A multi-wind-direction instrument self-correction method based on group consensus and credibility gating

Abstract

The application discloses a kind of multi-wind direction instrument self-correction methods based on group consensus and credibility gate, comprising the following steps: S1, multi-source data acquisition and sliding window construction;S2, data quality double-dimension modeling;S3, group consensus center construction;S4, credibility gate drift self-correction;S5, divergence suppression and global freezing mechanism;S6, double-layer abnormal data rejection mechanism;S7, anti-loop error vector fusion output;S8, closed-loop iteration and yaw control.The application gradually and safely updates the drift parameter of each wind direction indicator by constructing group consensus as virtual true value reference and driving through credibility gate mechanism, cooperates divergence suppression, double-layer abnormal rejection and hysteresis recovery mechanism, forms robust self-correction closed loop, and can fundamentally solve the three major core pain points of low precision, drift without online correction and poor fault tolerance of wind direction measurement.

Description

Technical Field

[0001] This invention relates to the technical field of wind turbine generator control, and in particular to a multi-wind vane self-calibration method based on group consensus and credibility gating. Background Technology

[0002] Wind energy, as a clean and renewable energy source, is a core support for the global energy transition. The power generation efficiency and operational stability of wind turbine generators are the core technological directions of the wind power industry. Wind direction measurement is the core input for wind turbine yaw control. Accurate wind direction data can drive the wind turbine to align with the prevailing wind direction in real time, maximizing wind energy capture and reducing unit fatigue load. If there is a deviation in wind direction measurement, it will directly lead to yaw alignment error, significantly reducing wind energy utilization, aggravating mechanical wear of the yaw mechanism and transmission chain, and even causing abnormal unit shutdown.

[0003] Currently, the mainstream solutions in the industry fall into two categories, both of which have insurmountable technical flaws:

[0004] 1. Single Anemometer Measurement Solution: Traditional wind turbines generally use a single mechanical or ultrasonic anemometer to collect wind direction data, which has three major problems:

[0005] Measurement accuracy is easily affected by environmental interference. Measurement errors increase significantly under harsh conditions such as high wind speed, turbulence, dust, and high humidity. Stability is extremely poor under low wind speed and sudden wind direction changes.

[0006] Long-term operation is prone to zero drift and range drift, and there is no online calibration capability, resulting in continuous accumulation of measurement deviations;

[0007] Without a redundancy fault tolerance mechanism, if the wind vane malfunctions, data gets stuck, or transmission is abnormal, the system will directly execute yaw actions based on erroneous data, causing serious wind deviation and even causing the unit to shut down.

[0008] 2. Existing multi-wind vane redundancy solutions: To address the shortcomings of single-sensor solutions, some solutions employ dual or multiple wind vane configurations. However, these solutions only perform basic processing such as simple arithmetic averaging and binary selection on the multiple data streams, resulting in four major flaws:

[0009] Without a quantitative consensus assessment mechanism, it is impossible to accurately determine the reliability of data from each sensor.

[0010] Without stability quantitative analysis, it is impossible to distinguish between normal and abnormal sensors with fluctuating data.

[0011] The problem of online self-calibration of wind vane drift cannot be solved. Drift correction relies on external high-precision reference sensors or requires manual calibration after shutdown, and closed-loop online calibration cannot be achieved.

[0012] In summary, existing technologies cannot fundamentally solve the three core problems of low accuracy in wind direction measurement, lack of online drift correction, and poor fault tolerance, making it difficult to meet the requirements for long-term, efficient, and stable operation of wind turbines. Therefore, there is an urgent need to develop a multi-wind vane self-calibration method and system with consistency assessment, drift self-correction, anomaly fault tolerance, and high-precision fusion. Summary of the Invention

[0013] The purpose of this invention is to overcome the shortcomings of the prior art and provide a multi-wind vane self-calibration method based on group consensus and credibility gating. By constructing group consensus as a virtual truth benchmark and driving the progressive and secure update of the drift parameters of each wind vane through a credibility gating mechanism, combined with divergence suppression, double-layer anomaly removal and hysteresis recovery mechanisms, a robust self-calibration closed loop is formed, which can fundamentally solve the three core pain points of low accuracy of wind direction measurement, lack of online drift correction and poor fault tolerance.

[0014] The objective of this invention is achieved through the following technical solution:

[0015] A multi-wind instrument self-calibration method based on group consensus and credibility gating includes the following steps:

[0016] S1. Deploy no fewer than two wind vanes on the top of the wind turbine nacelle, synchronously collect the raw wind direction angle data of each wind vane, construct a sliding time window based on the raw wind direction angle data and update it in real time, and output the raw wind direction angle data within the current sliding time window.

[0017] S2. Based on the original wind direction angle data within the current sliding time window, calculate the consistency index and stability weight of each wind vane.

[0018] S3. Based on the consistency index and stability weight obtained in step S2, calculate the comprehensive credibility weight of a single wind vane, and optimize based on the comprehensive credibility weight to obtain the group consensus benchmark that minimizes the total weighted angle dispersion of the wind vane in the current period.

[0019] S4. Calculate the annular angle deviation of each wind vane based on the group consensus benchmark and the original wind direction angle data obtained in step S3; calculate the credibility gating coefficient based on the consistency index and stability weight obtained in step S2; iteratively update the drift estimate and drift rate of each wind vane based on the annular angle deviation and credibility gating coefficient, and use the updated drift estimate to compensate for the original wind direction angle data to obtain the corrected wind direction data.

