Balance correction method for wind driven generator rotor

Abstract

The invention discloses a balance correction method of a wind driven generator rotor, relates to the technical field of balance correction of wind driven generator rotors, and aims to solve the problem of dynamic imbalance caused by blade mass distribution difference. A traditional static correction method cannot meet the real-time adjustment requirement of a large-size turbine, and the real-time adjustment requirement of the large-size turbine is met by collecting spindle vibration, blade root strain and pitch angle signals, constructing a real-time rotational inertia vector, calculating mass inertia change and a pneumatic cycle lag index and inputting a graph neural network to determine an unbalance coefficient and type. Generating a correction instruction for movable counterweight and pitch adjustment; the correction process is synchronously executed under rated conditions, and coordination of mechanical compensation and pneumatic compensation is ensured. After correction, the evaluation effect is compared through a vibration vector, and the result is stored in a self-adaptive database to optimize subsequent circulation; low-load operation can be maintained, the maintenance cost is reduced, and the service life of key parts is prolonged.

Description

Technical Field

[0001] The present invention relates to the technical field of wind generator rotor balance correction, and in particular to a wind generator rotor balance correction method. Background Art

[0002] During wind turbine operation, the mass distribution of rotor blades may vary due to factors such as manufacturing tolerances, material non-uniformity, or installation errors, leading to rotor imbalance. Traditional rotor balancing methods rely on static mass measurement and manual weighting, which makes it difficult to accurately and timely address dynamic changes in mass distribution during operation. This static correction method may not effectively eliminate dynamic imbalance caused by uneven mass distribution, which in turn affects the efficiency and lifespan of the wind turbine.

[0003] After searching, the patent application with the application publication number CN102539071B relates to an online dynamic balancing method for a fan rotor based on the principle of parallel decomposition of force, which includes the following steps: (1) using the three-circle opening method to determine the type of fan rotor imbalance: on the side of the shaft diameter where the fan vibration is larger, use a "scratching needle" to draw three circles at the shaft diameter between the casing and the bearing seat, with the distance between the circles selected to be 2mm; after stopping, judging whether the rotor is single-sided or double-sided unbalanced based on the opening of the circle - when there is a single opening, it is single-sided unbalanced, and when there are two openings, it is double-sided unbalanced; (2) the operation process of eliminating single-sided imbalance; (3) the operation process of eliminating double-sided imbalance. The present invention introduces the principle of parallel decomposition of force into the dynamic balancing and alignment of the fan when the fan is online, forming a set of online dynamic balancing and alignment methods; the method has the characteristics of high balancing efficiency, high balancing accuracy, standardized operation process and calculation formula, and is very easy to master for front-line equipment maintenance personnel.

[0004] The rotor balance of a wind turbine is one of the key factors in ensuring efficient and stable operation of the wind turbine. During the manufacturing and installation of the wind turbine, manufacturing tolerances of the blades, uneven material distribution, and errors during installation may lead to uneven mass distribution of the rotor blades, thus causing rotor imbalance. Traditional balancing correction methods are mainly achieved through static mass measurement and manual counterweighting, but this method is usually performed when the wind turbine is shut down and cannot respond to changes in mass distribution during operation in real time. As the scale of wind turbines continues to increase, the size and weight of rotor blades are also increasing, and traditional static balancing correction methods are facing increasing challenges.

[0005] Therefore, developing a method that can perform rotor balance correction in real time and accurately has become an important issue in improving the operating efficiency and extending the service life of wind turbines. Summary of the Invention

[0006] (1) Technical problems solved

[0007] To address the shortcomings of the existing technology, the present invention provides a method for balancing and correcting a wind turbine rotor. By collecting main shaft vibration, blade root strain, and pitch angle signals, the method constructs a real-time moment of inertia vector, calculates mass inertia change and aerodynamic cycle hysteresis index, and inputs a graph neural network to determine the imbalance coefficient and type, generating correction instructions for movable counterweights and pitch adjustment. The correction process is performed synchronously under rated conditions to ensure coordination between mechanical and aerodynamic compensation. After correction, the effect is evaluated through vibration vector comparison, and the results are stored in an adaptive database to optimize subsequent cycles. This method can maintain low-load operation, reduce maintenance costs, and extend the life of key components, thereby resolving the technical problems described in the background art.

[0008] (2) Technical solution

[0009] To achieve the above objectives, the present invention is implemented through the following technical solutions: A method for balancing a wind turbine rotor, comprising: under normal operation of the wind turbine, collecting three-dimensional vibration signals of the main shaft, blade root strain signals, and pitch angles in the same period to construct an instantaneous rotational inertia vector;

[0010] According to the rotational inertia vector, the mass inertia variable order index and the aerodynamic cycle hysteresis index are synchronously calculated and input into the graph neural network to output the coupling imbalance coefficient and determine the imbalance category;

[0011] The balance solution process receives the coupling imbalance coefficient and imbalance category, combines the aircraft structure parameters to deduce the movable counterweight displacement and target pitch compensation angle, and generates a correction instruction package;

[0012] Based on the imbalance type and intensity determined in the previous step, the slide rail adjustable counterweight device and the independent pitch servo are driven to collaboratively perform synchronous correction under rated operating conditions, ensuring that the timing of mechanical compensation and pneumatic compensation actions are consistent and do not interfere with each other.

[0013] The effect evaluation step compares the corrected vibration vector with the initial vibration vector and writes the result into the adaptive database as a parameter benchmark and basis for model updating.

[0014] Furthermore, under normal operation of the wind turbine, the three-dimensional vibration signal of the main shaft, the strain signal of the blade root and the pitch angles are collected synchronously in real time; the three-dimensional vibration signal of the main shaft is low-pass filtered, and the vibration intensity and vibration timing offset of the filtered signal at the rotor rotation frequency are extracted.

[0015] Furthermore, the average strain and peak strain of the blade root strain signal within the rotation period are calculated;

[0016] The horizontal vibration intensity of the main shaft, the horizontal vibration timing offset of the main shaft, the vertical vibration intensity of the main shaft, the vertical vibration timing offset of the main shaft, the axial vibration intensity of the main shaft, the axial vibration timing offset of the main shaft, the average strain of the first blade, the peak strain of the first blade, the pitch angle of the first blade, the average strain of the second blade, the peak strain of the second blade, the pitch angle of the second blade, the average strain of the third blade, the peak strain of the third blade and the pitch angle of the third blade are sequentially combined to form an instantaneous rotational inertia vector.

