A method, device, equipment and medium for wind turbine group clearance protection control based on tip deflection prediction

Abstract

The application discloses a wind turbine group clearance protection control method and device based on tip deflection prediction, equipment and medium, relates to the wind turbine group control technical field, including: based on the operation time series data of the wind turbine group and the corresponding tip deflection true value determine total relationship matrix and total eigenvalue vector;The target eigenvalue of the to-be-estimated time series data is sorted;Unit root test is carried out on the residual sequence corresponding to the obtained tip deflection estimated time series and real time series, parameter estimation and Ljung-Box white noise test are carried out on the obtained alternative ARIMA model set, the predicted value is determined by using the target ARIMA model, and the forecast value is obtained by combining the tip deflection estimated time series;When the wind wheel azimuth angle meets the preset interval, whether the difference between the forecast value and the preset static clearance value is less than the target clearance protection threshold is judged, so that the clearance protection control operation is carried out.In the case of radar failure, the clearance protection is realized based on tip deflection prediction.

Description

Technical Field

[0001] This invention relates to the field of wind turbine generator control technology, and in particular to a wind turbine generator headroom protection control method, device, equipment and medium based on blade tip deflection prediction. Background Technology

[0002] With the rapid development of wind power generation technology and the increasing demands for cost per kilowatt-hour, turbine capacity and blade length are gradually increasing. Large-capacity wind turbines can be equipped with airspace monitoring equipment to take protective pitch control actions when airspace is insufficient; however, airspace monitoring equipment is prone to failure under complex terrain and weather conditions, resulting in deficiencies in both safety and economy.

[0003] As can be seen from the above, how to achieve airspace protection based on blade tip deflection prediction in the event of radar failure is an urgent problem to be solved. Summary of the Invention

[0004] In view of this, the purpose of this invention is to provide a method, device, equipment, and medium for airspace protection control of wind turbine generators based on blade tip deflection prediction, which can achieve airspace protection based on blade tip deflection prediction in the event of radar failure. The specific solution is as follows: In a first aspect, this application provides a wind turbine generator headroom protection control method based on blade tip deflection prediction, including: Sample data is constructed based on the operating sequence data of wind turbine generators under different operating conditions and the corresponding true values ​​of blade tip deflection. The total relation matrix corresponding to all operating conditions is determined using the sample data. The total feature value vector corresponding to all operating conditions is determined based on the operating sequence data and preset weight coefficients. Determine the target feature value corresponding to the time series data to be estimated, sort the target feature value based on the total feature value vector, and use the sorting result to determine the working condition weight coefficient corresponding to the time series data to be estimated, and use the total relationship matrix and the working condition weight coefficient to determine the blade tip deflection estimation time series; The residual sequence is determined by the difference between the estimated time series of blade tip deflection and the corresponding true time series of blade tip deflection. The unit root test is performed on the residual sequence based on the augmented Dickey-Fowler test method. The target stationary residual sequence is determined by the obtained unit root test results. A set of candidate ARIMA models is constructed based on the autocorrelation function and partial autocorrelation function of the target stationary residual sequence. Parameter estimation is performed on each model in the candidate ARIMA model set, and Ljung-Box white noise test is performed on the model residual sequence of the estimated model. Based on the test results, the target ARIMA model is determined, and the future trend of the residual sequence is predicted using the target ARIMA model to obtain the corresponding predicted value. The predicted value and the blade tip deflection estimation time series are summed to obtain the blade tip deflection prediction value. When the wind turbine azimuth angle meets the preset range, it is determined whether the difference between the predicted blade tip deflection value and the preset static clearance value is less than the target clearance protection threshold, and corresponding clearance protection control operations are performed based on the obtained judgment result.

[0005] Optionally, the step of constructing sample data based on the operating sequence data of the wind turbine generator under different operating conditions and the corresponding true values ​​of blade tip deflection, using the sample data to determine the overall relational matrix corresponding to all operating conditions, and determining the overall eigenvalue vector corresponding to all operating conditions based on the operating sequence data and preset weight coefficients, includes: The system collects operational sequence data of wind turbine generators under different operating conditions. These different operating conditions include varying average wind speeds, turbulence intensities, and wind-related errors. The operational sequence data includes generator parameters, which include rotor speed, rotor azimuth angle, blade pitch angle, blade root flapping torque, blade root oscillation torque, and generator torque. All generator parameters are pairwise linearly uncorrelated. Based on the runtime sequence data under a single operating condition, the Gram-Schmidt orthogonalization method is used to decompose the data to obtain the corresponding relation matrix. The relation matrices under all operating conditions are then merged to obtain the total relation matrix. The mean of the runtime sequence data under a single operating condition is weighted by a preset weight coefficient to obtain the corresponding feature value. The feature values ​​under all operating conditions are then merged to obtain the total feature value vector.

