Cabin radar wind speed real-time correction method and system based on atmospheric stability grading

Abstract

The application discloses a kind of based on atmospheric stability grading cabin radar wind speed real-time correction method and system.The method of the application includes the following steps:S1, stability-dependent offline pairing calibration;S2, real-time stability grade discrimination;S3, dynamic calibration coefficient selection and transition zone smoothing;S4, real-time wind speed / wind direction correction output.The system of the application includes cabin Doppler wind lidar (DWL), auxiliary sensing unit and stability grading and calibration module.The application introduces atmospheric stability into the calibration parameter space of cabin radar, establishes the dynamic mapping of stability grade ↔ calibration coefficient set, and dynamically switches the corresponding calibration coefficient set to eliminate the systematic deviation caused by the refractive index structure change, aerosol distribution offset and wind shear under different stratification conditions.The application belongs to the field of wind turbine wind measurement and yaw control technology.

Description

Technical Field

[0001] This invention belongs to the field of wind turbine wind measurement and yaw control technology, specifically involving a method and system for real-time correction of wind speed of nacelle-mounted Doppler wind lidar (DWL) based on atmospheric stability classification. By classifying the real-time atmospheric stability, the corresponding calibration coefficient group is dynamically switched to eliminate the systematic deviations caused by changes in refractive index structure, aerosol distribution shifts, and wind shear on the DWL measurement path under different stratification conditions. Background Technology

[0002] The capture efficiency of a wind turbine's rated power directly depends on the accuracy of the effective incoming wind speed and direction reaching the rotor plane. For a long time, turbines have relied on cup anemometers and wind vanes on the nacelle roof to obtain measurements. However, due to fluid interference within the nacelle itself (obstruction effect, wake distortion, and flow field induced by blade rotation), these measurements exhibit non-negligible directional deviations under most operating conditions and are prone to failure in harsh environments such as freezing, salt spray, and insect infestation.

[0003] In recent years, Doppler wind lidar (DWL) has been introduced into aircraft rooftops as a high-order wind sensor due to its non-contact, forward-looking telemetry capabilities. Typically, a DWL emits a 1550nm coherent laser in a four-beam configuration (top / bottom / left / right), and radial wind speed is obtained by processing the echo Doppler spectrum using FFT. V IOSi Then, the horizontal wind speed is inverted using the least squares solution. U lidar With wind direction .

[0004] To eliminate installation tilt angle errors, beam pointing deviations, and local flow field distortions in DWL (Digital Wavelength Helicopter), the industry standard practice is to perform a one-time calibration using a nearby meteorological tower: using the true values ​​of the cup anemometers on the tower during the synchronous time period. U mast With DWL readings U lidar Perform linear fitting to obtain fixed calibration coefficients. k 1, b 1 (and wind direction) k 2, b 2): U mast = k 1· U lidar + b 1, The same set of methods was used throughout the entire operating cycle thereafter. k 1, b 1) This is the mainstream paradigm of existing technology.

[0005] However, after in-depth research, the applicant discovered a fundamental flaw in this one-time static calibration paradigm: the calibration coefficient... k 1, b 1 is not a true constant, but rather a variable that drifts with the atmospheric thermodynamic state. Its physical origins can be traced back to the following three interlocking mechanisms: First, the refractive index structure constant C n The stratification dependence of ². The velocity measurement principle of DWL is based on the existence of a uniformly distributed tracer aerosol in the atmosphere, and the echo amplitude is proportional to the backscattering coefficient. β Based on the assumption. C n ² (refractive index turbulent structure constant) is strongly correlated with atmospheric stability: under unstable convection conditions (heated during the day, strong turbulent mixing). C n ²The significant increase and uneven vertical distribution cause slight fluctuations in the effective sample volumes at different beam elevation angles, thus affecting the accuracy of each sample. V LOS,i When projected onto the horizontal plane, a stability-related system bias is generated; during stable temperature inversion nights, aerosols are often suppressed below the near-surface layer, forming a stratified structure, and the beam passes through different... β The effective path of the layer changes, introducing a different mode of bias. This is due to the use of least squares inversion. X =( A T A ) -1 A T b right b The uniform bias of each channel is extremely sensitive, and the final output U lidar will with k 1( s The form of ) exhibits stability dependence.

[0006] Second, the implicit influence of wind shear on projection geometry. In existing technologies, it is generally assumed that the horizontal wind vector at the center of the measurement volume for each beam […]. u , v The same applies. However, under strongly stable / strongly unstable conditions, vertical wind shear... The significant increase leads to a difference in the average atmospheric velocity observed by the upper and lower beams, which is equivalent to a small stability-related deviation between the nominal pointing angle and the actual effective pointing angle of matrix A. dth ( s This deviation was... The term ultimately manifests as a constant. b 1( s ) drift.