[0020] S5. Determine whether the corrected wind direction data and the current number of valid anemometers meet the preset freeze trigger conditions. If not, the global freeze mechanism is not triggered, and the current drift parameter update state is maintained. If it is met, the global freeze mechanism is triggered, the iterative update of the drift estimate and drift rate is stopped, and the drift estimate and drift rate are restricted. When the trigger conditions are lifted, the gradual recovery phase begins.

[0021] S6. Based on the consistency index and stability weight, perform the first-level statistical anomaly judgment on the wind vane, and based on the group consensus benchmark and the corrected wind direction data, perform the second-level structural anomaly judgment on the wind vane; if any judgment result is abnormal, mark the corresponding wind vane as an abnormal wind vane and remove it from the data fusion of the current period.

[0022] S7. Based on the corrected wind direction data corresponding to the valid wind vanes selected in step S6, and combined with the comprehensive credibility weight in step S3, the data is weighted and fused to obtain the fused wind direction data.

[0023] S8. The fused wind direction data obtained in step S7 is used as the initial search reference value for solving the consensus benchmark in step S3 of the next cycle, and the updated drift estimate and drift rate are stored to achieve continuous closed-loop iteration across cycles. The main control system of the wind turbine generates corresponding yaw control commands based on the yaw error between the fused wind direction data and the nacelle orientation, and drives the yaw system of the wind turbine to perform wind-fighting actions.

[0024] Furthermore, for the abnormal wind vanes that were removed in step S6, the calculation in step S2 is repeated; when the consistency index and stability weight of the abnormal wind vane reach the preset recovery threshold and the duration reaches the preset duration, the abnormal wind vane is remarked as a valid wind vane and re-included in the fusion calculation.

[0025] Furthermore, in step S2, the formula for calculating the consistency index is:

[0026] ;

[0027] in, As a consistency indicator; | | represents the absolute difference in the original wind direction angle between the i-th and j-th wind vanes;

[0028] The formula for calculating the stability weight is:

[0029] ;

[0030] in, For stability weights; This represents the standard deviation of all raw wind direction angle data for this wind vane within the current sliding time window. These are dimensionless coefficients used to convert the standard deviation into a dimensionless coefficient. The mapping is to dimensionless weights.

[0031] Furthermore, in step S3, the calculation formula for the group consensus benchmark is as follows:

[0032] ;

[0033] in, As a benchmark for group consensus; Let be the overall reliability weight of the i-th wind vane. , As a consistency indicator, For stability weights; It is a circular angular distance function. , The value range is [0°, 180°].

[0034] Furthermore, step S4 includes the following steps:

[0035] S4.1 Calculate the arithmetic mean of the original wind direction angle data within the current sliding time window of each wind vane. ;

[0036] The arithmetic mean is calculated using the circular difference formula. With group consensus benchmark The annular angular deviation is shown in the following formula:

[0037] ;

[0038] in, The circumferential angle deviation has a range of values ​​(−180°, +180°).

[0039] S4.2 Calculate the credibility gating coefficient based on the consistency index and stability weight, as shown in the following formula:

[0040] ;

[0041] in, This is the credibility gating coefficient; As a consistency indicator; For stability weights; and For sensitivity adjustment index, and >0, >0;

[0042] S4.3, Using credibility gating coefficients For ring angle deviation After weighting, the drift estimates and drift rates of each anemometer are updated as shown in the following formula:

[0043] ;

[0044] ;

[0045] in, The update cycle for the sliding time window; This is the updated drift estimate; Drift rate; The drift estimate before the update;

[0046] S4.4. The original wind direction angle data is compensated using the updated drift estimate to obtain the corrected wind direction data, as shown in the following formula:

[0047] ;

[0048] in, This is the corrected wind direction data; This is the original wind direction angle data.

[0049] Furthermore, in step S5, the preset freeze trigger condition includes: the variance of the corrected wind direction data. Exceeding the preset divergence threshold Or the number of valid wind vanes that have not yet been removed is less than the preset safety number Nmin;

[0050] When the variance of the corrected wind direction data Not exceeding the preset divergence threshold If the number of valid wind vanes that have not been removed is not less than the preset safe number Nmin, the preset freeze trigger condition is not met. In this case, the global freeze mechanism is not triggered and the current drift parameter update state is maintained.

[0051] When the variance of the corrected wind direction data Exceeding the preset divergence threshold If the number of valid wind vanes that have not yet been removed is lower than the preset safety number Nmin, the preset freeze trigger condition is met. At this time, the global freeze mechanism is triggered, immediately stopping the iterative update of the drift estimate and drift rate, keeping the drift parameters of the previous cycle unchanged, and forcibly freezing the absolute value of the drift estimate. |Limited to the maximum allowable drift value Within, and force the absolute value of the drift rate | |Limited to the maximum allowable drift rate within;

[0052] After the frozen state is maintained for a preset time, the variance of the corrected wind direction data Preset divergence threshold If the number of valid anemometers that have not yet been removed is greater than or equal to the preset safety number Nmin, the system enters the slow-release recovery phase, specifically as follows:

[0053] In the current cycle, only the group consensus benchmark calculation in step S3 is performed. If the deviation between the new group consensus benchmark and the group consensus benchmark maintained during the freeze period does not exceed the preset deviation value, the freeze is completely lifted and normal drift parameter updates are resumed. Otherwise, the freeze status is maintained until the next cycle for re-evaluation.