[0006] Optionally, the step of sorting the target feature values ​​based on the total feature value vector, determining the operating condition weight coefficients corresponding to the time series data to be estimated using the obtained sorting results, and determining the blade tip deflection estimation time series using the total relation matrix and the operating condition weight coefficients includes: The target feature values ​​are sorted into the total feature value vector using a sorting algorithm to obtain the location feature value vector. Based on the location feature value vector, the adjacent feature values ​​before and after the target feature value and the target operating conditions are determined. The corresponding working condition weight coefficient is determined based on the adjacent feature values ​​and the target feature value; The blade tip deflection estimation time series is determined based on the overall relationship matrix, the target operating conditions, the operating condition weighting coefficients, and the time series data to be estimated.

[0007] Optionally, the method of performing a unit root test on the residual sequence based on the augmented Dickey-Fowler test, and using the obtained unit root test results to determine the target stationary residual sequence, and constructing a set of candidate ARIMA models based on the autocorrelation function and partial autocorrelation function of the target stationary residual sequence, including: The unit root test was performed on the residual sequence based on the augmented Dickey-Fowler test method to obtain the corresponding test statistic; The target stationary residual sequence is determined using the test statistic, and the difference order is determined based on the target stationary residual sequence. Determine the autocorrelation function and partial autocorrelation function of the target stationary residual sequence, and draw the corresponding correlation plot based on the autocorrelation function and the partial autocorrelation function; Based on the truncation or trailing characteristics of the correlation graph and the target information criterion, the autoregressive order and the moving average order are determined, and a set of candidate ARIMA models is constructed using the difference order, the autoregressive order, and the moving average order; the target information criterion includes the Akaike information criterion and the Bayesian information criterion.

[0008] Optionally, determining the target stationary residual sequence using the test statistic includes: If the test statistic is not greater than the preset critical value, then the residual sequence is determined to be the target stationary residual sequence; If the test statistic is greater than the preset critical value, the residual sequence is determined to be a non-stationary residual sequence, and a difference operation is performed on the non-stationary residual sequence to obtain a new residual sequence. Then, the process jumps to the step of performing a unit root test on the residual sequence based on the augmented Dickey-Fowler test method until the test statistic is not greater than the preset critical value, so as to obtain the target stationary residual sequence.

[0009] Optionally, the step of estimating the parameters of each model in the candidate ARIMA model set, performing Ljung-Box white noise test on the model residual sequence of the estimated model, and determining the target ARIMA model based on the obtained test results includes: The parameters of each model in the set of candidate ARIMA models are estimated using the maximum likelihood estimation method or the conditional least squares method to obtain the autoregressive coefficients and moving average coefficients. An estimated model is constructed based on the autoregressive coefficients and the moving average coefficients. The Ljung-Box white noise test is performed on the model residual sequence of the estimated model to obtain the corresponding test value. If the test value is not greater than the preset significance level, the difference order, the autoregression order, and the moving average order are adjusted to obtain new difference order, autoregression order, and moving average order, and then the process jumps to the step of constructing a set of candidate ARIMA models using the difference order, the autoregression order, and the moving average order, until the tested model is obtained. If the test value is greater than the preset significance level, the test is considered passed, and the estimated model that passed the test is determined as the tested model. The target ARIMA model is determined based on the goodness of fit, parameter significance level, and information criterion value of the tested model.

[0010] Optionally, when the wind turbine azimuth angle meets a preset range, it is determined whether the difference between the predicted blade tip deflection value and the preset static clearance value is less than the target clearance protection threshold, and corresponding clearance protection control operations are performed based on the obtained determination result, including: When the wind turbine azimuth angle meets the preset range, it is determined whether the difference between the predicted blade tip deflection value and the preset static clearance value is not greater than the target clearance protection threshold. If the value is not greater than the target airspace protection threshold, then the original pitch angle in the time series data to be estimated is superimposed with a compensation value, and the protection window timer is started. During the protection window timer, the real-time runtime sequence data of the next sampling cycle is determined as the new time sequence data to be estimated, and then the process jumps to the step of determining the target feature value corresponding to the time sequence data to be estimated, so as to obtain a new blade tip deflection prediction value, until the wind turbine azimuth angle does not meet the preset range, so as to complete the corresponding airspace protection control operation.