[0017] Furthermore, the mass inertia variable order index and the aerodynamic cycle hysteresis index are calculated based on the instantaneous rotational inertia vector; the mass inertia variable order index and the aerodynamic cycle hysteresis index are input into the graph neural network, and the graph neural network outputs the coupling imbalance coefficient and determines the imbalance category.

[0018] Furthermore, under normal operation of the wind turbine, the three-dimensional vibration signal of the main shaft, the strain signal of the blade root and each pitch angle are collected synchronously in real time, and the instantaneous rotational inertia vector is constructed;

[0019] Based on the instantaneous rotating inertia vector, the mass inertia variation index and the aerodynamic cycle hysteresis index are calculated. The mass inertia variation index is determined by the weighted sum of the three-axis vibration intensity of the main shaft and the rate of change of the average strain of the blade root, and the aerodynamic cycle hysteresis index is determined by the average phase difference between the blade pitch angle and the peak strain at the rotor rotation frequency.

[0020] Furthermore, the mass inertia variable order index and aerodynamic cycle hysteresis index are input into the graph neural network. The graph neural network takes the rotor system as the graph structure, which includes blade nodes and main shaft nodes. The interaction relationship of node features is learned through the graph convolution layer, and the coupling imbalance coefficient and imbalance category are output.

[0021] Furthermore, the coupling unbalance coefficient and the unbalance type are received and combined with the structural parameters of the wind turbine; the displacement of the adjustable counterweight on the sliding guide rail is calculated by the moment balance principle;

[0022] The target pitch compensation angle of the blade is calculated based on the relative difference in blade strain, and a correction instruction packet including the displacement of the adjustable counterweight and the target pitch compensation angle of the blade is generated.

[0023] Furthermore, after receiving the correction instruction package, the slidable adjustable weight device is driven to adjust the weight position and / or the independent pitch control servo is driven to adjust the blade pitch angle, and the adjustment action is determined by the type of imbalance; the mechanical compensation and aerodynamic compensation actions are executed synchronously without interfering with each other, and all operations are completed under the rated operating conditions of the wind turbine.

[0024] Furthermore, after completing the correction action, the three-axis vibration signal of the main shaft is collected in real time, the vibration amplitude corresponding to the rotor rotation frequency is extracted, the corrected vibration vector is constructed, and compared with the initial vibration vector before correction, and the vibration amplitude change and correction effect index are calculated.

[0025] Furthermore, the correction effect index, initial vibration vector, corrected vibration vector, parameters in the correction instruction package, and current wind speed and direction data are recorded in the adaptive database to support the regular update of the graph neural network model and parameter optimization of subsequent correction cycles.

[0026] (3) Beneficial effects

[0027] The present invention provides a method for balancing a wind turbine rotor, which has the following beneficial effects:

[0028] By collecting the three-dimensional vibration signals of the main shaft, blade root strain signals, and various pitch angles in real time during normal wind turbine operation, an instantaneous rotational inertia vector is constructed. Based on this vector, the mass inertia variable order index and aerodynamic cycle hysteresis index are simultaneously calculated, enabling dynamic monitoring of rotor imbalance. This system can quickly respond to dynamic changes in mass distribution and aerodynamic characteristics during operation, making it highly adaptable to wind fluctuations or component wear. Compared to traditional periodic inspections, this real-time performance significantly improves the timeliness of corrections and avoids the accumulation and deterioration of imbalance.

[0029] Graph Neural Networks (GNNs) are used to deeply process the mass inertia variable order index and aerodynamic cycle hysteresis index, outputting the coupled imbalance coefficient and imbalance category, thereby accurately identifying the intensity and type of imbalance. This not only improves diagnostic accuracy but also comprehensively considers the coupling effects of mass distribution and aerodynamic factors. Compared to traditional single-parameter analysis, the application of GNNs can explore the complex relationships between multi-source data, ensuring the comprehensiveness and reliability of diagnostic results.

[0030] By adjusting the displacement of the slide's adjustable counterweight and the blade's target pitch compensation angle, a correction instruction package is generated. Mechanical and aerodynamic compensation are then executed collaboratively under rated operating conditions, ensuring timing consistency and non-interference between the two adjustments, minimizing vibration and torque fluctuations caused by rotor imbalance. The synergy of mechanical and pneumatic methods not only improves correction efficiency but also avoids secondary problems that can arise from a single correction method.

[0031] By comparing vibration vectors before and after correction, calculating correction effectiveness indicators, and writing the evaluation results and related parameters to an adaptive database, the system supports dynamic updates of the graph neural network model and parameter optimization in subsequent correction cycles. It continuously learns and improves based on operating status and wind conditions, ensuring not only the efficiency of a single correction but also long-term correction effectiveness and system stability. Compared to traditional fixed-parameter correction methods, this adaptive optimization capability significantly enhances the system's intelligence.

[0032] Through real-time, precise balancing, this solution significantly reduces mechanical losses and vibrations caused by rotor imbalance, thereby reducing maintenance downtime and associated costs. Furthermore, optimized operating conditions improve energy capture efficiency and extend the service life of key components. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 The present invention is a flow chart of a method for balancing and correcting a wind turbine rotor. DETAILED DESCRIPTION

[0034] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0035] See also Figure 1 The present invention provides a method for balancing a wind turbine rotor, comprising:

[0036] Step 1. Under normal operating conditions of the wind turbine, the three-dimensional vibration signal of the main shaft, the blade root strain signal and each pitch angle are collected synchronously in real time; the three-dimensional vibration signal of the main shaft is low-pass filtered to extract the vibration intensity and vibration timing offset of the filtered signal at the rotor rotation frequency; the average strain and peak strain within the rotation period of the blade root strain signal are calculated; the horizontal vibration intensity of the main shaft, the horizontal vibration timing offset of the main shaft, the vertical vibration intensity of the main shaft, the vertical vibration timing offset of the main shaft, the axial vibration intensity of the main shaft, the axial vibration timing offset of the main shaft, the average strain of the first blade, the peak strain of the first blade, the pitch angle of the first blade, the average strain of the second blade, the peak strain of the second blade, the pitch angle of the second blade, the average strain of the third blade, the peak strain of the third blade and the pitch angle of the third blade are sequentially combined into an instantaneous rotational inertia vector, which is transmitted to the subsequent steps as output.