[0007] Third, the signal-to-noise ratio (SNR) and spectral moment estimation bias. Different stability levels are accompanied by different temperature and humidity profiles and aerosol sources and sinks, leading to variations in echo SNR, which in turn affects the bias and variance structure of FFT spectral center estimation—this is particularly pronounced near low wind speeds (U < 4 m / s). b 1( s The intercept term of ) exhibits systemic oscillation.

[0008] Therefore, existing technologies are based on a single ( k 1, b 1) The fundamental problem with covering all atmospheric conditions is that the same yardstick is used under different physical mechanisms. When day and night alternate (sunrise is unstable ↔ night is stable), fronts pass through, or seasons change, the static calibration residual can accumulate and expand from ±0.1 m / s to ±0.3~0.5 m / s or even higher. Since the cube of the wind speed is included in the power calculation (P∝U³), it means that a system deviation of only 0.3 m / s can cause about 11% power estimation error under the 8 m / s condition - which is unacceptable for DWL-based yaw control and wake management. Summary of the Invention

[0009] The purpose of this invention is to overcome the shortcomings of the prior art, which uses a one-time fixed coefficient for DWL wind speed calibration and cannot track dynamic changes in atmospheric stability, resulting in measurement drift. This invention proposes a dynamic calibration coefficient switching method and system based on atmospheric stability classification, so that DWL can maintain the same level of measurement reliability as weather towers under all-weather operation.

[0010] The technical solution adopted in this invention is a real-time correction method for cabin radar wind speed based on atmospheric stability grading, which includes the following steps: S1. Stability-dependent offline pairing calibration: A nacelle Doppler wind-measuring radar is installed in the nacelle of the wind turbine. A meteorological calibration tower equipped with a cup anemometer and an ultrasonic anemometer is deployed within the wind farm. The horizontal distance between the two towers is no more than 500m, and they are located in the same prevailing wind upwind area. The calibration period is divided into several sub-periods based on atmospheric conditions and stability levels. For each stability level... s ∈{1,2,…, SThe original output wind speed sequence of DWL and the synchronous reference wind speed sequence of the weather tower were collected respectively, and the wind speed calibration coefficient set corresponding to this level was obtained by fitting using the least squares method. k 1( s ), b 1( s )] and orientation calibration coefficient group[ k 2( s ), b 2( s )],in: , Assemble the coefficients of all levels into an index table. ={[ s , k 1( s ), b 1( s ), k 2( s ), b 2( s The data is stored in the non-volatile memory of the cabin controller. S2. Real-time stability level determination: In each control window t Inside, the following stability parameters are calculated using the DWL sampling sequence or the cabin auxiliary sensor array: Turbulence intensity IT = s Vh / V h , s Vh For wind speed standard deviation, V h This represents the average wind speed at that window. Mourning-Obuhoff length L For the amount of agency business, choose one of the following three or a combination thereof: (a) Based on friction speed u Estimation of sensible heat flux: L =-( θv · u 3 ) / [ kg ( w ′ · i ′ v )],in θv This is the average virtual temperature. k ≈0.40, which is the von Kármán constant. g For gravitational acceleration, ( w ′ · i′ v This represents the virtual pulsatile flux. (b) Based on the gradient Richardson number Approximate sign determination; (c) Based on IT Practical grading criteria: when IT A value greater than 0.18 is considered unstable. s =1; when 0.10≤ IT A value ≤0.18 is considered near neutral. s =2; when IT A value <0.10 is considered stable. s =3; Output the current stability level label based on the judgment result. s t ∈{1,…, S}; S3. Selection of dynamic calibration coefficients and smoothing of the transition region: like s t Corresponding index table The only level in China s Then select the coefficient group of that level. k 1( s ), b 1( s ), k 2( s ), b 2( s )]; If the current L The value falls within the preset level boundary transition zone. L - L boundary |<Δ L hyst Then take the two adjacent levels. s a , s b The corresponding coefficient groups are weighted by distance. l Perform linear interpolation to output equivalent coefficients: k 1= λk 1( s b )+(1- l ) k 1( s a ), b 1= λb 1( s b )+(1- l ) b 1(s a ), l =| L - L sa | / | L sb - L sa |, b 2, k 2. By analogy, in, L boundary Let Δ be the boundary layer Moning-Obukhov length. L hyst To avoid hysteresis widths that frequently jump, a value of 50m to 100m is adopted; S4, Real-time wind speed / wind direction correction output: For the current output value of DWL U lidar and Perform dynamic calibration: , Will It is fed into the main control system of the wind turbine for yaw control, power curve correction and wake management.