[0011] Secondly, this application provides a wind turbine generator headroom protection control device based on blade tip deflection prediction, comprising: The eigenvalue vector determination module is used to construct sample data based on the operating sequence data of the wind turbine generator under different operating conditions and the corresponding true values ​​of the blade tip deflection, use the sample data to determine the total relation matrix corresponding to all operating conditions, and determine the total eigenvalue vector corresponding to all operating conditions based on the operating sequence data and preset weight coefficients. The deflection estimation time series determination module is used to determine the target feature value corresponding to the time series data to be estimated, sort the target feature value based on the total feature value vector, and use the sorting result to determine the working condition weight coefficient corresponding to the time series data to be estimated, and use the total relationship matrix and the working condition weight coefficient to determine the blade tip deflection estimation time series. The model set construction module is used to determine the residual sequence by using the difference between the estimated time series of leaf tip deflection and the corresponding true time series of leaf tip deflection, perform a unit root test on the residual sequence based on the augmented Dickey-Fowler test method, determine the target stationary residual sequence using the obtained unit root test results, and construct a candidate ARIMA model set based on the autocorrelation function and partial autocorrelation function of the target stationary residual sequence. The deflection prediction value determination module is used to estimate the parameters of each model in the candidate ARIMA model set, perform Ljung-Box white noise test on the model residual sequence of the estimated model, determine the target ARIMA model based on the test results, predict the future trend of the residual sequence using the target ARIMA model to obtain the corresponding predicted value, and sum the predicted value and the blade tip deflection estimation time series to obtain the blade tip deflection prediction value. The airspace protection control module is used to determine whether the difference between the predicted blade tip deflection value and the preset static airspace value is less than the target airspace protection threshold when the wind turbine azimuth angle meets the preset range, and to perform corresponding airspace protection control operations based on the obtained judgment result.

[0012] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor is used to execute the computer program to implement the aforementioned wind turbine generator airspace protection control method based on blade tip deflection prediction.

[0013] Fourthly, this application provides a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the aforementioned wind turbine generator headroom protection control method based on blade tip deflection prediction.

[0014] This application constructs sample data based on the operating sequence data of wind turbine generators under different operating conditions and the corresponding true values ​​of blade tip deflection. It uses this sample data to determine the overall relational matrix corresponding to all operating conditions, and determines the overall eigenvalue vector corresponding to all operating conditions based on the operating sequence data and preset weight coefficients. It then determines the target eigenvalues ​​corresponding to the time series data to be estimated, sorts these target eigenvalues ​​based on the overall eigenvalue vector, and uses the sorting results to determine the operating condition weight coefficients corresponding to the time series data to be estimated. Finally, it uses the overall relational matrix and the operating condition weight coefficients to determine the blade tip deflection estimation time series. The application uses the difference between the estimated blade tip deflection time series and the corresponding true blade tip deflection time series to determine the residual sequence, performs a unit root test on the residual sequence based on the augmented Dickey-Fowler test method, and uses the obtained unit root test results... The root position test results determine the target stationary residual sequence. Based on the autocorrelation function and partial autocorrelation function of the target stationary residual sequence, a set of candidate ARIMA models is constructed. Parameter estimation is performed on each model in the candidate ARIMA model set, and the model residual sequence of the estimated model is subjected to Ljung-Box white noise test. Based on the test results, the target ARIMA model is determined. The target ARIMA model is used to predict the future trend of the residual sequence to obtain the corresponding predicted value. The predicted value and the estimated blade tip deflection time series are summed to obtain the predicted blade tip deflection value. When the rotor azimuth angle meets the preset range, it is determined whether the difference between the predicted blade tip deflection value and the preset static clearance value is less than the target clearance protection threshold. Based on the judgment result, corresponding clearance protection control operations are performed.

[0015] As shown above, this application constructs sample data based on the operating sequence data of wind turbine generators under different operating conditions and the corresponding true values ​​of blade tip deflection. Based on the sample data, it determines the overall relational matrix and overall eigenvalue vector corresponding to all operating conditions. By sorting the data, it matches the operating conditions of the time series data to be estimated, and combines this with the overall relational matrix to determine the blade tip deflection estimation time series, capturing the deterministic linear trend of blade tip deflection. For nonlinear components that cannot be fitted, root tests and differencing operations are used to ensure that the input sequence meets the model assumptions. Ljung-Box white noise is used to verify whether the model has sufficiently extracted residual information. Based on the target ARIMA model, the future trend of the residual sequence is predicted to obtain the predicted value, which is then combined with the blade tip deflection estimation time series to obtain the predicted blade tip deflection value. In this way, protection is triggered in advance when the blade enters the dangerous azimuth angle range based on the predicted blade tip deflection value, providing sufficient response time for the pitch system and significantly reducing the risk of tower collision. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0017] Figure 1 This application discloses a flowchart of a wind turbine generator headroom protection control method based on blade tip deflection prediction. Figure 2 This is a schematic diagram illustrating one method for determining weighting coefficients as disclosed in this application; Figure 3 This is a schematic diagram illustrating the determination of a target ARIMA model disclosed in this application; Figure 4 This is a schematic diagram of the airspace protection control device for a wind turbine generator set based on blade tip deflection prediction disclosed in this application. Figure 5 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation

[0018] 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 only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Currently, large-capacity wind turbines can be equipped with air clearance monitoring equipment to take protective pitch control actions when air clearance is insufficient; however, air clearance monitoring equipment is prone to failure under complex terrain and weather conditions, resulting in deficiencies in both safety and economy. Therefore, this application provides a wind turbine air clearance protection control method based on blade tip deflection prediction. Based on the predicted blade tip deflection value, protection is triggered in advance when the blade enters the dangerous azimuth angle range, providing sufficient response time for the pitch system and significantly reducing the risk of tower collision.

[0020] See Figure 1 As shown, this invention discloses a wind turbine generator headroom protection control method based on blade tip deflection prediction, comprising: Step S11: Construct sample data based on the operating sequence data of the wind turbine generator under different operating conditions and the corresponding true values ​​of blade tip deflection. Use the sample data to determine the total relation matrix corresponding to all operating conditions. Based on the operating sequence data and preset weight coefficients, determine the total feature vector corresponding to all operating conditions.

[0021] In this embodiment, the operation sequence data of the wind turbine generator set under different operating conditions are obtained; the different operating conditions include, but are not limited to, differences in average wind speed, turbulence intensity, and wind error; the operation sequence data can come from dynamic simulation software or actual operating unit parameters; the unit parameters include rotor speed, rotor azimuth angle, blade pitch angle, blade root flapping torque, blade root oscillation torque, and generator torque. Figure 2 This is a schematic diagram illustrating the determination of weighting coefficients. The runtime sequence data includes an input vector and an output vector; the formula corresponding to the input vector is as follows: ; Where i is the number of input vectors. The output vector is , where is the blade deflection.

[0022] Understandably, this is an assumption. Various operating conditions, in a single operating condition Assuming the data length is L, then the input matrix... for: ; in, Let be the value of the i-th input vector at the k-th time position. Since the input vectors are pairwise linearly uncorrelated, the input matrix... It is a full-rank matrix. Output matrix. for: ; in, This is to output the value of the variable deflection at the k-th time position.

[0023] Furthermore, for a single operating condition and full-rank matrix It can be decomposed into an orthogonal matrix and an upper triangular matrix with positive diagonal elements using the Gram-Schmidt orthogonalization method. The product of, i.e. Based on the linear regression equation Solve the relation matrix under the operating conditions. The corresponding formula is as follows: ; in, The relation matrix is ​​described above; The upper triangular matrix; Let be the output matrix. For all Under various operating conditions, a total relation matrix containing N relation matrices is obtained. The corresponding formula is as follows: ; The feature values ​​are constructed from the input vector. The feature values ​​under a single operating condition are shown below: ; in, The feature value; The timing of the i-th input vector under a single operating condition The mean of the following; These are the weight coefficients corresponding to the mean of each input vector. Accordingly, for all... Under various operating conditions, a total eigenvalue vector containing N eigenvalues ​​is obtained. As shown below: ; By adjusting the weight coefficients, the total eigenvalue vector can be optimized. It achieves monotonically increasing.

[0024] Specifically, the step of constructing sample data based on the operating sequence data of the wind turbine generator under different operating conditions and the corresponding true values ​​of blade tip deflection, using the sample data to determine the overall relationship matrix corresponding to all operating conditions, and determining the overall feature vector corresponding to all operating conditions based on the operating sequence data and preset weight coefficients, includes: collecting operating sequence data of the wind turbine generator under different operating conditions; the different operating conditions include different average wind speeds, turbulence intensities, and windward errors; the operating sequence data includes generator parameters; the generator parameters include rotor speed. The parameters include the rotor azimuth angle, blade pitch angle, blade root flapping moment, blade root oscillation moment, and generator torque; all parameters are pairwise linearly uncorrelated; based on the operating sequence data under a single operating condition, the Gram-Schmidt orthogonalization method is used to decompose the data to obtain the corresponding relation matrix; the relation matrices under all operating conditions are merged to obtain the total relation matrix; the mean of the operating sequence data under a single operating condition is weighted by a preset weight coefficient to obtain the corresponding eigenvalues; the eigenvalues ​​under all operating conditions are merged to obtain the total eigenvalue vector.