[0037] The step 1 includes the following:

[0038] Step 101: Signal acquisition

[0039] When the wind turbine is operating normally, the three-dimensional vibration signal of the main shaft, the strain signal of the blade root and the pitch angles are collected in real time. The specific process is as follows:

[0040] A triaxial accelerometer mounted on the main shaft measures the vibration acceleration of the main shaft in three orthogonal directions. These three directions are typically defined as horizontal, vertical, and axial, corresponding to the X, Y, and Z directions in the wind turbine coordinate system, respectively. The collected data are represented as the horizontal, vertical, and axial vibration accelerations of the main shaft. Strain gauges mounted at the root of each blade measure the strain generated by the force changes on the three blades during rotation, recording them as the first, second, and third blade strain signals. The pitch control device acquires the real-time pitch angle of each blade and records them as the first, second, and third blade pitch angles.

[0041] The system collects three-dimensional main shaft vibration signals, blade root strain signals, and pitch angles to reflect the operating status of the wind turbine rotor from different dimensions. The three-dimensional main shaft vibration signals directly reflect the mechanical vibration characteristics of the rotor caused by uneven mass distribution or unbalanced operation. The blade root strain signals reflect the stress changes caused by mass differences or uneven force on each blade, and the pitch angles reflect the impact of blade posture on the rotor's aerodynamic performance. The comprehensive collection of these signals provides multi-dimensional information, ensuring a comprehensive description of the rotor's dynamic characteristics and supporting the accuracy and completeness of subsequent processing.

[0042] Step 102 : pre-processing the collected main shaft three-dimensional vibration signals, blade root strain signals, and pitch angles to ensure data synchronization and reliability.

[0043] The specific process is as follows: First, the horizontal, vertical, and axial vibration accelerations of the main shaft, the strain signals of the first, second, and third blades, and the pitch angles of the first, second, and third blades are aligned according to the timestamps at the time of acquisition, so that these signals correspond to the same rotational position of the rotor in time. Second, low-pass filtering is applied to the horizontal, vertical, and axial vibration accelerations of the main shaft. By setting an appropriate frequency threshold, the high-frequency noise in the signals is removed, while the low-frequency vibration information related to the rotor rotation is retained. The filtered horizontal, vertical, and axial vibration accelerations of the main shaft are obtained.

[0044] The reason for time alignment is to ensure the consistency of all signals in the time dimension, to avoid the situation where different signals cannot accurately correspond to the same state of the rotor due to sampling time deviation, thereby improving the comparability of data and analysis accuracy; the spindle vibration signal may contain high-frequency interference that is not related to rotor imbalance. Through filtering, the low-frequency vibration characteristics related to the rotor rotation frequency can be highlighted, so that the data can more realistically reflect the rotor operating status.

[0045] Step 103: extract key features from the preprocessed signal to quantify the dynamic characteristics of the rotor.

[0046] The specific process is as follows: frequency domain analysis is performed on the filtered spindle horizontal vibration acceleration, the filtered spindle vertical vibration acceleration and the filtered spindle axial vibration acceleration respectively. These time domain signals are converted into frequency domain signals through the fast Fourier transform method, and the vibration intensity and vibration timing offset information corresponding to the rotor rotation frequency are extracted, where the vibration intensity is the amplitude of the frequency domain signal at the rotor rotation frequency, and the vibration timing offset is the phase angle of the frequency domain signal at the rotor rotation frequency, which are recorded as the spindle horizontal vibration intensity, the spindle horizontal vibration timing offset, the spindle vertical vibration intensity, the spindle vertical vibration timing offset, the spindle axial vibration intensity and the spindle axial vibration timing offset respectively.

[0047] For the first, second, and third blade strain signals, time domain analysis is performed over a complete rotor rotation cycle. The average and maximum values ​​of each signal are calculated and recorded as the first blade average strain, first blade peak strain, second blade average strain, second blade peak strain, third blade average strain, and third blade peak strain, respectively. For the first, second, and third blade pitch angles, their real-time values ​​are directly used as features.

[0048] By extracting the vibration intensity and timing offset of the main shaft vibration signal, we can quantify the magnitude of the vibration caused by rotor imbalance and its distribution during the rotation cycle, providing direct evidence of the imbalance. The average value of the blade strain signal reflects the overall force level of the blade during the rotation cycle, while the peak strain reflects the extreme force conditions during operation. The combination of the two can comprehensively describe the impact of blade mass distribution on rotor balance.

[0049] Step 104: Construct instant rotational inertia vector

[0050] The extracted features are combined in sequence to construct a multidimensional vector, named the instantaneous rotational inertia vector;

[0051] The specific process is as follows: arrange the horizontal vibration intensity of the main shaft, the horizontal vibration timing offset of the main shaft, the vertical vibration intensity of the main shaft, the vertical vibration timing offset of the main shaft, the axial vibration intensity of the main shaft, the axial vibration timing offset of the main shaft, the average strain of the first blade, the peak strain of the first blade, the average strain of the second blade, the peak strain of the second blade, the average strain of the third blade, the peak strain of the third blade, the pitch angle of the first blade, the pitch angle of the second blade and the pitch angle of the third blade in sequence to form a vector containing fifteen components, which is expressed as the instantaneous rotational inertia vector.

[0052] By constructing an instantaneous rotational inertia vector and integrating the features extracted from multi-source signals into a unified structured representation, it is convenient to centrally reflect the rotor's vibration characteristics, blade stress state, and aerodynamic attitude information, facilitate the subsequent steps of systematic analysis of the rotor imbalance state, and improve the efficiency of data processing and the accuracy of analysis results.

[0053] The constructed instantaneous rotational inertia vector is used as the output of step 1 and directly passed to the subsequent step 2. Step 2 will use the characteristic information in the instantaneous rotational inertia vector to further calculate the rotor's mass distribution change index and aerodynamic performance lag index to complete the analysis and correction of the imbalance state.