[0011] Furthermore, in step S1, the sub-time period is divided using a sliding window method: with 10 minutes as the smallest calibration unit, historical data from the same period are traversed, and the time periods are divided according to the Morning-Obukhov length. L The interval will group the labeled data as follows: s =1: In the unstable stratification region, -∞< L <- L crit ; s =2: This is a near-neutral region. L crit ≤ L ≤+ L crit ; s =3: This is a stable stratification region, L>+ L crit ; in, L crit =50m~100m, A minimum of 30 effective paired samples (N≥30) must be accumulated within each group for a good fit. k 1( s ), b 1( s ).

[0012] Furthermore, in step S2, IT The calculation uses the horizontal wind speed sequence derived from DWL itself: for the same control window t DWL inversion wind speed within the system, and calculation of average wind speed. V h Standard deviation s Vh and turbulence intensity IT : V h =(1 / N τ )·∑ i=1 U lidar ( t i ), s Vh ={[1 / ( N τ -1)]·∑ i=1 [ U lidar ( t i )- V h ] 2} 1 / 2 , IT = s Vh / V h , And when V h When the wind speed is <2m / s, the stability assessment is skipped, and the coefficient group of the previous effective level remains unchanged to prevent issues at low wind speeds. IT Divergent.

[0013] Furthermore, in step S2, the friction speed u The estimation is based on the three-axis components of the existing cabin ultrasonic anemometer within the autonomous control system. u′ , v′ , w′ The residual covariance of ) u =[( u′w′ ) 2 +( v′w′ ) 2 ] 1 / 4 , Or, when no ultrasound device is available, it degenerates to the state described in step S2. IT Grading criteria (c) Direct grading.

[0014] Furthermore, in step S1, physical constraints are added during the coefficient fitting process: for all levelss All k 1( s )>0 and | k 1( s )-1|<0.15, b 1( s )∈[-0.5,+0.5]m / s, k 2( s )≈1, b 2( s )∈[-5°,+5°], to prevent ill-fitting from introducing non-physical solutions.

[0015] Furthermore, the method also includes an online drift monitoring step: in each control window t Calculate the correction value U corr Calculate its relationship with the previous window. U corr，prev First-order difference | U corr - U corr，prev If the difference of three consecutive windows is less than e =0.1m / s and U corr If the speed is greater than 3 m / s, it is determined that the current weather conditions are steady-state and homogeneous, triggering a lightweight online fine-tuning: using the most recent N i =30~60 groups[ U lidar , U mast ( s The current grade coefficients are updated using recursive least squares, allowing the coefficients to evolve slowly following long-term instrument drift.

[0016] Furthermore, in step S4, the corrected U corr Further wind shear consistency verification: using the ambient wind shear simultaneously measured at the nacelle. α Verification U corr The margin of the ratio between the wind speed calculated from the SCADA side pitch angle / rotation speed and the actual wind speed is considered. If it exceeds [0.85, 1.15], the window is marked as suspicious, and the interpolation coefficient is used instead of the direct calibration coefficient. Output quality flags (including at least four states: normal / transition band interpolation / suspicious downgrade / low wind speed freeze) are provided for arbitration by the main control system. Among these, environmental wind shear... α It is obtained by inversion of DWL multi-elevation echo intensity or multi-height layer.

[0017] Furthermore, the index table The lookup table (LUT) structure is embedded in the cabin controller Flash memory, with the table entries being { s , L low , L high , k 1, b 1, k 2, b 2, R ²}, where R ² The coefficients of determination fitted for each level; only one table lookup is needed for comparison during runtime. L or IT The coefficient group can be locked within the specified range, and the single selection latency is less than 1ms, which meets the refresh rate requirements of 1Hz and above.

[0018] Furthermore, the stability level s Choose 3 or 5; s When the value is 5, two semi-stable transitional levels are added to the level boundaries on both sides of the near-neutral level, making... L <-200m indicates strong instability, 200m< L <∞ represents strong stability, to cover the refractive index structure constant during extreme temperature inversion nights. C n 2 The DWL backscattering bias caused by the dramatic change.

[0019] Furthermore, the method temporarily disables stability switching during wind turbine yaw: when the yaw rate | Oh When the freezing rate is greater than 0.5° / s, the freezing level is... s t The value is the effective value before yaw. After yaw ends, the level is refreshed and reactivated within a preset window of 60s to 120s to avoid pollution from unstable wind fields induced by yaw. IT and L estimate.

[0020] Furthermore, in step S2, when using the Moning-Obukhov length... L When the proxy quantity (a) or TI criterion (c) is difficult to stably determine the level, the gradient Richardson number is adopted. Ri g Perform auxiliary rating: when Ri g When the value is less than 0, it is considered an unstable stratification. Ri g When ≈0, it is considered nearly neutral. Ri g When the value is greater than 0, it is considered a stable stratification, and | Ri g The size of the | distinguishes between strong and weak stratification. Ri g The rating result andL , IT The grading results are combined by majority vote or priority, and the final grade label is output. s t .

