Impeller unbalance monitoring method based on fusion of wind speed and blade root load

Abstract

The invention provides an impeller unbalance monitoring method based on fusion of wind speed and blade root load, which comprises the following steps of: firstly, constructing a dynamic synthesis model of local effective wind speed of a blade, and establishing a dynamic effective wind speed model of each point of the blade by combining wind shear, a pitch angle, a rotation effect and an incoming flow wind direction so as to provide high-precision input for theoretical aerodynamic load calculation; then, a multi-sensor dynamic data fusion method is fully adopted, real-time wind speed field data and direct measurement data of a blade root load sensor are dynamically fused, a real-time correction algorithm based on a physical model is constructed, interference of wind speed fluctuation on load judgment is eliminated, and unbalance detection precision is improved. And finally, based on a dynamic threshold early warning mechanism adopted by theoretical-actual load deviation, a sliding window standard deviation analysis and machine learning algorithm is introduced to set a dynamic abnormal threshold of the load deviation. Therefore, the unbalance state of the impeller is recognized in real time, and real-time monitoring, analysis and early warning of pneumatic unbalance of the fan are achieved.

Description

Technical Field

[0001] The present invention relates to the field of operation monitoring of wind turbines, in particular to a monitoring method for aerodynamic imbalance of wind turbine blades. Background Art

[0002] During the operation of existing wind turbines, due to the influence of uneven wind speed, blade surface contamination, icing or other environmental factors, the aerodynamic loads on the blades may become unbalanced, resulting in increased impeller vibration, accelerated structural fatigue and increased failure risk. Traditional monitoring methods mainly rely on blade surface monitoring and simple vibration analysis. These methods have great limitations in real-time evaluation of the influence of wind speed changes on blade loads, and it is also difficult to accurately distinguish the load changes caused by environmental wind speed fluctuations from the real aerodynamic imbalance problems.

[0003] In recent years, due to its long-range and non-contact real-time wind measurement ability, lidar (LiDAR) technology has been widely used in the wind power field. LiDAR can accurately collect the incoming wind speed and direction information, providing important data for evaluating the characteristics of the wind farm, but it cannot directly reflect the actual force condition of the wind turbine blades. The root load sensor installed at the blade root can directly measure the blade bending moment and shear force, and obtain the actual working conditions of the blades under the action of aerodynamic loads. However, when using the root load sensor alone, it is difficult to separate the influence brought by the wind speed change, resulting in possible deviation in the determination of the unbalanced state.

[0004] Therefore, if the LiDAR real-time wind measurement technology is combined with the root load sensor and the method of multi-sensor data fusion is used, accurate wind speed and direction information can be obtained in the same system, and at the same time, the load data reflecting the actual force condition of the blades can be obtained. However, there is no relevant report in this field at present. Summary of the Invention

[0005] The present invention provides an impeller imbalance monitoring method based on the fusion of wind speed and root load, which solves the deficiencies in the prior art. By comparing the root load data of each blade under the same wind speed condition, this method can more accurately calculate the deviation between the theoretical and actual loads, so as to identify the impeller imbalance state in real time, realize the real-time monitoring, analysis and early warning of the aerodynamic imbalance of the wind turbine, and thus improve the safety and reliability of the unit operation.

[0006] The technical solution adopted to achieve the above object of the present invention is as follows:

[0007] An impeller imbalance monitoring method based on the integration of wind speed and blade root load, comprising the following steps: (1) Install a LiDAR lidar sensor on the top of the wind turbine nacelle to measure the oncoming wind speed in front of the blade, calculate the average oncoming wind speed V0 at the hub height, and further calculate the wind speed V(h) at any height;

[0008] (2) Obtain the angular velocity ω of the current blade through the central control system of the wind turbine, and further obtain the rotational linear velocity V rot,i (r) of the point at the radius position r on any blade i;