[0054] The reason for using the instantaneous rotational inertia vector as output is that it contains multi-dimensional characteristics of the rotor's operating state, providing comprehensive and structured input data for step two, ensuring continuity and consistency in the analysis process. The benefit of this connection method is that it maintains the complete transfer of data from acquisition to analysis, supporting real-time response and efficient execution of the entire balancing correction process during dynamic operation.

[0055] Step 1 collects the three-dimensional vibration signals of the main shaft, blade root strain signals, and pitch angles in real time, and then performs time alignment, low-pass filtering, feature extraction, and vector construction to ultimately generate the instantaneous rotational inertia vector. This process integrates the rotor's vibration characteristics, blade stress state, and aerodynamic attitude information to fully reflect the rotor's mass distribution and dynamic state during operation, ensuring the accuracy and real-time performance of subsequent imbalance analysis.

[0056] Step 2: Calculate the mass inertia variable order index and aerodynamic cycle hysteresis index based on the instantaneous rotational inertia vector; input the mass inertia variable order index and aerodynamic cycle hysteresis index into the graph neural network; the graph neural network outputs the coupling imbalance coefficient and determines the imbalance category.

[0057] Among them: under the normal operating state of the wind turbine, the three-dimensional vibration signal of the main shaft, the blade root strain signal and each pitch angle are collected synchronously in real time, and the instantaneous rotational inertia vector is constructed; based on the instantaneous rotational inertia vector, the mass inertia change order index and the aerodynamic circulation hysteresis index are calculated, wherein the mass inertia change order index is determined by the weighted sum of the change rate of the three-dimensional vibration intensity of the main shaft and the average strain of the blade root, and the aerodynamic circulation hysteresis index is determined by the average phase difference between the blade pitch angle and the peak strain at the rotor rotation frequency; the mass inertia change order index and the aerodynamic circulation hysteresis index are input into the graph neural network, which uses the rotor system as a graph structure, including blade nodes and main shaft nodes, and learns the interaction relationship of the node features through the graph convolution layer, and outputs the coupling imbalance coefficient and imbalance category; the coupling imbalance coefficient and imbalance category are passed to the subsequent steps to generate a correction instruction package.

[0058] The second step includes the following:

[0059] Step 201: Calculate the mass inertia variable order index

[0060] The mass inertia variable order index is used to quantify the degree of change in rotor mass distribution over time and reflect the characteristics of mass imbalance during dynamic operation. The specific calculation process is as follows:

[0061] First, the horizontal, vertical, and axial vibration intensities of the main shaft, as well as the average strains of the first, second, and third blades, are extracted from the instantaneous rotating inertia vector. Then, within a time window equal to the rotor's rotation period, the components of these vibration intensities and blade average strains are analyzed, and their rates of change within that time window are calculated. The rate of change is calculated by performing a linear regression on the data for each component within the time window, obtaining the slope of the regression curve, which is the rate of change. The absolute values ​​of these rates of change are then multiplied by a preset weighting coefficient and summed to obtain the mass inertia order index. The weighting coefficient is set based on the specific model of the wind turbine and historical operating data to ensure that the contribution of different components to mass imbalance is reasonably assessed.

[0062] By analyzing the rate of change of vibration intensity and blade strain, we can capture the impact of dynamic changes in rotor mass distribution on system operation. Using linear regression to calculate the rate of change effectively smooths short-term fluctuations, highlights long-term trends, and improves the stability and reliability of the analysis. The introduction of weighting coefficients allows for adjustments based on the characteristics of different wind turbines, ensuring the universality and specificity of the mass inertia variable order index.

[0063] Step 202: Calculate the aerodynamic cycle hysteresis index

[0064] The aerodynamic cycle hysteresis index is used to quantify the timing deviation between the blade aerodynamic response and pitch adjustment, reflecting the degree of aerodynamic imbalance. The specific calculation process is as follows:

[0065] First, the first blade pitch angle, second blade pitch angle, third blade pitch angle, and first blade peak strain, second blade peak strain, and third blade peak strain are extracted from the instantaneous rotating inertia vector. Then, within a complete rotor rotation cycle, the pitch angle and peak strain of each blade are analyzed in the frequency domain, and their phase angles at the rotor rotation frequency are extracted through fast Fourier transform. Next, the difference between the phase of each blade pitch angle and the phase of the peak strain is calculated to obtain the phase difference. Finally, the phase difference of the three blades is averaged to obtain the aerodynamic cycle hysteresis index.

[0066] The timing relationship between the blade's aerodynamic response and pitch adjustment directly affects the rotor's aerodynamic balance. Calculating the phase difference quantifies this timing deviation and reveals the degree of aerodynamic imbalance. Using a fast Fourier transform to extract the phase angle accurately captures the signal's periodic characteristics at the rotational frequency, ensuring accurate analysis. The use of an average phase difference integrates information from multiple blades, providing a holistic assessment of aerodynamic lag and contributing to a comprehensive understanding of the rotor's aerodynamic state.

[0067] Step 202: Input graph neural network

[0068] The mass inertia variable order index and aerodynamic cycle hysteresis index are used as features and input into the pre-trained graph neural network to output the coupling imbalance coefficient and imbalance category.

[0069] The specific process is as follows: First, a graph structure is constructed, which contains four nodes, representing three blades and a main shaft. The characteristics of each node are composed of the mass inertia variable order index and the aerodynamic cycle hysteresis index. The edges in the graph represent the physical connection relationship between the blades and the main shaft. Then, these node features are processed through multiple layers of graph convolution layers to learn the interaction and dependencies between nodes. Finally, the graph neural network outputs the coupling imbalance coefficient and imbalance category through the regression head and classification head respectively. The coupling imbalance coefficient is a scalar value that represents the comprehensive intensity of the imbalance; the imbalance category is a classification result, including mass imbalance, aerodynamic imbalance, and coupling imbalance.

[0070] There are complex physical and functional connections between the blades and the main shaft in the rotor system. The graph structure can naturally represent the relationship between these multiple nodes. The graph neural network can effectively capture the local and global dependencies of node features through the graph convolution layer, improving the ability to analyze imbalance states. The design of the output coupling imbalance coefficient and imbalance category can simultaneously provide quantitative and qualitative assessments of the imbalance, supporting the precise formulation of subsequent corrective measures.