[0025] Step S12: Determine the target feature value corresponding to the time series data to be estimated, sort the target feature value based on the total feature value vector, and use the sorting result to determine the working condition weight coefficient corresponding to the time series data to be estimated, and use the total relationship matrix and the working condition weight coefficient to determine the blade tip deflection estimation time series.

[0026] In this embodiment, for the time series data to be estimated The corresponding target feature values ​​can be determined using the method described above. The target feature values ​​are sorted into the total feature value vector using a sorting algorithm. In the process, a localization feature vector of length N+1 is obtained, as shown below: ; in, The location feature value vector; Let be the target feature value. It can be obtained that... Record the index q of the adjacent feature values ​​before and after the target feature value, that is, the target operating condition corresponding to the target feature value.

[0027] It is understood that the corresponding working condition weight coefficient is determined based on the adjacent feature values ​​and the target feature value, and the corresponding formula is as follows: ; ; in, , This refers to the weighting coefficient for operating conditions. , These are adjacent feature values; The target feature value is defined as follows. After obtaining the operating condition weighting coefficients, the timing sequence for tip deflection estimation is further determined, and the corresponding formula is as follows: ; in, For the timing of the blade tip deflection estimation; , This refers to the weighting coefficient for operating conditions. , These are adjacent feature values; The time series data to be estimated is referred to here.

[0028] Specifically, the step of sorting the target feature values ​​based on the total feature value vector and determining the operating condition weight coefficients corresponding to the time series data to be estimated using the obtained sorting results, and determining the blade tip deflection estimation time series using the total relation matrix and the operating condition weight coefficients, includes: sorting the target feature values ​​into the total feature value vector using a sorting algorithm to obtain a positioning feature value vector, and determining the adjacent feature values ​​and target operating conditions corresponding to the target feature values ​​based on the positioning feature value vector; determining the corresponding operating condition weight coefficients based on the adjacent feature values ​​and the target feature value; and determining the blade tip deflection estimation time series based on the total relation matrix, the target operating conditions, the operating condition weight coefficients, and the time series data to be estimated.

[0029] Step S13: Determine the residual sequence using the difference between the estimated blade tip deflection time series and the corresponding true blade tip deflection time series; perform a unit root test on the residual sequence based on the augmented Dickey-Fowler test method; determine the target stationary residual sequence using the obtained unit root test results; and construct a set of candidate ARIMA models based on the autocorrelation function and partial autocorrelation function of the target stationary residual sequence.

[0030] In this embodiment, Figure 3 A schematic diagram is provided for determining a target ARIMA model. The time series of the estimated blade tip deflection is subtracted from the actual time series to obtain the residual time series, and the corresponding formula is as follows: ; in, The residual timing sequence; This represents the actual timing sequence; The time series for estimating the leaf tip deflection is then performed. A unit root test is then conducted on the residual sequence, which can be achieved using the augmented Dickey-Fowler test or other methods. If the obtained test statistic is greater than the critical value, the residual sequence is determined to be non-stationary. The residual sequence is then differencing until it passes the unit root test to obtain the target stationary residual sequence. , where d is the optimal difference order that makes the residual sequence stationary.

[0031] Understandably, the autocorrelation function and partial autocorrelation function of the target stationary residual sequence are determined, and the corresponding autocorrelation analysis plots and partial autocorrelation analysis plots are plotted. Based on the truncation or tailing characteristics of the above analysis plots, and the Akaike information criterion or Bayesian information criterion, the autoregression order p and the moving average order q are initially determined, forming a set of candidate ARIMA models. Specifically, the step of performing a unit root test on the residual sequence based on the augmented Dickey-Fowler test method, and using the obtained unit root test results to determine the target stationary residual sequence, and constructing a set of candidate ARIMA models based on the autocorrelation function and partial autocorrelation function of the target stationary residual sequence, includes: performing a unit root test on the residual sequence based on the augmented Dickey-Fowler test method to obtain the corresponding test statistic; using the test statistic to determine the target stationary residual sequence, and determining the difference order based on the target stationary residual sequence; determining the autocorrelation function and partial autocorrelation function of the target stationary residual sequence, and drawing the corresponding correlation plot based on the autocorrelation function and the partial autocorrelation function; determining the autoregression order and moving average order based on the truncation or tailing characteristics of the correlation plot and the target information criterion, and constructing a set of candidate ARIMA models using the difference order, the autoregression order, and the moving average order; the target information criterion includes the Akaike information criterion and the Bayesian information criterion.