[0009] (3) According to the height h blade (φ) from the ground of any part of the blade at different azimuth angles φ and the pitch angle θ of the blade, correct the wind speed at this any part to obtain the oncoming wind speed V blade ;

[0010] (4) According to the rotational linear velocity V rot,i (r) of any part of the blade, and the included angle β i between the wind direction and the tangential direction of the blade rotation, further correct the oncoming wind speed V blade to obtain the effective oncoming wind speed V corr，i (r, φ, θ, β i );

[0011] (5) Discretize the blade radially into micro-segments with a length of dr, and further calculate the aerodynamic force dF corr，i (r) received by this micro-segment according to the effective oncoming wind speed V i (r, φ, θ, β aero ), which is the bending moment generated by this micro-segment on the blade root. By integrating the bending moments of all micro-segments in each blade, the theoretical blade root bending moment M theory,i is obtained;

[0012] (6) Install blade root load sensors at the roots of each blade to measure the blade root bending moment and shear force in real time; compare the measured actual blade root bending moment M yi with the theoretical blade root bending moment M theory,i , calculate the load deviation ΔM i , ΔM i = M yi - M theory，i , and preprocess the load deviation ΔM i to filter out noise and outliers;

[0013] (7) For the load deviation ΔM iPerform data monitoring. Through moving window standard deviation analysis and machine learning clustering analysis, based on the results of moving window standard deviation analysis and / or machine learning clustering analysis, judge the current load deviation ΔM in real time. i Whether it exceeds the dynamic threshold range. If it exceeds the threshold range, the system will automatically trigger an alarm.

[0014] Furthermore, in step (1), the LiDAR sensor obtains wind speed data at multiple measurement points within the hub sweep area, denoted as V ∞，j (j = 1, 2,..., N), where N is the number of measurement points; further calculate the average oncoming wind speed V0 at the hub height, and the calculation formula is as follows:

[0015] Furthermore, the calculation formula for the wind speed V(h) at any height in step (1) is as follows: where h0 is the height of the hub center from the ground, and α is the wind shear index obtained by data fitting of the LiDAR sensor, which is a constant.

[0016] Furthermore, in step (2), the rotational linear velocity V rot,i (r) of the point at the radius position r on any blade i is calculated by the formula: V rot，i (t) = ω·r.

[0017] Furthermore, in step (3), the ground clearance h blade (φ) of any part of the blade at different azimuth angles φ is calculated by the formula: h blade (φ) = h0 + R·sin(φ), where R is the impeller radius and h0 is the height of the hub center from the ground; first, correct the wind speed according to h blade (φ) to obtain:

[0018] Then, further correct according to the pitch angle θ of the blade to obtain the corrected oncoming wind speed V blade :

[0019]

[0020] Furthermore, in step (4), the angle β i between the wind direction and the tangential direction of the blade rotation is calculated by the formula: where γ is the oncoming wind direction; after correction, the effective oncoming wind speed V corr，i (r, φ, θ, β i ) at this any part is obtained, and the correction calculation formula is:

[0021]

[0022] Further, the aerodynamic force dF aero received by the micro-segment in step (5) is calculated by the following formula:

[0023] where ρ is the air density, c(r) is the chord length of the blade at the position of the micro-segment in the blade; C eff (r) is the effective aerodynamic coefficient at the position of the micro-segment in the blade, which is a constant.

[0024] Further, the theoretical root bending moment M theory,i in step (5) is calculated by the following formula:

[0025] where r min is the starting position of the root, and r max is the maximum radial length of the blade.

[0026] Further, the moving window standard deviation analysis method in step (7) is as follows: First, set a moving window with the past T minutes or N data points, calculate the mean μ and standard deviation σ of the load deviation ΔM i within the window, then set the dynamic threshold upper limit T uper = μ + k·σ, and the dynamic threshold lower limit T lower = μ - k·σ, where k is an adjustable parameter for controlling the warning sensitivity; if the current load deviation ΔM i exceeds the dynamic threshold range T lower —T upper , it is considered that there is an abnormal load deviation and an alarm is automatically triggered.