[0071] The output of step 2 is the coupling imbalance coefficient and imbalance category, which are directly passed to the subsequent step 3. Based on these output results and the specific structural parameters of the wind turbine, step 3 calculates the displacement of the movable counterweight and the target pitch compensation angle, thereby generating a correction instruction package.

[0072] The coupled imbalance coefficient and imbalance category provide information on the intensity and type of imbalance, respectively, providing clear direction and basis for formulating corrective measures. Directly transferring these results to step three ensures timely application of analysis results, supporting the continuity and real-time nature of the entire balance correction process. This integrated approach effectively reduces data processing and decision-making delays, improving the stability and efficiency of wind turbine operation.

[0073] In the second step, an in-depth analysis of the instantaneous rotational inertia vector is performed to calculate the mass inertia variable order index and the aerodynamic cycle hysteresis index. The graph neural network is then used to output the coupled imbalance coefficient and imbalance category, thereby achieving real-time quantification and classification of the rotor mass imbalance and aerodynamic imbalance. This ensures the accuracy and comprehensiveness of the imbalance diagnosis and provides reliable decision-making support for subsequent corrective measures.

[0074] Step 3: Receive the coupling imbalance coefficient and imbalance type determined in the previous step, combine the structural parameters of the wind turbine, calculate the displacement of the adjustable counterweight on the sliding guide rail based on the torque balance principle, and calculate the target pitch compensation angle of the blade based on the relative strain difference of the blade, and generate a correction instruction package containing the displacement of the adjustable counterweight and the target pitch compensation angle of the blade to guide subsequent correction actions;

[0075] The step three includes the following:

[0076] Step 301: Receive input data

[0077] Step 3 first receives the coupling imbalance coefficient and imbalance category output from step 2 and simultaneously obtains the wind turbine's structural parameters. These parameters include the rotor radius, blade mass, main shaft moment of inertia, movable counterweight rail length, and counterweight mass. The coupling imbalance coefficient reflects the degree of rotor imbalance, and the imbalance categories are categorized as mass imbalance, aerodynamic imbalance, and coupling imbalance.

[0078] The purpose of obtaining the structural parameters of the model is to combine the correction calculation with the specific physical characteristics of the wind turbine, which can ensure that the correction measures match the actual equipment.

[0079] Step 302: Determine a correction strategy based on the imbalance category

[0080] According to the different types of imbalance, step three determines the specific correction strategy. For mass imbalance, displacement adjustment of the movable counterweight device is adopted; for aerodynamic imbalance, blade pitch angle adjustment is adopted; for coupling imbalance, displacement adjustment of the movable counterweight device and blade pitch angle adjustment are adopted simultaneously.

[0081] The purpose of determining the correction strategy based on the imbalance type is to adopt appropriate treatment methods for different types of imbalance problems. Mass imbalance is mainly caused by uneven distribution of rotor mass. Adjusting the counterweight position can effectively offset the unbalanced torque. Aerodynamic imbalance is caused by uneven distribution of aerodynamic loads on the blades. Adjusting the pitch angle can optimize aerodynamic characteristics. Coupled imbalance has characteristics of both and requires a combination of both methods for correction. This classification method ensures that the correction measures correspond to the imbalance type, improves correction efficiency, and avoids ineffective adjustments caused by inappropriate measures.

[0082] Step 303: Calculate the displacement of the movable counterweight device

[0083] For imbalances classified as mass or coupling, step three calculates the displacement of the movable counterweight to offset the torque generated by the rotor imbalance. The calculation logic is as follows: First, multiply the coupling imbalance coefficient by the spindle moment of inertia to calculate the torque generated by the rotor imbalance. This torque is then divided by the product of the counterweight mass and the rotor radius to determine the displacement of the movable counterweight. If the calculated displacement exceeds the length of the movable counterweight rail, the maximum rail length is used as the final displacement.

[0084] The purpose of calculating the displacement of the movable counterweight is to restore the rotor to equilibrium by applying the principle of moment balance, using the movement of the counterweight to generate a torque equal in magnitude and opposite in direction to the rotor's unbalanced torque. Multiplying the coupling unbalance coefficient by the spindle's moment of inertia reflects the total amount of unbalanced torque, while dividing by the product of the counterweight's mass and the rotor radius converts the torque into the distance the counterweight travels on the rail.

[0085] Step 304: Calculate the target pitch compensation angle of the blade

[0086] For the case where the imbalance type is aerodynamic imbalance or coupled imbalance, step three calculates the target pitch compensation angle of each blade to adjust the aerodynamic load distribution.

[0087] The calculation logic is as follows: first determine the relative deviation of the blade peak strain, that is, calculate the difference between the ratio of the peak strain of each blade to the total peak strain of all blades and the ideal uniform distribution ratio; then multiply the pitch compensation coefficient, the coupling imbalance coefficient and the relative deviation of the blade peak strain to obtain the target pitch compensation angle of each blade.

[0088] The purpose of calculating the target blade pitch compensation angle is to modify aerodynamic characteristics by adjusting the blade pitch angle, optimize aerodynamic load distribution, and reduce vibration and torque fluctuations caused by aerodynamic imbalance. Using the relative deviation of blade peak strain as a basis accurately reflects the differences in aerodynamic loads on each blade, ensuring that the compensation angle corresponds to the actual imbalance condition. Including the pitch compensation coefficient and the coupled imbalance coefficient in the calculation allows adjustments based on the aircraft characteristics and operating conditions, improving the flexibility and accuracy of the correction.

[0089] Based on the imbalance classification, step three integrates the calculated displacement of the movable counterweight and / or the target blade pitch compensation angle to generate a correction instruction packet. For mass imbalance, only the displacement is included; for aerodynamic imbalance, only the compensation angle is included; for coupled imbalance, both the displacement and compensation angle are included.

[0090] The purpose of generating the correction instruction package is to integrate the scattered correction parameters into a unified instruction set, so that the wind turbine can efficiently implement balance correction in the operating state.

[0091] Step 3 directly passes the generated correction instruction packet to step 4, where it is used by the movable counterweight system's slide rail adjustment and the independent pitch servo system to perform synchronous correction. Passing the correction instruction packet as an output to step 4 ensures that the correction parameters are promptly applied to the actual equipment, supporting synchronous correction of the wind turbine under rated operating conditions. This direct connection reduces time delays during the correction process, improves correction efficiency, and enhances the stability of wind turbine operation.