[0054] Furthermore, in step S6, a first-level statistical anomaly judgment is performed on the wind vane based on the consistency index and stability weight to determine whether the wind vane has any inherent quality problems, as shown in the following formula:

[0055] ≤ or ≤ ;

[0056] in, As a consistency indicator; For stability weights; and These are the dynamic removal thresholds for stability weight and consistency index, respectively;

[0057] Based on the consensus benchmark and the corrected wind direction data, a second-level structural anomaly assessment is performed on the wind vane to determine whether the wind vane still significantly deviates from the consensus even after drift compensation, as shown in the following formula:

[0058] | - |> ;

[0059] in, This is the corrected wind direction data; As a benchmark for group consensus; The structural anomaly threshold;

[0060] When performing the second-layer structure anomaly judgment, if the removal operation will cause the number of valid wind vanes to fall below the preset safe number Nmin, then the removal will be abandoned and the global freeze mechanism in step S5 will be triggered to retain the data of the minimum number of wind vanes to participate in the fusion, and at the same time, an "insufficient redundancy" alarm will be reported.

[0061] Furthermore, in step S7, the corrected wind direction data corresponding to the valid anemometers selected in step S6 are converted into unit vectors in the Cartesian coordinate system and weighted averaged to obtain a weighted average unit vector, as shown in the following formula:

[0062] ;

[0063] ;

[0064] in, and These are the x-axis and y-axis components of the weighted average unit vector, respectively. Let be the overall reliability weight of the i-th wind vane; This is the corrected wind direction data;

[0065] The weighted average unit vector is then converted back to the standard wind direction angle to obtain the fused wind direction data within the standard definition range, as shown in the following formula:

[0066] ;

[0067] in, If wind direction data is integrated, If the value is negative, 360° is automatically added to ensure that the fused wind direction data is within the standard defined range of (0°, 360°) to completely avoid calculation errors when crossing the ring angle.

[0068] Furthermore, in step S7, the fused wind direction data is sent to the main control system of the wind turbine. The main control system calculates the yaw error between the fused wind direction data and the nacelle orientation in real time. When the yaw error exceeds the preset yaw start threshold, a corresponding yaw control command is generated and the yaw system is driven to perform wind-fighting actions.

[0069] Furthermore, it also includes step S9, real-time monitoring of the drift rate of each wind vane continuously updated in step S4, and triggering a wind vane slow drift fault warning when the drift rate of a single wind vane exceeds the preset health warning threshold and the duration reaches the preset diagnostic confirmation window.

[0070] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0071] 1. Achieves fully autonomous drift self-correction closed loop without requiring an external truth benchmark.

[0072] This invention employs a "group consensus building - credibility gating update - closed-loop iteration" mechanism. Under conditions where there is no GPS or absolute wind direction reference, the system can automatically identify and correct the slow drift of each wind vane by relying solely on mutual verification between wind vanes. This is a fundamental advancement compared to traditional calibration methods that rely on external benchmarks.

[0073] 2. Prevent group drift and dispersion at the source

[0074] This invention ensures that the benchmark will not be "pulled off course" by a deviating wind vane through a consensus-based group approach, ensures that erroneous information from abnormal wind vanes will not contaminate the drift model through a credibility gating mechanism, sets a final "safety valve" for extreme cases through divergent freezing, and solves the most challenging group offset problem in the long-term operation of multi-wind vane redundant systems through three layers of progressive protection.

[0075] 3. Extremely high anti-interference and adaptive capabilities

[0076] This invention can distinguish between the instability (statistical anomaly) of the wind vane itself and structural deviation (structural anomaly) through a two-layer rejection mechanism, and its dynamic threshold can be adjusted according to wind conditions. Through a hysteresis recovery mechanism, the system can make reasonable use of the wind vane that has recovered to normal, rather than permanently discarding it.

[0077] 4. Possesses predictive maintenance value

[0078] This invention provides a direct and continuous monitoring method for slow drift rate from the perspective of multi-wind instrument fusion for the first time. Maintenance personnel can receive early warnings before the wind instruments completely fail, turning unplanned downtime into planned maintenance, significantly improving unit availability and reducing maintenance costs.

[0079] 5. Complete solution for ring data processing

[0080] This invention considers the cyclic characteristics of wind direction angles throughout the entire process, from calculating the circumferential angle deviation to the final vector fusion to resist circumferential errors. It avoids jump errors when crossing 0° / 360° and achieves high accuracy.

[0081] 6. Wide applicability and scalability

[0082] The core architecture of this method does not depend on the specific characteristics of the measured physical quantity. It is also applicable to any multi-redundant measurement scenario with slow-varying drift characteristics, such as anemometers, temperature sensors, and vibration sensors, and has cross-domain promotion value. Attached Figure Description

[0083] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0084] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0085] Example 1:

[0086] like Figure 1 As shown, this embodiment provides a multi-wind vane self-calibration method based on group consensus and credibility gating, including the following steps:

[0087] S1. Multi-source data acquisition and sliding window construction

[0088] At least two wind vanes, which can be mechanical or ultrasonic, are installed on the top of the wind turbine nacelle. The main control PLC of the wind turbine synchronously collects the raw wind direction angle data of each wind vane at a fixed frequency of not less than 1Hz. The collected raw wind direction and angle data are stored in the PLC data buffer, and a sliding time window with a duration of T (preferably 30 seconds) is constructed. The window is updated every second as time progresses, ensuring that the data pool always reflects the latest wind condition information.

