectively the mass radii of gyration about the principal neutral axis and the axis perpendicular to the elastic axis, m is the mass per unit length of the blade, p is the density of the blade, A is the blade cross-sectional area, e is the centroid offset, Q is the wind turbine rotational speed, the lower limit of the integral x is the distance of the calculation section from the center of the wind turbine in the blade spanwise direction, R is the length of the blade in the spanwise direction, L is the aerodynamic force, and M is the pitch moment.

5. The low frequency noise based wind turbine blade flutter warning method of claim 4, wherein, The Hamilton principle formula (1) is: （1） In formula (1), t1 is the starting time of the blade motion, t2 is the ending time of the blade motion, δU is the variation of strain energy, δT is the variation of kinetic energy, and δW is the virtual work.

6. The low frequency noise based wind turbine blade flutter warning method of claim 4, wherein, The control differential equations of the variable moment M and aerodynamic force L of the aerodynamic load acting on the 2D airfoil are shown in equations (4) and (5) according to Theodosen theory. （4） （5） a is the distance of elastic axis and chord center, b is half chord length, u is the incoming flow velocity, C La is the lift coefficient, p ∞ is the air density, C(k) = F(k) + iG(k) is the Theodorsen function, where , f f is the natural frequency of the airfoil, i is the imaginary unit, F(k) and G(k) are the real and imaginary parts of the Theodorsen function, respectively, and are the first order derivatives of the flapping displacement and the torsional displacement with respect to time, respectively, and other symbols have the same meanings as in equations (2) and (3).

7. The low frequency noise based wind turbine blade flutter warning method of claim 5, wherein, The step 2.4 utilizes the finite element method to discretize the idealized beam structure in space, and converts it into a linear equation, specifically: put formulas (4) and (5) into formulas (2) and (3), combine the finite element method, and regard the blade as a variable cross-section linear Euler-Bernoulli beam, with 3 degrees of freedom at each node, namely w, , ; discretize the structure region into a 1D 2-node element, and assume the shape function of the degree of freedom of each node and w(x) to be a linear function and a cubic polynomial respectively, as shown in formulas (6) and (7); (6) (7) In equations (6), (7) N i e is the edgewise shape function, d i e is the edgewise degree of freedom, P i e is the torsional shape function, is the torsional degree of freedom, the superscript 'e' means 'element' in the finite element.

8. The low frequency noise based wind turbine blade flutter warning method of claim 6, wherein, The linear analysis of step 2.5 is carried out to obtain the modal shape and modal frequency of the blade, specifically, equations (6) and (7) are brought into equations (2) and (3) and weighted integration is carried out on the element to obtain the differential equation shown in equation (8), (8) where M s and M a are the structural and aerodynamic mass matrices, C is the aerodynamic damping matrix, K s and K a are the structural and aerodynamic stiffness matrices, and are the second order derivatives of the in-plane and torsional displacements with respect to time, and are the first order derivatives of the in-plane and torsional displacements with respect to time, w and are the in-plane and torsional displacements, the eigenvalues and eigenvectors are obtained by solving equation (8), and then the modal frequencies and modal shapes are obtained; the modal frequencies are recorded in the order of f1 n , and the 4th to 6th order modal frequencies are taken as the monitoring objects.

9. The low frequency noise based wind turbine blade flutter warning method of claim 8, wherein, The characteristic value and characteristic vector of the undamped system are solved according to equation (8), as shown in equation (9): The characteristic value and characteristic vector of the undamped system are further solved according to equation (8), as shown in equation (9): (9) In equation (9), K t is the total stiffness matrix, M t is the total mass matrix, is the natural angular frequency. The differential equation (9) is solved to obtain the mode shapes and modal frequencies of the blade.

10. The low frequency noise based wind turbine blade flutter warning method of claim 1, wherein, In step (d), the 1st sensor in the array is set as a reference microphone, cross-correlation analysis is carried out on the noise signals measured by the microphone group to obtain the time delay of different sensors in the sensor group relative to the reference sensor, delay accumulation technology is used to enhance the low-frequency noise generated by the blade, and finally a low-pass filter is set to carry out low-pass filtering on the enhanced sound signals to obtain the low-frequency noise.

Images