[0092] Step 3 receives the coupling imbalance coefficient and imbalance category from step 2, as well as the wind turbine's structural parameters, and calculates the displacement of the movable counterweight and the target blade pitch compensation angle for mass imbalance, aerodynamic imbalance, and coupling imbalance, respectively. It then generates a correction instruction packet and transmits it to step 4. This calculation process, based on the optimization principles of moment balance and aerodynamic load distribution, ensures the accuracy and specificity of the correction measures, providing an actionable instruction basis for wind turbine balance correction.

[0093] Step 4: Based on the correction instruction package generated in the previous step, the slidable adjustable weight device is driven to adjust the weight position and / or the independent pitch control servo is driven to adjust the blade pitch angle. The specific action is determined by the type of imbalance, ensuring that the mechanical compensation and aerodynamic compensation actions are performed synchronously and do not interfere with each other. All operations are completed under the rated operating conditions of the wind turbine.

[0094] The step 4 includes the following contents:

[0095] Step 401: Receive a calibration instruction packet

[0096] Step 4 first obtains the correction instruction packet from step 3. This instruction packet contains the following specific parameters: the displacement of the movable counterweight device, the target pitch compensation angle of the blade, and the imbalance type. The displacement of the movable counterweight device refers to the distance the counterweight block needs to move along the slide rail; the target pitch compensation angle of the blade is the pitch angle value that needs to be adjusted for each blade; and the imbalance type is the specific type of current rotor imbalance, including mass imbalance, aerodynamic imbalance, or coupling imbalance. These parameters are generated in step 3 based on the rotor state analysis and are used to guide the execution of subsequent correction actions.

[0097] The correction instruction package integrates the diagnostic results and correction calculations into a unified data set, which can ensure that the correction process is carried out based on accurate data, reduce execution deviations caused by information transmission errors, and thus improve the accuracy and efficiency of correction.

[0098] Step 402: Determine the correction action type

[0099] Step 4: Based on the imbalance category specified in the correction instruction package, determine and execute the corrective action type. Specifically, when the imbalance category is mass imbalance, only the slide rail adjustable counterweight device is driven to adjust the position of the counterweight block; when the imbalance category is aerodynamic imbalance, only the independent pitch servo is driven to adjust the pitch angle of the blade; when the imbalance category is coupled imbalance, both the slide rail adjustable counterweight device and the independent pitch servo are driven simultaneously to adjust the position of the counterweight block and the pitch angle of the blade, respectively. The imbalance category is determined directly based on the parameters in the correction instruction package, without the need for additional calculations.

[0100] Corrective measures are tailored to the characteristics of different imbalance issues, ensuring that the action matches the cause. Mass imbalance is caused by uneven rotor mass distribution; adjusting the counterweight position can effectively offset the unbalanced torque. Aerodynamic imbalance is caused by uneven distribution of aerodynamic loads on the blades; adjusting the pitch angle can optimize aerodynamic characteristics. Coupled imbalance combines the characteristics of both and requires a combination of both approaches. By implementing these adjustments in a categorized manner, corrections can be more targeted and mitigate poor results due to inappropriate measures.

[0101] Step 403: Drive the slide rail with adjustable counterweight device

[0102] For imbalances requiring counterweight adjustment, the slide-adjustable counterweight assembly moves the counterweight to the target position based on the displacement specified in the correction instruction package. The counterweight's movement speed is predetermined by the slide-adjustable counterweight assembly's mechanical design parameters, such as motor power and rail friction coefficient. The movement time is calculated by dividing the required counterweight movement distance by the movement speed. This is used to control the timing of subsequent actions, ensuring that the counterweight reaches the specified position within the predetermined time.

[0103] By changing the position of the counterweight, a torque equal in magnitude and opposite in direction to the rotor's unbalanced torque is generated, restoring the rotor's equilibrium. The movement speed is determined by the mechanical design parameters, ensuring a smooth movement and avoiding any additional interference with the rotor's dynamic characteristics.

[0104] For imbalance categories that require adjustment of the blade pitch angle, the independent pitch servo adjusts the pitch of each blade to the corresponding value based on the target pitch compensation angle specified in the correction instruction packet.

[0105] The pitch adjustment speed is predetermined by the system parameters of the independent pitch servo, such as the servo motor's response speed and the performance of the transmission mechanism. The adjustment time is calculated by dividing the absolute value of the target pitch compensation angle by the pitch adjustment speed. This is used to control the timing of subsequent actions, ensuring that each blade is adjusted within the predetermined time. By adjusting the blade pitch angle, the aerodynamic characteristics of the blade are changed, optimizing the aerodynamic load distribution and reducing vibration and torque fluctuations caused by aerodynamic imbalance. The pitch adjustment speed is determined by system parameters, ensuring a smooth and controllable adjustment process.

[0106] Step 404: Collaboratively perform mechanical and aerodynamic compensation

[0107] When the imbalance type is coupled imbalance, step 4 requires performing the counterweight position adjustment and blade pitch angle adjustment simultaneously, and ensuring that the timing of the two actions is consistent.

[0108] The collaborative execution process is as follows: first calculate the movement time of the counterweight position adjustment and the longest pitch adjustment time among all blades; then compare the movement time and the longest pitch adjustment time, and take the larger value of the two as the total duration of the synchronous correction; finally adjust the execution speed so that the counterweight position adjustment and the blade pitch angle adjustment are completed within the total duration of the synchronous correction. The specific adjustment method is: if the counterweight movement time is less than the longest pitch adjustment time, reduce the movement speed of the counterweight; if the longest pitch adjustment time is less than the counterweight movement time, reduce the blade pitch adjustment speed. By adjusting the execution speed so that the two actions are completed in the same time, the dynamic disturbance in the correction process can be reduced, and the stability and reliability of the correction effect can be improved. This collaborative method effectively copes with the complexity of coupled imbalance and enhances the overall efficiency of the correction.