[0089] S2, Data Quality Dual-Dimensional Modeling

[0090] Based on the original wind direction angle data within the current sliding time window, two basic quality factors are calculated for each wind vane: a consistency index and a stability weight, which serve as the basis for all subsequent judgments. The consistency index measures the degree of data fit between the i-th wind vane and all other wind vanes, with a value range of (0,1], and its calculation formula is as follows:

[0091] ;

[0092] in, As a consistency indicator; | | represents the absolute difference in the original wind direction angle between the i-th and j-th wind vanes; The number of wind vanes; The closer it is to 1, the more consistent the weather vane is with the other members of the group;

[0093] The stability weight is used to measure the fluctuation of the data from the i-th wind vane, and its calculation formula is as follows:

[0094] ;

[0095] in, For stability weights; This represents the standard deviation of all raw wind direction angle data for this wind vane within the current sliding time window. These are dimensionless coefficients used to convert the standard deviation into a dimensionless coefficient. Mapped to dimensionless weights, The recommended values ​​are based on the typical fluctuation range of the anemometer under normal wind conditions. =0.5~2, preferred =1. The larger the value, the more sensitive the stability weight is to fluctuations. The smaller the value, the smaller the data fluctuation. The closer it is to 1.

[0096] S3, Construction of Group Consensus Center

[0097] Based on the consistency index and stability weight obtained in step S2, the comprehensive credibility weight of a single wind vane is calculated. Optimization is then performed based on this comprehensive credibility weight, which can be achieved through a golden section search or by traversing a finite grid (e.g., a 0.1° step size). This yields the group consensus benchmark that minimizes the total weighted angle dispersion of the wind vanes in the current period. The formula for calculating this group consensus benchmark is as follows:

[0098] ;

[0099] in, As a benchmark of group consensus, ; Let be the overall reliability weight of the i-th wind vane. , As a consistency indicator, For stability weights; It is a circular angular distance function. , The value range is [0°, 180°], which conforms to the cyclic characteristics of wind direction angle.

[0100] S4, Reliability-Gated Drift Self-Correction

[0101] The annular angle deviation of each wind vane is calculated based on the consensus benchmark obtained in step S3 and the original wind direction angle data; the credibility gating coefficient is calculated based on the consistency index and stability weight obtained in step S2; the drift estimate and drift rate of each wind vane are iteratively updated based on the annular angle deviation and the credibility gating coefficient, and the updated drift estimate is used to compensate for the original wind direction angle data to obtain the corrected wind direction data; including the following steps:

[0102] S4.1 Calculate the arithmetic mean of the original wind direction angle data within the current sliding time window of each wind vane. ;

[0103] Since the wind direction angle is a circular data point, the arithmetic mean is calculated using the circular difference formula. With group consensus benchmark The annular angle deviation is adjusted to avoid calculation errors when crossing 0° / 360°, as shown in the following formula:

[0104] ;

[0105] in, For the circumferential angle deviation, this formula normalizes the deviation to the range of (−180°, +180°) to ensure that the direction and magnitude of the deviation are correct;

[0106] S4.2 Calculate the credibility gating coefficient based on the consistency index and stability weight. This credibility gating coefficient is a monotonically increasing nonlinear function of the consistency index and stability weight, and is used to determine the annular angle deviation of this round. The extent to which it is adopted for updating drift values ​​is shown in the following formula:

[0107] ;

[0108] in, This is the credibility gating coefficient; As a consistency indicator; For stability weights; and For sensitivity adjustment index, and >0, >0, preferred

[0109] S4.3, Using credibility gating coefficients For ring angle deviation After weighting, the drift estimates and drift rates of each anemometer are updated as shown in the following formula:

[0110] ;

[0111] ;

[0112] in, The update period for the sliding time window is 1 second in this embodiment; This is the updated drift estimate; Drift rate; The drift estimate before the update;

[0113] S4.4. The original wind direction angle data is compensated using the updated drift estimate to obtain the corrected wind direction data, as shown in the following formula:

[0114] ;

[0115] in, This is the corrected wind direction data; This is the original wind direction angle data.

[0116] S5, Divergence Suppression and Global Freeze Mechanism

[0117] Configure system-level security monitoring. When a risk of data divergence is detected, pause all drift status updates and set up a "safety valve" for the worst-case scenario, as follows:

[0118] Determine whether the corrected wind direction data and the current number of valid anemometers meet the preset freeze trigger conditions; if not, the global freeze mechanism is not triggered, and the current drift parameter update state is maintained; if it meets the conditions, the global freeze mechanism is triggered, the iterative update of drift estimate and drift rate is stopped, and the drift estimate and drift rate are restricted. When the trigger conditions are lifted, the gradual recovery phase begins.