[0032] Specifically, determining the target stationary residual sequence using the test statistic includes: if the test statistic is not greater than a preset critical value, the residual sequence is determined to be the target stationary residual sequence; if the test statistic is greater than the preset critical value, the residual sequence is determined to be a non-stationary residual sequence, and a differencing operation is performed on the non-stationary residual sequence to obtain a new residual sequence. Then, the process jumps to the step of performing a unit root test on the residual sequence based on the augmented Dickey-Fowler test method, until the test statistic is not greater than the preset critical value, thus obtaining the target stationary residual sequence. It is worth noting that the preset critical value can be set according to actual conditions.

[0033] Step S14: Estimate the parameters of each model in the candidate ARIMA model set, and perform Ljung-Box white noise test on the model residual sequence of the estimated model. Based on the test results, determine the target ARIMA model, use the target ARIMA model to predict the future trend of the residual sequence to obtain the corresponding predicted value, and sum the predicted value and the blade tip deflection estimation time series to obtain the blade tip deflection forecast value.

[0034] In this embodiment, after obtaining the set of candidate ARIMA models, the parameters of each model in the set are estimated using maximum likelihood estimation or conditional least squares to obtain autoregressive coefficients and moving average coefficients. Based on these autoregressive and moving average coefficients, an estimated model is constructed. The Ljung-Box white noise test is performed on the residual sequence of the estimated model to obtain the corresponding test value. If the test value is greater than a preset significance level, the residual sequence is considered a white noise sequence, indicating that the model has fully extracted the autorelated information from the residuals. The estimated model that passes the test is determined as the tested model. If the test value is not greater than the preset significance level, the model order is readjusted, and the model is refitted until the tested model is obtained. Considering the model's goodness of fit, parameter significance, and information criterion value, the model with the smallest Akaike information criterion or Bayesian information criterion value and that passes the white noise test is selected as the optimal ARIMA model. This optimal ARIMA model is denoted as the autoregressive integral moving average model, and is expressed as: ; in, It is an autoregressive polynomial; It is a lag operator; It is the difference order; It is a moving average polynomial; The sequence is a white noise sequence. The optimal ARIMA model is used to predict the future trend of the residual sequence to obtain the predicted value. , representing the predicted value of the residual at time t+k based on the information at time t. The predicted value and the estimated blade tip deflection time series are summed to obtain the predicted blade tip deflection value, and the corresponding formula is as follows: ; in, This refers to the predicted value of the blade tip deflection; For the timing of the blade tip deflection estimation; The predicted value is denoted as .

[0035] Specifically, the step of estimating the parameters of each model in the candidate ARIMA model set and performing an Ljung-Box white noise test on the model residual sequence of the estimated model, and determining the target ARIMA model based on the test results, includes: estimating the parameters of each model in the candidate ARIMA model set using maximum likelihood estimation or conditional least squares to obtain autoregressive coefficients and moving average coefficients; constructing the estimated model based on the autoregressive coefficients and the moving average coefficients; performing an Ljung-Box white noise test on the model residual sequence of the estimated model to obtain the corresponding test values; if If the test value is not greater than the preset significance level, the difference order, the autoregression order, and the moving average order are adjusted to obtain new difference orders, autoregression orders, and moving average orders. Then, the process jumps to the step of constructing a set of candidate ARIMA models using the difference order, the autoregression order, and the moving average order until the tested model is obtained. If the test value is greater than the preset significance level, the test is considered passed, and the estimated model that passed the test is determined as the tested model. The target ARIMA model is determined based on the goodness of fit, parameter significance level, and information criterion value of the tested model.

[0036] Step S15: When the wind turbine azimuth angle meets the preset range, determine whether the difference between the predicted blade tip deflection value and the preset static clearance value is less than the target clearance protection threshold, and perform corresponding clearance protection control operations based on the obtained judgment result.

[0037] In this embodiment, when the wind turbine azimuth angle is within the danger zone, which can be [160°, 200°] or adjusted according to the actual situation, if the difference between the preset static clearance value and the predicted blade tip deflection value is not greater than the target clearance protection threshold, a fixed compensation value is added to the current pitch angle output value, and a protection window timer is started. During the protection window timer's timing, real-time runtime sequence data is continuously collected as new time sequence data to be estimated, and the process jumps to the step of determining the target feature value corresponding to the time sequence data to be estimated, so as to obtain a new predicted blade tip deflection value. If the wind turbine azimuth angle turns out of the preset danger zone, or the protection window timer ends and the recalculated predicted clearance value is greater than the target clearance protection threshold, the protection mode is exited and the original pitch angle output value is restored.