[0021] Furthermore, in step S1, at the stability level s Based on the average wind speed V h Wind speed zones b ∈{1,2,…, B}, for each ( s , b Two-dimensional units are fitted to the calibration coefficient set respectively. k 1( s , b ), b 1( s , b ), k 2( s , b ), b 2( s , b [], based on real-time stability level during runtime s t A coefficient group is selected by looking up a table together with the real-time wind speed segment to further compensate for the intercept drift caused by the decrease in signal-to-noise ratio in the low wind speed segment.

[0022] Furthermore, step S2 introduces a priori timeframes based on day and night: combining local sunrise and sunset times, the priori weights for determining stable stratifications are increased during the nighttime hours, and the priori weights for determining unstable stratifications are increased around noon. This timeframe prior is then combined with… IT or L The measured classification results are weighted and fused to suppress erroneous jumps in stability levels during the sunrise / sunset transition period.

[0023] Furthermore, the quality flag is a multi-level flag, including at least four states: normal, transition band interpolation, suspected degradation, and low wind speed freeze. The main control system adjusts the flags according to different flag levels. U corr Different confidence weights are assigned for data arbitration in yaw wind control, power curve correction, and wake management.

[0024] A cabin Doppler wind-measuring radar wind measurement system for performing the method described above, the system comprising: The nacelle Doppler wind radar, installed on the top panel of the nacelle, emits a laser beam forward and upward to output radial wind speed. V LOS With inverted horizontal wind speed U lidar ,wind direction ; Auxiliary sensing unit: temperature, humidity and pressure sensor and optional ultrasonic anemometer, outputting environmental parameters for stability calculation; The stability grading and calibration module is configured to perform the method steps described above and output the corrected value. U corr and To the main control system.

[0025] Compared with existing technologies (i.e., static calibration methods based on fixed coefficients), this invention represents a significant improvement in measurement accuracy, environmental adaptability, and control reliability. The specific beneficial effects are described below: 1. This invention overcomes the inherent limitations of static calibration by establishing a mapping relationship between environmental conditions and measurement deviations. Existing technologies generally employ one-time static linear calibration, implicitly assuming that the calibration coefficients are constant and independent of the environment. This invention, for the first time, introduces atmospheric stability into the calibration parameter space of the cabin radar, establishing stability levels. The dynamic mapping model of the calibration coefficient set breaks the limitation of the traditional technology of using a single standard to measure everything, and eliminates the systematic measurement error caused by environmental state drift (such as day-night cycle, frontal passage) from a physical mechanism perspective.

[0026] 2. Significantly improves wind speed measurement accuracy across all wind speed ranges and under all weather conditions, with effective error control capability improved by approximately one order of magnitude. Quantitative effect: This invention reduces the absolute error of wind speed measurement by the cabin radar under different atmospheric strata from 0.25–0.5 m / s in existing technologies to within 0.08–0.12 m / s. Power calculation advantage: Given… P ∝ U 3 (Therefore) dP / P ≈3· dU / U This invention eliminates the 0.3 m / s level deviation generated during the stability switching of static calibration, thereby avoiding approximately 8% to 15% (approximately 11% for the 8 m / s operating condition) of power estimation error, enabling the unit's main control system to perform power tracking based on real wind resource data, and directly improving wind energy capture efficiency.

[0027] 3. By introducing hysteresis detection and interpolation smoothing algorithms, the continuity and stability of the control signal are ensured. To address the potential high-frequency, small-amplitude fluctuations in stability parameters near level boundaries, this invention sets a hysteresis width. ΔL hyst Furthermore, in the transition zone, equivalent coefficients are dynamically generated using linear interpolation based on distance weights, rather than a simple step switch, to ensure... U corrIt is physically continuous, meeting the main control system's requirements for input signal smoothness and real-time performance (selecting a delay of <1ms).

[0028] 4. An online drift monitoring and multi-level quality flag mechanism have been added, enhancing the system's robustness and adaptability. The system can identify long-term hardware drift of the instruments and update parameters using the recursive least squares method. At the same time, it eliminates abnormal data through wind shear consistency verification. The output multi-level quality flag provides the main control system with the basis for data reliability arbitration. When yaw interference or suspicious data is detected, the system automatically degrades the quality flag to prevent erroneous wind speed data from misleading yaw control and to ensure the safe operation of the unit.