[0119] The preset freeze trigger conditions include: the variance of the corrected wind direction data. Exceeding the preset divergence threshold In this embodiment =200 degrees², or the number of valid wind vanes that have not been eliminated is less than the preset safety number Nmin. In this embodiment, Nmin=2, that is, the minimum redundancy is retained.

[0120] When the variance of the corrected wind direction data Not exceeding the preset divergence threshold If the number of valid wind vanes that have not been removed is not less than the preset safe number Nmin, the preset freeze trigger condition is not met. In this case, the global freeze mechanism is not triggered and the current drift parameter update state is maintained.

[0121] When the variance of the corrected wind direction data Exceeding the preset divergence threshold If the number of valid wind vanes that have not yet been removed is lower than the preset safety number Nmin, the preset freeze trigger condition is met. At this time, the global freeze mechanism is triggered, immediately stopping the iterative update of the drift estimate and drift rate, keeping the drift parameters of the previous cycle unchanged, and forcibly freezing the absolute value of the drift estimate. |Limited to the maximum allowable drift value Within this embodiment =20°, and force the absolute value of the drift rate | |Limited to the maximum allowable drift rate Within this embodiment =0.5° / s;

[0122] After the frozen state is maintained for a preset time, the variance of the corrected wind direction data Preset divergence threshold If the number of valid wind vanes that have not yet been removed is greater than or equal to the preset safety number Nmin, the system will not immediately resume a full update, but will instead enter a gradual recovery phase, specifically:

[0123] In the current cycle, only the group consensus benchmark calculation in step S3 is performed. If the deviation between the new group consensus benchmark and the group consensus benchmark maintained during the freeze period does not exceed the preset deviation value (in this embodiment, the preset deviation value is 5°), then the freeze is completely lifted and normal drift parameter updates are restored. Otherwise, the freeze state is maintained until the next cycle for re-evaluation.

[0124] S6, Two-layer anomaly removal mechanism

[0125] The first layer of statistical anomaly detection for wind vanes is performed based on consistency indices and stability weights, while the second layer of structural anomaly detection is performed based on a consensus benchmark and corrected wind direction data. If either detection result is anomaly, the corresponding wind vane is marked as an anomalous wind vane and removed from the data fusion of the current period. Specifically:

[0126] Based on consistency indicators and stability weights, a first-level statistical anomaly detection is performed on the anemometer to determine whether the anemometer has inherent quality problems, as shown in the following formula:

[0127] ≤ (Stability weight too low, anemometer data fluctuation abnormal) or ≤ (The consistency index is too low, and the wind vane deviates significantly from the population.)

[0128] in, As a consistency indicator; For stability weights; and These are dynamic elimination thresholds for stability weights and consistency indicators, respectively, which can be automatically adjusted according to the current wind conditions and turbulence intensity; under low-turbulence wind conditions, the threshold settings are more stringent (e.g., ...). =0.3, =0.4); Under high turbulent wind conditions, the threshold should be appropriately relaxed to avoid false rejection caused by normal wind fluctuations.

[0129] Based on the consensus benchmark and the corrected wind direction data, a second-level structural anomaly assessment is performed on the wind vane to determine whether the wind vane still significantly deviates from the consensus even after drift compensation, as shown in the following formula:

[0130] | - |> ;

[0131] in, This is the corrected wind direction data; As a benchmark for group consensus; As a structural anomaly threshold, this embodiment sets =15°;

[0132] When performing the second-layer structure anomaly judgment, if the removal operation would cause the number of valid wind vanes to fall below the preset safe number Nmin, the removal is abandoned and the global freeze mechanism in step S5 is triggered to retain the data of the minimum number of wind vanes to participate in the fusion. At the same time, an "insufficient redundancy" alarm is reported. This ensures that the system still has a minimum amount of data available in the worst case.

[0133] S7, Vector fusion output with anti-circular error

[0134] Based on the corrected wind direction data corresponding to the valid anemometers selected in step S6, and combined with the comprehensive reliability weight from step S3, the data is weighted and fused to obtain the fused wind direction data; specifically as follows:

[0135] The corrected wind direction data corresponding to the valid anemometers selected in step S6 are converted into unit vectors in Cartesian coordinates and then weighted averaged to obtain the weighted average unit vector, as shown in the following formula:

[0136] ;

[0137] ;

[0138] in, and These are the x-axis and y-axis components of the weighted average unit vector, respectively. Let be the overall reliability weight of the i-th wind vane; This is the corrected wind direction data;

[0139] The weighted average unit vector is then converted back to the standard wind direction angle to obtain the fused wind direction data within the standard definition range, as shown in the following formula:

[0140] ;

[0141] in, If wind direction data is integrated, If the value is negative, 360° is automatically added to ensure that the fused wind direction data is within the standard defined range of (0°, 360°), thus completely avoiding calculation errors when crossing annular angles. =0, output the previous cycle. .

[0142] S8, Closed-Loop Iteration and Yaw Control

[0143] The fused wind direction data obtained in step S7 is stored in the PLC data buffer as the initial search reference value (hot start) for solving the consensus benchmark in the next cycle step S3, and the updated drift estimate and drift rate are stored to achieve continuous closed-loop iteration across cycles. The fused wind direction data is sent to the main control system of the wind turbine. The main control system calculates the yaw error between the fused wind direction data and the nacelle orientation in real time. When the yaw error exceeds the preset yaw start threshold, this embodiment sets the yaw start threshold to 8°. At this time, the corresponding yaw control command is generated and the yaw system is driven to perform wind action.