[0038] Specifically, when the wind turbine azimuth angle meets the preset range, it is determined whether the difference between the predicted blade tip deflection value and the preset static clearance value is less than the target clearance protection threshold, and corresponding clearance protection control operations are performed based on the obtained judgment result. This includes: when the wind turbine azimuth angle meets the preset range, it is determined whether the difference between the predicted blade tip deflection value and the preset static clearance value is not greater than the target clearance protection threshold; if it is not greater than the target clearance protection threshold, a compensation value is added to the original blade pitch angle in the time series data to be estimated, and a protection window timer is started; during the timer period of the protection window timer, the real-time running time series data of the next sampling period is determined as the new time series data to be estimated, and then the process jumps to the step of determining the target feature value corresponding to the time series data to be estimated, so as to obtain a new predicted blade tip deflection value, until the wind turbine azimuth angle does not meet the preset range, so as to complete the corresponding clearance protection control operation.

[0039] As shown above, this application constructs sample data based on the operating sequence data of wind turbine generators under different operating conditions and the corresponding true values ​​of blade tip deflection. Based on the sample data, it determines the overall relational matrix and overall eigenvalue vector corresponding to all operating conditions. By sorting the data, it matches the operating conditions of the time series data to be estimated, and combines this with the overall relational matrix to determine the blade tip deflection estimation time series, capturing the deterministic linear trend of blade tip deflection. For nonlinear components that cannot be fitted, root tests and differencing operations are used to ensure that the input sequence meets the model assumptions. Ljung-Box white noise is used to verify whether the model has sufficiently extracted residual information. Based on the target ARIMA model, the future trend of the residual sequence is predicted to obtain the predicted value, which is then combined with the blade tip deflection estimation time series to obtain the predicted blade tip deflection value. In this way, protection is triggered in advance when the blade enters the dangerous azimuth angle range based on the predicted blade tip deflection value, providing sufficient response time for the pitch system and significantly reducing the risk of tower collision.

[0040] Accordingly, see Figure 4As shown, this application also provides a wind turbine generator headroom protection control device based on blade tip deflection prediction, comprising: The eigenvalue vector determination module 11 is used to construct sample data based on the operating sequence data of the wind turbine generator under different operating conditions and the corresponding true value of the blade tip deflection, use the sample data to determine the total relation matrix corresponding to all operating conditions, and determine the total eigenvalue vector corresponding to all operating conditions based on the operating sequence data and the preset weight coefficient. The deflection estimation time series determination module 12 is used to determine the target feature value corresponding to the time series data to be estimated, sort the target feature value based on the total feature value vector, and use the sorting result to determine the working condition weight coefficient corresponding to the time series data to be estimated, and use the total relationship matrix and the working condition weight coefficient to determine the blade tip deflection estimation time series. The model set construction module 13 is used to determine the residual sequence by using the difference between the estimated time series of blade tip deflection and the corresponding true time series of blade tip deflection, perform a unit root test on the residual sequence based on the augmented Dickey-Fowler test method, determine the target stationary residual sequence by using the obtained unit root test results, and construct a candidate ARIMA model set based on the autocorrelation function and partial autocorrelation function of the target stationary residual sequence. The deflection prediction value determination module 14 is used to estimate the parameters of each model in the candidate ARIMA model set, perform Ljung-Box white noise test on the model residual sequence of the estimated model, determine the target ARIMA model based on the test results, use the target ARIMA model to predict the future trend of the residual sequence to obtain the corresponding predicted value, and sum the predicted value and the blade tip deflection estimation time series to obtain the blade tip deflection prediction value. The airspace protection control module 15 is used to determine whether the difference between the predicted blade tip deflection value and the preset static airspace value is less than the target airspace protection threshold when the wind turbine azimuth angle meets the preset range, and to perform corresponding airspace protection control operations based on the obtained judgment result.