[0029] This invention targets the closest prior art (i.e., using a single fixed coefficient). k 1, b 1) A method for one-time static calibration of DWL), the distinguishing technical features of this invention are: (i) classifying the calibration data according to atmospheric stability, fitting an independent calibration coefficient set for each stability level, and constructing an index table of stability level ↔ calibration coefficient set; (ii) online real-time determination of atmospheric stability level and dynamic switching of coefficient set accordingly; (iii) setting a hysteresis band at the level boundary and performing distance-weighted interpolation to ensure continuous output. The technical effect brought about by the above distinguishing technical features, which is difficult for those skilled in the art to anticipate, is that the existing technology has long assumed that the calibration coefficient is a constant independent of the environment, while this invention reveals and utilizes the neglected physical law of calibration coefficient drift with stability, eliminating the systematic measurement deviation under cross-layer structure conditions from the generation mechanism (rather than post-processing filtering), and improving the all-weather measurement accuracy by about an order of magnitude. In summary, by using atmospheric stability as the dynamic dimension of calibration, this invention creatively solves the long-standing problem of environmental state-measurement bias coupling in the field, enabling the cabin Doppler wind radar to maintain measurement accuracy on par with meteorological towers under all-weather operation, resulting in significant economic benefits and technological advancements. Attached Figure Description

[0030] Figure 1 This is a simplified structural block diagram of the system of the present invention; Figure 2 This is a logic control flowchart of the method of the present invention. Detailed Implementation

[0031] This invention transfers the calibration coefficient from a single point (a single...) k 1, b 1) Expand into a family of functions indexed by stability {( s , k 1( s ), b 1( s ))}, determined in real time during runtime st Look up the corresponding coefficients in the table, and perform interpolation smoothing in the transition zone if necessary.

[0032] like Figure 1 As shown, the nacelle Doppler wind radar system of the present invention includes a nacelle Doppler wind radar (DWL), which is installed on the top plate of the nacelle and emits a laser beam forward and upward to output radial wind speed. V LOS With inverted horizontal wind speed U lidar ,wind direction ; Auxiliary sensing unit 1: temperature, humidity and pressure sensor and optional ultrasonic anemometer, outputting environmental parameters for stability calculation; Stability grading and calibration module 2 is configured to perform the following method steps and output the corrected value. U corr and To the main control system 3.

[0033] The method of the present invention will be described in more detail below with specific embodiments.

[0034] Example 1: 3MW onshore unit, S=3, three-level classification

[0035] The method of the present invention includes the following steps: S1. Stability-dependent offline pairwise calibration (establishing a coefficient lookup table)

[0036] After installing the DWL (Digital Weather Monitoring System) on the nacelle roof, and utilizing synchronous observation data from a metrologically certified meteorological calibration tower (≤500m from the unit, in the same upwind region as the prevailing wind), the historical operating period was divided into S=3 levels of sub-periods (unstable / near-neutral / stable) based on atmospheric stability. For each level s, the following was performed independently: collecting synchronous data pairs; and using least squares fitting to obtain... U mast ( s )= k 1( s )· U lidar + b 1( s Similarly, fitting k 2( s ), b 2( s ); Calculate the goodness of fit for this level. R 2 As a quality control threshold ( R 2 (A value >0.95 is acceptable for storage). A final stability-coefficient lookup table is then created. In the cabin controller Flash:

[0037] The numerical ranges in the table above are merely examples of the order of magnitude of typical wind field measurements and are not restrictive.

[0038] S2. Real-time stability level determination: In each control window t (Preferably within 600 seconds, or 10 minutes): Path A (with ultrasound / flux meter) — based on L Estimate friction speed u With sensible heat flux ( w′θ v ′ ),calculate L =-( θv · u 3 ) / [ kg ( w ′ · i ′ v )], In the formula k ≈0.40 is the von Kármán constant, derived from... L Determining which interval the sign and absolute value fall into s t .

[0039] Path B (degradation run using only DWL's own data) — based on IT Classification: TI= s Vh / V h ; when IT When >0.18, s t =1 (unstable); when 0.10≤ IT When ≤0.18, s t =2 (near neutral); when IT When <0.10, s t =3 (stable); And a low wind speed latch is added: when V h When the speed is less than 2m / s, the handover is paused, and the previous valid handover is maintained. s t In case of IT Divergence-based misjudgment. When the judgment level is unstable, gradient Richardson's number can be used as a supplement. Rig It is adjusted by integrating prior knowledge of day and night time periods.

[0040] S3. Coefficient selection and transition region smoothing: Lookup table have to s t Corresponding coefficient group; if the current L At the level boundary ±Δ L hyst Within the hysteresis band, take the coefficients of two adjacent levels as... l -Weighted linear interpolation output equivalent ( k 1, b 1) Guarantee U corr continuous.

[0041] S4. Real-time correction and output: , Will Send to the main controller to replace the original U lidar Participate in yaw wind and power curve compensation and wake control, and record ( U corr , , s t (Quality mark) to the circular buffer for future reference.