[0144] In a specific example, a hysteresis recovery mechanism is also included:

[0145] For the abnormal wind vanes removed in step S6, repeat the calculation in step S2; when the consistency index and stability weight of the abnormal wind vanes reach the preset recovery threshold, and the duration of good condition reaches the preset duration, as shown in the following formula:

[0146] > ;;

[0147] > ;

[0148] Duration of good condition ≥ preset duration ;

[0149] Stability weights are preset to a recovery threshold. The consistency index has a preset recovery threshold, which is set in this embodiment. =0.5~0.6, =0.5~0.6, =10 minutes; the above preset recovery thresholds should be set slightly higher than the removal threshold ( > , > This forms a hysteresis loop, preventing the wind vane from frequently switching between rejection and recovery states at critical points. This is a classic practice in industrial control, ensuring stable system operation.

[0150] The abnormal wind vane was then remarked as a valid wind vane and reintegrated into the fusion calculation.

[0151] In a specific example, step S9, slow drift diagnosis and predictive maintenance triggering, is also included:

[0152] The drift rate of each wind vane is continuously updated in real-time monitoring step S4. When the drift rate of a single wind vane exceeds the preset health warning threshold and the duration reaches the preset diagnostic confirmation window, a wind vane slow drift fault warning is triggered, as shown in the following formula:

[0153]

[0154] in, To preset the health warning threshold, this embodiment sets... =0.05° / s~0.1° / s, This embodiment sets up a preset diagnostic confirmation window. =30 minutes.

[0155] In this embodiment, taking the deployment of three mechanical wind vanes (numbered 1, 2, and 3 respectively) as an example, the specific working process of the above method is as follows:

[0156] Initial power-on (seconds 0-30):

[0157] After the system powers on and starts up, the first 30-second cycle is used to fill the sliding time window. During this period:

[0158] Step S1 continuously collects raw wind direction and angle data, with the window data volume gradually increasing from 0 to 30 records; Step S2 only calculates the consistency index. and stability weights However, no output is output; steps S3 to S4 do not perform drift parameter updates; when the window data reaches 30 records (i.e., 30 seconds), the system officially enters the initialization phase.

[0159] Initialization phase (the first complete window period, i.e., the interval from 30 to 60 seconds):

[0160] Step S3 uses the current window data to solve for the group consensus benchmark through the golden section search. The search range is [0°, 360°), and the convergence accuracy is set to 0.1°. Step S4 receives the group consensus benchmark output in step S3. Subsequently, since there were no historical drift estimates (initial value was 0), the gating update was executed normally, and the annular angle deviation of each wind vane was calculated. And multiplied by the confidence gating coefficient This value is then directly used as the initial drift estimate.

[0161] Normal operation phase:

[0162] The data from the three wind vanes were of good quality and consistent with the data. All are above 0.85, stability weight All are above 0.80; estimated drift values ​​for each anemometer Fluctuating slightly around 0°, drift rate ≈0; Integrating wind direction data The yaw system is operating normally, remaining stable near the actual wind direction.

[0163] Slow drift initiation phase:

[0164] The bearing of wind vane No. 2 began to experience slight jamming due to dust accumulation, and its window mean θ̄2 began to slowly deviate from the consensus center at a rate of approximately 0.05 / s. The annular angle deviation Δ2 of wind vane No. 2 gradually increased from ±2° to approximately ±20°. At the same time, the consistency index C2 of wind vane No. 2 gradually decreased from 0.85 to approximately 0.4 due to the increased difference from other wind vanes, while its stability weight S2 may temporarily remain above 0.75 due to the jamming (because the jamming caused the change in reading value to be smaller). The reliability gating coefficient G2=C2¹×S2²≈0.4×0.56≈0.224 of wind vane No. 2 dropped to less than 1 / 3 of the normal level. The effective update amount G2·Δ2 of wind vane No. 2 was significantly compressed, and the growth rate of its drift estimate b2 was much lower than the actual drift rate—this is an intentional "conservative update" by the gating mechanism to avoid over-self-correction caused by wind vane drift.

[0165] Through slow drift diagnosis and predictive maintenance triggering continuous monitoring, the drift rate r2 of the No. 2 wind vane was stabilized at approximately 0.05~0.08° / s, and cumulative timing began;

[0166] Drift acceleration and automatic isolation phase:

[0167] The No. 2 wind vane malfunctioned, with the annular angle deviation Δ2 further increasing to over 30°; the consistency index C2 dropped below 0.2, and the stability weight S2 also began to fluctuate significantly due to intermittent malfunctions, dropping below 0.3; the reliability gating coefficient G2 = 0.2 × 0.09 ≈ 0.018, only about 3% of the normal level, and drift updates were almost completely frozen; at this point, the second-layer structural anomaly detection was initiated: °, Wind vane #2 was officially removed from the system, and its data will not be included in subsequent data fusion; The accumulated time through slow drift diagnosis and predictive maintenance exceeded 30 minutes, triggering a "Wind vane #2 slow drift fault" warning, which was communicated to maintenance personnel via the SCADA system; Wind vanes #1 and #3 continue to provide reliable data, and the wind direction data will be fused. Accuracy was not significantly affected; the system automatically limited the weight limit of the No. 2 wind vane to 0.3 to prevent it from repeatedly entering / exiting the fusion process when in a critical state.