[0027] Further, the machine learning clustering analysis in step (7) uses a clustering algorithm to cluster the historical load deviation ΔM i data, distinguish normal working condition data and abnormal data, determine the dynamic threshold range according to the distribution of normal working condition data, ensure that the threshold can adapt to complex working condition changes, and when the newly generated load deviation ΔM i does not belong to the normal cluster, it is marked as an abnormal load deviation and an alarm is automatically triggered.

[0028] Compared with the prior art, the impeller imbalance monitoring method based on the fusion of wind speed and root load provided by this application has the following advantages: In this application, the LiDAR real-time wind measurement technology is combined with the root load sensor, and the method of multi-sensor data fusion is used to obtain both accurate wind speed and wind direction information and load data reflecting the actual force on the blades in the same system. By comparing the root load data of each blade under the same wind speed condition, the deviation between the theoretical and actual loads can be calculated more accurately, so as to identify the impeller imbalance state in real time and provide early warnings and optimization suggestions to the operation and maintenance system in a timely manner. This technology fusion method not only improves the monitoring accuracy but also enhances the real-time diagnosis ability of the aerodynamic imbalance state, providing more effective technical support for the safe and stable operation of wind turbines.

[0029] In this application, the multi-sensor dynamic data fusion method is fully adopted. Through the dynamic fusion of the LiDAR real-time wind speed field data and the direct measurement data of the root load sensor, a real-time correction algorithm based on a physical model is constructed to eliminate the interference of wind speed fluctuations on load determination and improve the imbalance detection accuracy. In this application, a dynamic synthesis model of the local effective wind speed of the blade is constructed. Combining wind shear, pitch angle, rotation effect, and incoming flow wind direction, a dynamic effective wind speed model for each point of the blade is established to provide high-precision input for theoretical aerodynamic load calculation. In this application, based on the dynamic threshold warning mechanism adopted for the theoretical-actual load deviation, statistical methods (such as sliding window standard deviation analysis) and machine learning algorithms (cluster analysis) are introduced to set the dynamic abnormal threshold of the load deviation ΔM i to avoid the insufficient adaptability of the fixed threshold to complex working conditions. Brief Description of the Drawings

[0030] Figure 1 It is a flowchart of the impeller imbalance monitoring method provided in this application. Detailed Embodiment

[0031] The following will make a detailed and specific description of this application in combination with the drawings and specific embodiments.

[0032] In this embodiment, a method and system for monitoring the aerodynamic imbalance of a wind turbine based on the combination of LiDAR and a root load sensor are proposed. Through physical modeling and machine learning, real-time monitoring, analysis, and early warning of the aerodynamic imbalance of the wind turbine are realized, thereby improving the safety and reliability of the operation of the unit. Referring to the attached Figure 1 , the specific steps of the impeller imbalance monitoring method provided in this embodiment are as follows:

[0033] (1) Obtain the incoming flow wind speed field of the wind turbine measured by LiDAR

[0034] The LiDAR sensor is installed on the top of the wind turbine nacelle and is used to measure the incoming wind speed in front of the wind wheel. Wind speed data is obtained at multiple measurement points within the hub sweep area by LiDAR, denoted as: V ∞，j (j = 1, 2,..., N), where N is the number of measurement points. Thus, the average incoming wind speed at the hub height is calculated as:

[0035] (2) Calculation of blade rotation speed

[0036] Assume the angular velocity of the wind turbine is ω. For a point at the radius position r on blade i, its rotational linear velocity is: V rot，i (r) = ω·r. (3) Wind shear correction (Power Law model)

[0037] The variation of wind speed with height can be described by the Power Law model. The wind speed at any height h is:

[0038] where: V0 is the incoming wind speed at the hub height h0; h0 is the height of the hub center from the ground; α is the wind shear exponent (a constant obtained by fitting LiDAR data).