[0042] The following explanation, in conjunction with the hardware configuration, further illustrates this: The DWL on the top of the cabin sequentially drives four telescopes via optical switches, controlling the azimuth angle... f ≈±30°, pitch angle i ≈±12.5°, FFT points 512, interpolation zero-padded to 1024, output V LOS1 to V LOS4 The original data is obtained through least squares inversion. U lidar and The data is sent to the engine room PLC (main frequency ≥ 800MHz). The same PLC reads the engine room temperature, humidity, and pressure, as well as the optional ultrasonic module, via Modbus.

[0043] After grid connection, maintenance personnel select continuous time periods covering day and night and various wind conditions, and perform sliding grouping with 10-minute windows to exclude precipitation, shading, and other factors. SNR After a bad value of <-25dB, a linear fit was performed using the true value from the tower cup anemometer, resulting in level 1 (unstable, warm zone during the day). L =-120m example) k 1(1) = 1.042 b 1(1) = +0.15 m / s; Level 2 (near neutral, morning / evening transition, | L |=28m) k1(2) = 1.004 b 1(2) = +0.01 m / s; Level 3 (Stable inversion night, L =+185m) k 1(3) = 0.973 b 1(3) = -0.08 m / s, together with IT The boundary (0.18 / 0.10) is burned into the Flash.

[0044] Example 2: 5MW offshore turbine unit, S=5 five-level subdivision In response to temperature inversion at sea at night C n 2 In cases of dramatic changes and significant aerosol stratification, this embodiment uses S=5. Based on the three levels of Embodiment 1, two transitional levels, strongly unstable (L<-200m) and strongly stable (L>+200m), are added on both sides of the near-neutral level. Each level is further categorized according to the average wind speed. V h The wind speed is divided into four ranges: 3–5 m / s, 5–8 m / s, 8–12 m / s, and >12 m / s, forming ( s , b A two-dimensional coefficient table is used to compensate for the estimation bias of the spectrum center in the low wind speed range. Real-time operation is employed during runtime. s t The coefficient set was selected by looking up a table together with the real-time wind speed segment, and the a priori function for day and night periods was used to suppress erroneous jumps during the sunrise / sunset transition period. Field measurements show that the five-level subdivision can further reduce the absolute wind speed error by approximately 20%–30% during the highly stable nighttime period compared to the three-level scheme.

[0045] Comparison with the present invention Comparative example (traditional static calibration): By performing a unique global fit on all data over the entire 10-day period, we obtain... k 1 = 1.008, b 1 = +0.03. This ( k 1, b 1) When applied to the nighttime stable phase: The actual relationship measured at night is k 1(3) = 0.973, b 1(3) = -0.08, while the global fit forces (1.008, +0.03), which will produce a positive bias. Δ≈(1.008-0.973)×U≈+0.035×8m / s≈+0.28m / s; This means that DWL has always assumed that the wind is stronger than it actually is, and the main controller yaws accordingly to counter the wind, resulting in a systematic over-biasing of the actual yaw angle; the wake model uses this over-biased U as the inflow condition, which leads to an underestimation of the wake loss ΔV / U, causing the wake control strategy to be biased towards drag reduction, resulting in the downstream unit power loss not being fully recovered.

[0046] The processing of this invention: nighttime L >50m s t If the value is 3, it will automatically switch to (0.973, -0.08), and the deviation will be compressed from +0.28 m / s to about +0.03 to 0.05 m / s, which is about 1 / 5 to 1 / 8 of the comparative example.

[0047] Comparative conclusion: Static calibration represented by a scale will systematically deceive itself when crossing stability boundaries. This invention uses weather tower data as the truth anchor and stability as the switching dimension, decoupling the different physical mechanisms that were originally mixed together into their respective calibration subspaces, thereby restoring the consistency and reliability of DWL measurements in an all-weather sense.