[0109] During the correction action, step four uses sensors to monitor the position of the counterweight and the pitch angle of the blade in real time. The position of the counterweight is detected and fed back by the slide rail position sensor, and the pitch angle of the blade is detected and fed back by the encoder of the independent pitch servo. If a deviation is detected between the counterweight position or the blade pitch angle and the target value in the correction instruction package, the execution speed is adjusted according to the size of the deviation, or the correction instruction is reissued to ensure that the target value is ultimately achieved. Ensure the accurate execution of the correction action to avoid deviations from the expected correction effect due to execution errors. Through real-time feedback and dynamic adjustment, deviations can be corrected in a timely manner, improving the accuracy and reliability of the correction.

[0110] After the correction is complete, step 4 confirms that the position of the counterweight and the pitch angle of the blades have reached the target values ​​specified in the correction instruction package. After confirmation, step 5 is performed to evaluate the correction effect to verify whether the correction is effective.

[0111] In step 4, the correction command packet is received, and the slide-type adjustable counterweight device and independent pitch servo are driven to adjust the counterweight position and blade pitch angle, respectively, based on the imbalance type. A coordinated strategy is used to adjust the execution speed, ensuring that mechanical and aerodynamic compensation are completed synchronously under coupled imbalance conditions. Real-time monitoring and feedback mechanisms ensure the accuracy of the correction. After the correction is completed, the results are evaluated in step 5. This technical logic can effectively reduce vibration caused by rotor imbalance and improve the operating efficiency and service life of the wind turbine.

[0112] Step 5. After completing the correction action, collect the three-axis vibration signals of the main shaft in real time, extract the vibration amplitude corresponding to the rotor rotation frequency, construct the corrected vibration vector, and compare it with the initial vibration vector before correction. Calculate the vibration amplitude change and correction effect index. At the same time, record the correction effect index, initial vibration vector, corrected vibration vector, parameters in the correction instruction package, and current wind speed and wind direction data into the adaptive database to support the regular update of the graph neural network model and parameter optimization of subsequent correction cycles.

[0113] The step five includes the following:

[0114] Step 501: Collect the corrected vibration vector

[0115] After completing the correction action in step 4, the vibration acceleration signals of the spindle in the horizontal, vertical and axial directions are immediately collected through the three-axis acceleration sensor installed on the spindle; these signals are respectively called the corrected spindle horizontal vibration acceleration, the corrected spindle vertical vibration acceleration and the corrected spindle axial vibration acceleration.

[0116] These vibration acceleration signals are then analyzed in the frequency domain. This involves converting the time-domain signals into frequency-domain signals. The vibration amplitudes corresponding to the rotor's rotational frequency are then extracted and labeled as the corrected horizontal, vertical, and axial spindle vibration amplitudes. These three vibration amplitudes are combined into a three-dimensional vector, called the corrected vibration vector.

[0117] When used, the vibration amplitude directly reflects the mechanical vibration intensity caused by rotor imbalance. By extracting the amplitude corresponding to the rotor rotation frequency through frequency domain analysis, the vibration characteristics directly related to rotor imbalance can be accurately captured.

[0118] Step 502: Obtain initial vibration vector

[0119] Extract the pre-correction vibration data from the analysis results of step 1. Specifically, this includes the pre-correction spindle horizontal vibration amplitude, pre-correction spindle vertical vibration amplitude, and pre-correction spindle axial vibration amplitude. These vibration amplitudes are derived from the spindle vibration acceleration signals collected before correction and then subjected to frequency domain analysis. These amplitudes correspond to the rotor rotational frequency and are combined into a three-dimensional vector, called the initial vibration vector.

[0120] The purpose of obtaining the initial vibration vector is to provide benchmark data for evaluating the correction effect. The initial vibration vector represents the unbalanced vibration state of the rotor before correction. Comparison with the vibration vector after correction can clearly reflect the improvement effect of the correction action.

[0121] Step 503: Comparing and analyzing the corrected vibration vector and the initial vibration vector, specifically including calculating a vibration amplitude change and a correction effect index; the vibration amplitude change is calculated by subtracting the horizontal vibration amplitude of the initial vibration vector from the horizontal vibration amplitude of the corrected vibration vector to obtain the horizontal vibration amplitude change;

[0122] The vertical vibration amplitude of the corrected vibration vector is subtracted from the vertical vibration amplitude of the initial vibration vector to obtain the change in the vertical vibration amplitude. The axial vibration amplitude of the corrected vibration vector is subtracted from the axial vibration amplitude of the initial vibration vector to obtain the change in the axial vibration amplitude. The correction effect index is calculated by first calculating the modulus of the initial vibration vector and the corrected vibration vector, respectively. The modulus is calculated by adding the squares of the vibration amplitudes in the three directions and taking the square root. The modulus is then subtracted from the modulus of the corrected vibration vector to obtain the modulus difference. Finally, the modulus difference is divided by the modulus of the initial vibration vector and multiplied by 100% to convert the result into a percentage, indicating the percentage reduction in vibration intensity.

[0123] The effectiveness of corrective actions is quantitatively evaluated. The change in vibration amplitude reflects the specific improvement in spindle vibration in each direction after correction, facilitating analysis of the effects of corrective actions in different directions. The Correction Effectiveness Index, calculated as the percentage change in the vibration vector modulus, comprehensively reflects the degree of reduction in overall vibration intensity.

[0124] Relevant data for the correction effect evaluation is recorded in an adaptive database. This includes the correction effect index, the initial vibration vector, the corrected vibration vector, the parameters in the correction instruction package, and the current wind speed and direction data. The parameters in the correction instruction package include the displacement of the movable counterweight device and the target pitch compensation angle of the blades. The adaptive database structure contains multiple fields, such as correction time, imbalance category, correction parameters, initial vibration vector, corrected vibration vector, correction effect index, wind speed, and wind direction, for storing and managing this data. By recording the vibration vectors, correction parameters, and wind condition data before and after correction, the correlation between the correction effect and operating conditions can be analyzed, thereby improving the correction strategy.

[0125] Step 504: Update model parameters

[0126] Based on historical data accumulated in the adaptive database, the graph neural network model used in step 2 is regularly retrained to update its parameters. Retraining is performed monthly, and the training samples include the initial vibration vectors, corrected vibration vectors, correction parameters, and wind speed and direction data from the database. The training process optimizes the graph neural network, enabling it to more accurately predict the coupling imbalance coefficient and imbalance category.

[0127] By regularly retraining with historical data, the model can learn the latest vibration characteristics and correction effect feedback, thereby improving its prediction accuracy in subsequent correction cycles; this adaptive update mechanism ensures that the correction method can continuously improve as operating conditions change, maintaining high diagnostic and correction efficiency.