[0039] (4) Calculation of local blade height from the ground

[0040] At different azimuth angles φ of different blades, their local heights from the ground are determined by the impeller center height and the blade geometric position. Assume the impeller radius is D, then: h blade (φ) = h0 + R·sin(φ), where h blade (φ) is the height of the blade from the ground at the azimuth angle φ.

[0041] (5) Correction of wind speed for height and pitch angle

[0042] At the local blade position, first correct the wind speed according to the local height from the ground, obtaining:

[0043]

[0044] At the same time, considering the correction of the wind speed by the pitch angle θ at the blade position, the corrected local incoming wind speed is:

[0045]

[0046] (6) Synthesis of local effective incoming wind speed of the blade

[0047] The local effective incoming wind speed is affected by both the corrected incoming wind speed V blade and the linear velocity V rot,i (r) generated by the blade rotation. The vector synthesis method is adopted, and the included angle β i(This angle is determined by the relationship between the oncoming flow direction and the rotational tangential direction), then the local effective oncoming flow velocity at a certain position of blade i is:

[0048]

[0049] Where: β i is the angle between the wind direction and the rotational speed considered during local correction. Let the oncoming flow direction measured by LiDAR be γ (with due north as 0°, increasing counterclockwise); the current azimuth angle of the blade is φ, and its tangential velocity direction is: Therefore, the local angle β i can be expressed as:

[0050] Based on the wind speed field measured by LiDAR, combined with wind shear, pitch angle correction, and blade rotation effects, the local effective oncoming flow velocities at different positions of different blades are derived, providing a basis for further calculating the theoretical aerodynamic loads, correcting the root loads, and eliminating the influence of wind speed fluctuations on the imbalance judgment.

[0051] (7) Theoretical Aerodynamic Loads of Blades and Root Bending Moments

[0052] Based on the blade element theory of aerodynamics, the blade is discretized radially into tiny element segments. Let the length of the micro-segment at position r be dr, and the local effective oncoming flow velocity be: V corr，i (r, φ, θ, β i ), then the aerodynamic force on this micro-segment can be expressed as:

[0053] Where: ρ is the air density; c(r) is the chord length of the blade at the radial position r; C eff (r) is the effective aerodynamic coefficient at this position, usually obtained by comprehensively considering the lift coefficient C L (r), the drag coefficient C D (r), and the corresponding angle of attack, and is a constant.

[0054] Since this aerodynamic force generates a bending moment on the blade root, its theoretical bending moment at the blade root can be obtained by integrating the moments of each blade element segment, that is:

[0055]

[0056] Where: r min is the starting position of the blade root; r max is the maximum radial length of the blade. This theoretical blade root bending moment M theory,i can reflect the cumulative effect of the aerodynamic force caused by the local effective oncoming flow velocity on the blade root after considering wind shear, pitch angle correction, and blade rotation effects.

[0057] (8) Measurement of Root Loads and Judgment of Aerodynamic Imbalance

[0058] Install a root load sensor at the blade root to measure the root bending moment (e.g., M xi , M yi ) and shear force (e.g., F xi , F yi ) in real time. Compare the measured actual bending moment with the theoretical root bending moment and calculate the deviation:

[0059] ΔM i = M yi - M theory，i ;

[0060] If ΔM i deviates from the normal range for a long time, it can be determined that there is an aerodynamic imbalance problem with the blade. By this method, using accurate wind speed data to correct the root load can effectively eliminate the load changes caused by wind speed fluctuations, so as to more accurately identify the aerodynamic imbalance state of the wind turbine blade.

[0061] (9) Early warning

[0062] After obtaining the load deviation, monitor and early warn the unbalanced state of the unit in four parts:

[0063] 9.1 Data acquisition and preprocessing

[0064] Preprocess the data through the load deviation ΔM i to filter out noise and outliers and ensure data quality.