[0048] Finally, it should be emphasized that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for real-time correction of cabin radar wind speed based on atmospheric stability grading, characterized in that, The method includes the following steps: S1. Stability-dependent offline pairing calibration: A nacelle Doppler wind-measuring radar is installed in the nacelle of the wind turbine. A meteorological calibration tower equipped with a cup anemometer and an ultrasonic anemometer is deployed within the wind farm. The calibration period is divided into several sub-periods based on atmospheric conditions and stability levels. For each stability level... s ∈{1,2,…, S The original output wind speed sequence of DWL and the synchronous reference wind speed sequence of the weather tower were collected respectively, and the wind speed calibration coefficient set corresponding to this level was obtained by fitting using the least squares method. k 1( s ), b 1( s )] and orientation calibration coefficient group[ k 2( s ), b 2( s ] , assemble the coefficients of all levels into an index table Stored in the non-volatile memory of the cabin controller; S2. Real-time stability level determination: In each control window τ Inside, the following stability parameters are calculated using the DWL sampling sequence or the cabin auxiliary sensor array: Turbulence intensity TI = σ Vh / V h , σ Vh For wind speed standard deviation, V h This represents the average wind speed at that window. Mourning-Obuhoff length L For the amount of agency business, choose one of the following three or a combination thereof: (a) Based on friction speed u Estimation of sensible heat flux; (b) Based on the gradient Richardson number Ri g Approximate sign determination; (c) Based on TI Practical grading criteria; Output the current stability level label based on the judgment result. s t ∈{1,…, S }; S3. Selection of dynamic calibration coefficients and smoothing of the transition region: like s t Corresponding index table ℒ The only level in China s Then select the coefficient group of that level. k 1( s ), b 1( s ), k 2( s ), b 2( s )]; If the current L The value falls within the preset level boundary transition zone. L - L boundary |<Δ L hyst Then take the two adjacent levels. s a , s b The corresponding coefficient groups are weighted by distance. λ Perform linear interpolation to output equivalent coefficients: k 1= λk 1( s b )+(1- λ ) k 1( s a ), b 1= λb 1( s b )+(1- λ ) b 1( s a ), λ =| L - L sa | / | L sb - L sa |， b 2, k 2. By analogy, in, L boundary Let Δ be the boundary layer Moning-Obukhov length. L hyst To avoid hysteresis widths that frequently jump, a value of 50m to 100m is adopted; S4, Real-time wind speed / wind direction correction output: For the current output value of DWL U lidar and Perform dynamic calibration: ， Will It is fed into the main control system of the wind turbine for yaw control, power curve correction and wake management.

2. The method according to claim 1, characterized in that, In step S1, the horizontal distance between the cup anemometer and the meteorological calibration tower is no more than 500m, and they are located in the same prevailing wind upwind area, wherein: ， Index Table ={[ s , k 1( s ), b 1( s ), k 2( s ), b 2( s )]}, The sub-time period is divided using the sliding window method: with 10 minutes as the smallest calibration unit, historical data from the same period are traversed, and the time periods are divided according to the Morning-Obukhov length. L The interval will group the labeled data as follows: s =1: In the unstable stratification region, -∞< L <- L crit ; s =2: This is a near-neutral region. L crit ≤ L ≤+ L crit ; s =3: This is a stable stratification region, L>+ L crit ; in, L crit =50m~100m, A minimum of 30 effective paired samples (N≥30) must be accumulated within each group for a good fit. k 1( s ), b 1( s ).

3. The method according to claim 1, characterized in that, In step S2 TI The calculation uses the horizontal wind speed sequence derived from DWL itself: for the same control window τ DWL inversion wind speed within the system, and calculation of average wind speed. V h Standard deviation σ Vh and turbulence intensity TI : V h =(1 / N τ )·∑ i=1 U lidar ( t i ), σ Vh =｛[1 / ( N τ -1)]·∑ i=1 [ U lidar ( t i )- V h ] 2 ｝ 1 / 2 , TI = σ Vh / V h ， And when V h When the wind speed is <2m / s, the stability assessment is skipped, and the coefficient group of the previous effective level remains unchanged to prevent issues at low wind speeds. TI Divergence; Calculate the Möning-Obukhov length L The three methods for increasing agency volume are as follows: (a) Based on friction speed u Estimation of sensible heat flux: L =-( θv · u 3 ) / [ kg ( w ′ · θ ′ v )],in θv This is the average virtual temperature. k ≈0.40, which is the von Kármán constant. g For gravitational acceleration, ( w ′ · θ ′ v This represents the virtual pulsatile flux. (b) Based on the gradient Richardson number Approximate sign determination; (c) Based on TI Practical grading criteria: when TI A value greater than 0.18 is considered unstable. s =1; when 0.10≤ TI A value ≤0.18 is considered near neutral. s =2; when TI A value <0.10 is considered stable. s =3.

4. The method according to claim 3, characterized in that, In step S2, the friction speed u The estimation is based on the three-axis components of the existing cabin ultrasonic anemometer within the autonomous control system. u′ , v′ , w′ The residual covariance of ) u =[( u′w′ ) 2 +( v′w′ ) 2 ] 1 / 4 ， Or, when no ultrasound device is available, it degenerates to the state described in step S2. TI Grading criteria (c) Direct grading.

5. The method according to claim 1, characterized in that, In step S1, physical constraints are added to the coefficient fitting process: for all levels s All k 1( s )>0 and | k 1( s )-1|<0.15, b 1( s )∈[-0.5,+0.5]m / s, k 2( s )≈1, b 2( s )∈[-5°,+5°], to prevent ill-fitting from introducing non-physical solutions.