[0168] Divergence suppression test under extreme wind conditions:

[0169] During an extreme turbulence event, the instantaneous turbulence intensity exceeded 20%, and the data dispersion of the three anemometers increased sharply within a 30-second window. Step S5 detected that the variance of the corrected wind direction data > 200 degrees², and immediately performed the following actions: paused all drift parameter updates, maintaining the drift parameters from the previous cycle; enforced a check on the amplitude constraints of the drift estimate and drift rate; and continued to use the drift estimate from the previous cycle for data correction, fusing the wind direction data. The fluctuation range has been effectively controlled;

[0170] Approximately 2 minutes later, the turbulence intensity returned to normal levels, and the variance of the corrected wind direction data decreased to below 150 degrees²; the system then performed a slow-release recovery: calculating the current consensus baseline. The deviation from the consensus benchmark maintained during the freeze was only 3° (within the safe range of 5°), confirming that it can be safely restored; after the freeze was lifted, the drift update was restored normally, and the system transitioned smoothly.

[0171] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope disclosed in the present invention, based on the technical solution and inventive concept of the present invention, shall fall within the scope of protection of the present invention.

Claims

1. A self-calibration method for multi-wind vanes based on group consensus and credibility gating, characterized in that: Includes the following steps, S1. Deploy no fewer than two wind vanes on the top of the wind turbine nacelle, synchronously collect the raw wind direction angle data of each wind vane, construct a sliding time window based on the raw wind direction angle data and update it in real time, and output the raw wind direction angle data within the current sliding time window. S2. Based on the original wind direction angle data within the current sliding time window, calculate the consistency index and stability weight of each wind vane. S3. Based on the consistency index and stability weight obtained in step S2, calculate the comprehensive credibility weight of a single wind vane, and optimize based on the comprehensive credibility weight to obtain the group consensus benchmark that minimizes the total weighted angle dispersion of the wind vane in the current period. S4. Calculate the annular angle deviation of each wind vane based on the group consensus benchmark and the original wind direction angle data obtained in step S3; calculate the credibility gating coefficient based on the consistency index and stability weight obtained in step S2; iteratively update the drift estimate and drift rate of each wind vane based on the annular angle deviation and the credibility gating coefficient, and use the updated drift estimate to compensate for the original wind direction angle data to obtain the corrected wind direction data. S5. Determine whether the corrected wind direction data and the current number of valid anemometers meet the preset freeze trigger conditions. If the conditions are not met, the global freeze mechanism will not be triggered, and the current drift parameter update state will be maintained; if the conditions are met, the global freeze mechanism will be triggered, the iterative update of the drift estimate and drift rate will be stopped, and the drift estimate and drift rate will be restricted. When the triggering conditions are lifted, the slow recovery phase will begin. S6. Based on the consistency index and stability weight, perform the first-level statistical anomaly judgment on the wind vane, and based on the group consensus benchmark and the corrected wind direction data, perform the second-level structural anomaly judgment on the wind vane; if any judgment result is abnormal, mark the corresponding wind vane as an abnormal wind vane and remove it from the data fusion of the current period. S7. Based on the corrected wind direction data corresponding to the valid wind vanes selected in step S6, and combined with the comprehensive credibility weight in step S3, the data is weighted and fused to obtain the fused wind direction data. S8. The fused wind direction data obtained in step S7 is used as the initial search reference value for solving the consensus benchmark in step S3 of the next cycle, and the updated drift estimate and drift rate are stored to achieve continuous closed-loop iteration across cycles. The main control system of the wind turbine generates corresponding yaw control commands based on the yaw error between the fused wind direction data and the nacelle orientation, and drives the yaw system of the wind turbine to perform wind-fighting actions.

2. The multi-wind vane self-calibration method based on group consensus and credibility gating according to claim 1, characterized in that: For the abnormal wind vanes removed in step S6, repeat the calculation in step S2; when the consistency index and stability weight of the abnormal wind vane reach the preset recovery threshold and the duration reaches the preset duration, re-mark the abnormal wind vane as a valid wind vane and re-include it in the fusion calculation.

3. The multi-wind vane self-calibration method based on group consensus and credibility gating according to claim 1, characterized in that: In step S2, the formula for calculating the consistency index is: ； in, As a consistency indicator; | | represents the absolute difference in the original wind direction angle between the i-th and j-th wind vanes; The number of wind vanes; The formula for calculating the stability weight is: ； in, For stability weights; This represents the standard deviation of all raw wind direction angle data of the anemometer within the current sliding time window. These are dimensionless coefficients used to convert the standard deviation into a dimensionless coefficient. The mapping is to dimensionless weights.

4. The multi-wind vane self-calibration method based on group consensus and credibility gating according to claim 1, characterized in that: In step S3, the formula for calculating the group consensus benchmark is: ； in, As a benchmark for group consensus; Let be the overall reliability weight of the i-th wind vane. , As a consistency indicator, For stability weights; It is a circular angular distance function. , The value range is [0°, 180°].