[0065] 9.2 Moving window standard deviation analysis

[0066] Set a moving window (e.g., the past T minutes or N data points), and calculate the mean μ and standard deviation σ of the load deviation within the window; set dynamic thresholds:

[0067] Upper limit: T upper = μ + k·σ

[0068] Lower limit: T lower = μ - k·σ

[0069] where k is an adjustable parameter used to control the early warning sensitivity.

[0070] If the new data ΔM i exceeds this dynamic threshold range, it is considered that there is an abnormal load deviation.

[0071] 9.3 Machine learning clustering analysis

[0072] Use a clustering algorithm (such as K-means) for historical ΔM iCluster the data to distinguish normal operating condition data from abnormal data. Determine the dynamic threshold range based on the distribution of normal operating condition data to ensure that the threshold can adapt to complex operating condition changes. When a new data point does not belong to the normal cluster, mark it as abnormal.

[0073] 9.4 Anomaly Detection and Early Warning

[0074] Based on the comprehensive results of sliding window statistical analysis and clustering analysis, continuously judge whether the current ΔM i exceeds the dynamic threshold range; if it exceeds the threshold, the system automatically triggers an early warning.

[0075] (10) Specific Implementation Case

[0076] The following table shows the data results monitored for three blades of a wind turbine in a certain wind farm using the monitoring method in this embodiment, as shown in the following table:

[0077]

[0078] As can be seen from the above table, starting from 15:00, the ΔM1 of Blade 1 increased rapidly and exceeded the set dynamic threshold of 8 kNm, so it was determined that the blade state was abnormal and there was an imbalance in Blade 1. After subsequent investigation, it was found that there was local icing on Blade 1, resulting in an abnormal increase in aerodynamic load. Therefore, the impeller imbalance monitoring method provided in this application can realize real-time monitoring, analysis and early warning of wind turbine aerodynamic imbalance, thereby improving the safety and reliability of unit operation.

Claims

1. An impeller imbalance monitoring method based on the fusion of wind speed and blade root load, characterized in that Including the following steps: (1) Install the LiDAR lidar sensor on the top of the wind turbine nacelle, measure the oncoming wind speed in front of the blade, calculate the average oncoming wind speed V0 at the hub height, and further calculate the wind speed V(h) at any height h; (2) Obtain the angular velocity ω of the current blade through the central control system of the wind turbine, and further obtain the rotational linear velocity V rot,i (r) at the point with a radius position of r on any blade i; (3) According to the height h from the ground of any part of the blade at different azimuth angles φ blade (φ) and the pitch angle θ of the blade, correct the wind speed at any part to obtain the incoming flow wind speed V at any part after correction blade ; (4) According to the rotational linear velocity V rot,i (r) at any part of the blade, and the angle β i between the wind direction and the tangential direction of the blade rotation, blade further correct the oncoming flow velocity V corr,i (r, φ, θ, β i ) to obtain the effective oncoming flow velocity V (5) Discretize the blade radially into micro-segments with a length of dr, and based on the effective oncoming flow velocity V of this micro-segment corr,i (r, φ, θ, β i ) further calculate the aerodynamic force dF received by this micro-segment aero (r), which is the bending moment generated by this micro-segment on the blade root. By integrating the bending moments of all micro-segments in each blade, the theoretical blade root bending moment M received by each blade is obtained theory,i ; (6) Install root load sensors at the roots of each blade to measure the root moment and shear force in real time; compare the measured actual root moment M yi with the theoretical root moment M theory,i to calculate the load deviation ΔM i , where ΔM i = M yi - M theory，i , and preprocess the load deviation ΔM i to filter out noise and outliers; (7) For the load deviation ΔM i Perform data monitoring. Through moving window standard deviation analysis and machine learning clustering analysis, based on the results of moving window standard deviation analysis and / or machine learning clustering analysis, judge in real time whether the current load deviation ΔM i exceeds the dynamic threshold range. If it exceeds the threshold range, the system automatically triggers an alarm.