6. The method according to claim 1, characterized in that, The method also includes an online drift monitoring step: in each control window τ Calculate the correction value U corr Calculate its relationship with the previous window. U corr，prev First-order difference | U corr - U corr，prev If the difference of three consecutive windows is less than ε =0.1m / s and U corr If the speed is greater than 3 m / s, it is determined that the current weather conditions are steady-state and homogeneous, triggering a lightweight online fine-tuning: using the most recent N i =30~60 groups[ U lidar , U mast ( s The current grade coefficients are updated using recursive least squares, allowing the coefficients to evolve slowly following long-term instrument drift.

7. The method according to claim 1, characterized in that, In step S4, the corrected U corr Further wind shear consistency verification: using the ambient wind shear simultaneously measured at the nacelle. α Verification U corr The margin of the ratio between the wind speed calculated from the SCADA side pitch angle / rotation speed and the wind speed calculation is used. If it exceeds [0.85, 1.15], the window is marked as suspicious, and the interpolation coefficient is used instead of the direct calibration coefficient. The output quality flag is used for arbitration by the main control system. Among them, the environmental wind shear is... α It is obtained by inversion of DWL multi-elevation echo intensity or multi-height layer.

8. The method according to claim 1, characterized in that, The index table The lookup table (LUT) structure is embedded in the cabin controller Flash memory, with the table entries being { s , L low , L high , k 1, b 1, k 2, b 2, R ²}, where R ² The coefficients of determination fitted for each level; only one table lookup is needed for comparison during runtime. L or TI The coefficient group can be locked within the specified range, and the single selection latency is less than 1ms, which meets the refresh rate requirements of 1Hz and above.

9. The method according to claim 1, characterized in that, The stability level s Choose 3 or 5; s When the value is 5, two semi-stable transitional levels are added to the level boundaries on both sides of the near-neutral level, making... L <-200m indicates strong instability, 200m< L <∞ represents strong stability, to cover the refractive index structure constant during extreme temperature inversion nights. C n 2 The DWL backscattering bias caused by the dramatic change.

10. The method according to claim 1, characterized in that, The method described above temporarily disables stability switching during wind turbine yaw: when the yaw rate | Ω When the speed is greater than 0.5° / s, the freezing level is... s t The value is the effective value before yaw. After yaw ends, the level is refreshed and reactivated within a preset window of 60s to 120s to avoid pollution from unstable wind fields induced by yaw. TI and L estimate.

11. The method according to claim 1, characterized in that, In step S2, when the Moning-Obukhov length is used... L When the proxy quantity (a) or TI criterion (c) is difficult to stably determine the level, the gradient Richardson number is adopted. Ri g Perform auxiliary rating: when Ri g When the value is less than 0, it is considered an unstable stratification. Ri g When ≈0, it is considered nearly neutral. Ri g When the value is greater than 0, it is considered a stable stratification, and | Ri g The size of the | distinguishes between strong and weak stratification. Ri g The rating result and L , TI The grading results are combined by majority vote or priority, and the final grade label is output. s t .

12. The method according to claim 1, characterized in that, In step S1, at the stability level s Based on the average wind speed V h Wind speed zones b ∈{1,2,…, B }, for each ( s , b Two-dimensional units are fitted to the calibration coefficient set respectively. k 1( s , b ), b 1( s , b ), k 2( s , b ), b 2( s , b [], based on real-time stability level during runtime s t A coefficient group is selected by looking up a table together with the real-time wind speed segment to further compensate for the intercept drift caused by the decrease in signal-to-noise ratio in the low wind speed segment.

13. The method according to claim 1, characterized in that, Step S2 introduces a time-night prior: combining local sunrise and sunset times, the prior weight for determining a stable stratification is increased during the nighttime period, and the prior weight for determining an unstable stratification is increased around noon. This time-night prior is then combined with… TI or L The measured classification results are weighted and fused to suppress erroneous jumps in stability levels during the sunrise / sunset transition period.

14. The method according to claim 7, characterized in that, The quality flags are multi-level flags, including at least four states: normal, transition band interpolation, suspected degradation, and low wind speed freeze. The main control system adjusts the flags accordingly. U corr Different confidence weights are assigned for data arbitration in yaw wind control, power curve correction, and wake management.

15. A cabin Doppler wind-measuring radar wind measurement system for performing the method as described in any one of claims 1 to 14, characterized in that, The system includes: The nacelle Doppler wind radar (DWL), installed on the top panel of the nacelle, emits a laser beam forward and upward to output radial wind speed. V LOS With inverted horizontal wind speed U lidar ,wind direction ; Auxiliary sensing unit (1): temperature, humidity and pressure sensor and optional ultrasonic anemometer, outputting environmental parameters for stability calculation; The stability grading and calibration module (2) is configured to perform the steps of the method described in any one of claims 1 to 14 and output the corrected value. U corr and To the main control system (3).

Images