5. The multi-wind vane self-calibration method based on group consensus and credibility gating according to claim 1, characterized in that: Step S4 includes the following steps: S4.1 Calculate the arithmetic mean of the original wind direction angle data within the current sliding time window of each wind vane. ; The arithmetic mean is calculated using the circular difference formula. With group consensus benchmark The annular angular deviation is shown in the following formula: ； in, The circumferential angle deviation has a range of values ​​(−180°, +180°). S4.2 Calculate the credibility gating coefficient based on the consistency index and stability weight, as shown in the following formula: ； in, This is the credibility gating coefficient; As a consistency indicator; For stability weights; and The sensitivity adjustment index, and >0, >0; S4.3, Using credibility gating coefficients For ring angle deviation After weighting, the drift estimates and drift rates of each anemometer are updated as shown in the following formula: ； ； in, The update cycle for the sliding time window; This is the updated drift estimate; For drift rate; The drift estimate before the update; S4.

4. The original wind direction angle data is compensated using the updated drift estimate to obtain the corrected wind direction data, as shown in the following formula: ； in, This is the corrected wind direction data; This is the original wind direction angle data.

6. The multi-wind instrument self-calibration method based on group consensus and credibility gating according to claim 1, characterized in that: In step S5, the preset freeze trigger conditions include: the variance of the corrected wind direction data. Exceeding the preset divergence threshold Or the number of valid wind vanes that have not been removed is less than the preset safety number Nmin; When the variance of the corrected wind direction data Not exceeding the preset divergence threshold If the number of valid wind vanes that have not been removed is not less than the preset safe number Nmin, the preset freeze trigger condition is not met. In this case, the global freeze mechanism is not triggered and the current drift parameter update state is maintained. When the variance of the corrected wind direction data Exceeding the preset divergence threshold If the number of valid wind vanes that have not yet been removed is lower than the preset safety number Nmin, the preset freeze trigger condition is met. At this time, the global freeze mechanism is triggered, immediately stopping the iterative update of the drift estimate and drift rate, keeping the drift parameters of the previous cycle unchanged, and forcibly freezing the absolute value of the drift estimate. |Limited to the maximum allowable drift value Within, and force the absolute value of the drift rate | |Limited to the maximum allowable drift rate within; After the frozen state is maintained for a preset time, the variance of the corrected wind direction data Preset divergence threshold If the number of valid anemometers that have not yet been removed is greater than or equal to the preset safety number Nmin, the system enters the slow-release recovery phase, specifically as follows: In the current cycle, only the group consensus benchmark calculation in step S3 is performed. If the deviation between the new group consensus benchmark and the group consensus benchmark maintained during the freeze period does not exceed the preset deviation value, the freeze is completely lifted and normal drift parameter updates are resumed. Otherwise, the freeze status is maintained until the next cycle for re-evaluation.

7. The multi-wind instrument self-calibration method based on group consensus and credibility gating according to claim 1, characterized in that: In step S6, the first level of statistical anomaly judgment is performed on the anemometer based on the consistency index and stability weight to determine whether the anemometer has any quality problems, as shown in the following formula: ≤ or ≤ ; in, As a consistency indicator; For stability weights; and These are the dynamic removal thresholds for stability weight and consistency index, respectively; Based on the consensus benchmark and the corrected wind direction data, a second-level structural anomaly assessment is performed on the wind vane to determine whether the wind vane still significantly deviates from the consensus even after drift compensation, as shown in the following formula: | − |> ； in, This is the corrected wind direction data; As a benchmark for group consensus; This is the threshold for structural anomalies; When performing the second-layer structure anomaly judgment, if the removal operation will cause the number of valid wind vanes to fall below the preset safe number Nmin, then the removal will be abandoned and the global freeze mechanism in step S5 will be triggered to retain the data of the minimum number of wind vanes to participate in the fusion, and at the same time, an "insufficient redundancy" alarm will be reported.

8. The multi-wind vane self-calibration method based on group consensus and credibility gating according to claim 1, characterized in that: In step S7, the corrected wind direction data corresponding to the valid anemometers selected in step S6 are converted into unit vectors in the Cartesian coordinate system and then weighted and averaged to obtain the weighted average unit vector, as shown in the following formula: ； ； in, and These are the x-axis and y-axis components of the weighted average unit vector, respectively. Let be the overall reliability weight of the i-th wind vane; This is the corrected wind direction data; The weighted average unit vector is then converted back to the standard wind direction angle to obtain the fused wind direction data within the standard definition range, as shown in the following formula: ； in, If wind direction data is integrated, If the value is negative, 360° is automatically added to ensure that the fused wind direction data is within the standard defined range of (0°, 360°) to completely avoid calculation errors when crossing the ring angle.

9. The multi-wind vane self-calibration method based on group consensus and credibility gating according to claim 1, characterized in that: In step S7, the fused wind direction data is sent to the main control system of the wind turbine. The main control system calculates the yaw error between the fused wind direction data and the nacelle orientation in real time. When the yaw error exceeds the preset yaw start threshold, the corresponding yaw control command is generated and the yaw system is driven to perform wind-fighting actions.

10. The multi-wind vane self-calibration method based on group consensus and credibility gating according to claim 1, characterized in that: It also includes step S9, real-time monitoring of the drift rate of each wind vane continuously updated in step S4, and triggering a wind vane slow drift fault warning when the drift rate of a single wind vane exceeds the preset health warning threshold and the duration reaches the preset diagnostic confirmation window.

Images