2. The impeller imbalance monitoring method based on the fusion of wind speed and blade root load according to claim 1, wherein: In step (1), the LiDAR lidar sensor obtains wind speed data at multiple measurement points within the hub sweep area, denoted as V ∞，j (j = 1, 2,..., N), where N is the number of measurement points; further calculate the average oncoming wind speed V0 at the hub height, and the calculation formula is as follows:

3. The impeller imbalance monitoring method based on the fusion of wind speed and blade root load according to claim 1, characterized in that: The calculation formula for the wind speed V(h) at any height h in step (1) is as follows: where h0 is the height of the hub center from the ground, α is the wind shear index obtained by data fitting of the LiDAR sensor, and is a constant.

4. The impeller imbalance monitoring method based on the fusion of wind speed and blade root load according to claim 1, wherein: The rotational linear velocity V of a point at a radius position r on any blade i in step (2) rot,i (r) is calculated by the formula: V rot，i (r) = ω·r.

5. The impeller imbalance monitoring method based on the fusion of wind speed and blade root load according to claim 1, characterized in that: In step (3), the ground clearance h of any part of the blade at different azimuth angles φ blade (φ) is calculated by the formula: h plade (φ) = h0 + R·sin(φ), where R is the impeller radius and h0 is the ground clearance of the hub center; First, correct the wind speed according to h blade (φ) to obtain: Further correct it according to the pitch angle θ of the blade to obtain the corrected oncoming flow velocity V at any position blade :

6. The impeller imbalance monitoring method based on the fusion of wind speed and blade root load according to claim 1, wherein: The included angle β between the wind direction and the tangential direction of blade rotation in step (4) i has the following calculation formula: where γ is the incoming flow wind direction; the effective incoming flow wind speed V at any position is obtained after correction corr,i (r, φ, θ, β i ), and the corrected calculation formula is:

7. The impeller imbalance monitoring method based on the fusion of wind speed and blade root load according to claim 1, characterized in that: The aerodynamic force dF on the micro-segment in step (5) aero (r) is calculated by the formula: Where ρ is the air density and c(r) is the chord length of the blade at the position of the micro-segment in the blade; C eff (r) is the effective aerodynamic coefficient at the position of the micro-segment in the blade and is a constant.

8. The impeller imbalance monitoring method based on the fusion of wind speed and blade root load according to claim 7, characterized in that: In step (5), the theoretical blade root moment M theory,i is calculated by the formula: where r min is the starting position of the blade root, and r max is the maximum radial length of the blade.

9. The impeller imbalance monitoring method based on the fusion of wind speed and blade root load according to claim 1, characterized in that: The method of analyzing the standard deviation of the sliding window in step (7) is as follows: First, set a sliding window with the past T minutes or N data points, and calculate the mean μ and standard deviation σ of the load deviation ΔM within the window. i Then set the dynamic threshold upper limit T upper = μ + k·σ, and the dynamic threshold lower limit T lower = μ - k·σ, where k is an adjustable parameter used to control the warning sensitivity; if the current load deviation ΔM i exceeds the dynamic threshold range T lower —T upper , it is considered that there is an abnormal load deviation and the warning is automatically triggered.

10. The impeller imbalance monitoring method based on the fusion of wind speed and blade root load according to claim 1, characterized in that: In step (7), the machine learning clustering analysis uses a clustering algorithm to cluster the historical load deviation ΔM i data, distinguish normal operating condition data from abnormal data, determine the dynamic threshold range according to the distribution of the normal operating condition data, ensure that the threshold can adapt to complex operating condition changes, and when the newly generated load deviation ΔM i does not belong to the normal clustering, it is marked as an abnormal load deviation and an alarm is automatically triggered.

Images