[0128] After completing the correction evaluation and recording the data, the system enters the next correction cycle, continuing steps 1 through 5 to ensure continuous rotor balancing. Since the rotor state and wind conditions change during wind turbine operation, the cyclic execution of the correction process allows for timely detection and resolution of new imbalances, maintaining the rotor's dynamic balance and improving the wind turbine's operating efficiency and service life.

[0129] Step five collects the vibration acceleration signals of the corrected main shaft in the horizontal, vertical and axial directions, extracts the vibration amplitude corresponding to the rotor rotation frequency, constructs the corrected vibration vector, and compares it with the initial vibration vector obtained from step one. Then, the effect of the correction action is quantified by calculating the change in vibration amplitude and the correction effect index. The evaluation results, together with the correction parameters and wind condition data, are recorded in the adaptive database to support the regular update of the graph neural network model and provide data support for the execution of the next correction cycle. This technical logic realizes the accurate evaluation of the correction effect and the continuous improvement of the correction method, ensuring the long-term effectiveness of the wind turbine rotor balance correction.

[0130] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0131] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0132] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is only for some logical functions. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0133] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0134] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for balancing a wind turbine rotor, characterized in that: include, When the wind turbine is operating normally, the three-dimensional vibration signal of the main shaft, the strain signal of the blade root and each pitch angle are collected in the same period to construct the instantaneous rotational inertia vector; According to the rotational inertia vector, the mass inertia variable order index and the aerodynamic cycle hysteresis index are synchronously calculated and input into the graph neural network to output the coupling imbalance coefficient and determine the imbalance category; The balance solution process receives the coupling imbalance coefficient and imbalance category, combines the aircraft structure parameters to deduce the movable counterweight displacement and target pitch compensation angle, and generates a correction instruction package; According to the type and intensity of imbalance, the slide rail adjustable counterweight device and the independent pitch servo are driven to complete synchronous correction in a coordinated manner under rated working conditions, so that the timing of mechanical compensation and pneumatic compensation actions are consistent and do not interfere with each other; The corrected vibration vector is compared with the initial vibration vector, and the result is written into the adaptive database as a parameter benchmark and basis for model update.

2. The method for balancing a wind turbine rotor according to claim 1, wherein: Under normal operating conditions of the wind turbine, the three-dimensional vibration signal of the main shaft, the strain signal of the blade root and the pitch angles are collected synchronously in real time. The three-dimensional vibration signal of the main shaft is low-pass filtered to extract the vibration intensity and vibration timing offset of the filtered signal at the rotor rotation frequency.

3. The method for balancing a wind turbine rotor according to claim 2, wherein: Calculate the average strain and peak strain of the blade root strain signal during the rotation cycle; The horizontal vibration intensity of the main shaft, the horizontal vibration timing offset of the main shaft, the vertical vibration intensity of the main shaft, the vertical vibration timing offset of the main shaft, the axial vibration intensity of the main shaft, the axial vibration timing offset of the main shaft, the average strain of the first blade, the peak strain of the first blade, the pitch angle of the first blade, the average strain of the second blade, the peak strain of the second blade, the pitch angle of the second blade, the average strain of the third blade, the peak strain of the third blade and the pitch angle of the third blade are sequentially combined to form an instantaneous rotational inertia vector.

4. The method for balancing a wind turbine rotor according to claim 3, wherein: The mass inertia variable order index and aerodynamic cycle hysteresis index are calculated based on the instantaneous rotational inertia vector; the mass inertia variable order index and aerodynamic cycle hysteresis index are input into the graph neural network, and the graph neural network outputs the coupling imbalance coefficient and determines the imbalance category.

5. The method for balancing a wind turbine rotor according to claim 4, wherein: Under normal operation of the wind turbine, the three-dimensional vibration signal of the main shaft, the strain signal of the blade root and each pitch angle are collected synchronously in real time, and the instantaneous rotational inertia vector is constructed; Based on the instantaneous rotating inertia vector, the mass inertia variation index and the aerodynamic cycle hysteresis index are calculated. The mass inertia variation index is determined by the weighted sum of the three-axis vibration intensity of the main shaft and the rate of change of the average strain of the blade root, and the aerodynamic cycle hysteresis index is determined by the average phase difference between the blade pitch angle and the peak strain at the rotor rotation frequency.

6. The method for balancing a wind turbine rotor according to claim 5, wherein: The mass inertia variable order index and aerodynamic cycle hysteresis index are input into the graph neural network. The graph neural network uses the rotor system as a graph structure, which includes blade nodes and main shaft nodes. The interactive relationship of node features is learned through the graph convolution layer, and the coupling imbalance coefficient and imbalance category are output.

7. The method for balancing a wind turbine rotor according to claim 6, wherein: The coupling unbalance coefficient and unbalance type are received and combined with the structural parameters of the wind turbine; the displacement of the adjustable counterweight on the sliding guide rail is calculated by the moment balance principle; The target pitch compensation angle of the blade is calculated based on the relative difference in blade strain, and a correction instruction packet including the displacement of the adjustable counterweight and the target pitch compensation angle of the blade is generated.

8. The method for balancing a wind turbine rotor according to claim 7, wherein: After receiving the correction instruction packet, driving the slidable adjustable weight device to adjust the weight position and / or driving the independent pitch control servo to adjust the blade pitch angle, the adjustment action being determined by the type of imbalance; The mechanical compensation and the aerodynamic compensation are performed synchronously without interfering with each other, and all operations are completed under the rated operating conditions of the wind turbine.

9. The method for balancing a wind turbine rotor according to claim 8, wherein: After completing the correction action, the three-axis vibration signal of the main shaft is collected in real time, the vibration amplitude corresponding to the rotor rotation frequency is extracted, the corrected vibration vector is constructed, and compared with the initial vibration vector before correction, and the vibration amplitude change and correction effect index are calculated.

10. The method for balancing a wind turbine rotor according to claim 9, wherein: The correction effect indicators, initial vibration vector, corrected vibration vector, parameters in the correction instruction package, and current wind speed and direction data are recorded in the adaptive database to support the regular update of the graph neural network model and parameter optimization in subsequent correction cycles.

Images