[0041] In some specific embodiments, the feature vector determination module 11 may specifically include: The time-series data acquisition unit is used to collect the operating sequence data of the wind turbine generator under different operating conditions; the different operating conditions include different average wind speeds, turbulence intensities, and windward errors; the operating sequence data includes generator parameters; the generator parameters include rotor speed, rotor azimuth angle, blade pitch angle, blade root flapping moment, blade root oscillation moment, and generator torque; the generator parameters are all linearly uncorrelated. The relation matrix merging unit is used to decompose the runtime sequence data under a single operating condition using the Gram-Schmidt orthogonalization method to obtain the corresponding relation matrix. The relation matrices under all operating conditions are then merged to obtain the total relation matrix. The eigenvalue merging unit is used to weight the runtime sequence data under a single operating condition with a preset weight coefficient to obtain the corresponding eigenvalue, and to merge the eigenvalues ​​under all operating conditions to obtain the total eigenvalue vector.

[0042] In some specific embodiments, the deflection estimation timing determination module 12 may specifically include: The target condition determination unit is used to sort the target feature values ​​into the total feature value vector using a sorting algorithm to obtain a positioning feature value vector, and to determine the adjacent feature values ​​before and after the target feature value and the target operating conditions based on the positioning feature value vector. The weighting coefficient determination unit is used to determine the corresponding working condition weighting coefficient based on the adjacent feature values ​​and the target feature value. The timing estimation unit is used to determine the tip deflection estimation timing based on the total relationship matrix, the target operating conditions, the operating conditions weight coefficients, and the timing data to be estimated.

[0043] In some specific embodiments, the model collection construction module 13 may specifically include: The sequence test unit is used to perform unit root tests on the residual sequence based on the augmented Dickey-Fowler test method to obtain the corresponding test statistic. The difference order determination submodule is used to determine the target stationary residual sequence using the test statistic, and to determine the difference order based on the target stationary residual sequence; The correlation plotting unit is used to determine the autocorrelation function and partial autocorrelation function of the target stationary residual sequence, and plot the corresponding correlation plot based on the autocorrelation function and the partial autocorrelation function; The model set unit is used to determine the autoregressive order and the moving average order based on the truncation or trailing features of the correlation graph and the target information criterion, and to construct a set of candidate ARIMA models using the difference order, the autoregressive order and the moving average order; the target information criterion includes the Akaike information criterion and the Bayesian information criterion.

[0044] In some specific implementations, the difference order determination submodule may specifically include: The first target sequence determination unit is used to determine the residual sequence as the target stationary residual sequence if the test statistic is not greater than a preset critical value. The second target sequence determination unit is used to determine the residual sequence as a non-stationary residual sequence if the test statistic is greater than the preset critical value, and to perform a difference operation on the non-stationary residual sequence to obtain a new residual sequence. Then, it jumps to the step of performing a unit root test on the residual sequence based on the augmented Dickey-Fowler test method until the test statistic is not greater than the preset critical value, so as to obtain the target stationary residual sequence.

[0045] In some specific embodiments, the deflection prediction value determination module 14 may specifically include: The parameter estimation unit is used to estimate the parameters of each model in the set of candidate ARIMA models using the maximum likelihood estimation method or the conditional least squares method, so as to obtain the autoregressive coefficients and moving average coefficients. The white noise test unit is used to construct the estimated model based on the autoregressive coefficients and the moving average coefficients, and to perform Ljung-Box white noise test on the model residual sequence of the estimated model to obtain the corresponding test value. The first post-test model determination unit is used to adjust the difference order, the autoregression order, and the moving average order if the test value is not greater than the preset significance level value, so as to obtain new difference order, autoregression order, and moving average order, and then jump to the step of constructing a set of candidate ARIMA models using the difference order, autoregression order, and moving average order, until the post-test model is obtained; The second post-test model determination unit is used to characterize the test as passed if the test value is greater than the preset significance level value, and to determine the estimated post-test model that has passed the test as the post-test model. The target model determination unit is used to determine the target ARIMA model based on the goodness of fit, parameter significance level, and information criterion value of the tested model.

[0046] In some specific embodiments, the airspace protection control module 15 may specifically include: The difference judgment unit is used to determine whether the difference between the blade tip deflection prediction value and the preset static clearance value is not greater than the target clearance protection threshold when the wind turbine azimuth angle meets the preset range. If the value is not greater than the target airspace protection threshold, then the original pitch angle in the time series data to be estimated is superimposed with a compensation value, and the protection window timer is started. The airspace protection control unit is used to determine the real-time running sequence data of the next sampling cycle as the new time sequence data to be estimated during the timekeeping of the protection window timer, and then jump to the step of determining the target feature value corresponding to the time sequence data to be estimated, so as to obtain a new blade tip deflection prediction value, until the wind turbine azimuth angle does not meet the preset range, so as to complete the corresponding airspace protection control operation.

[0047] Furthermore, embodiments of this application also disclose an electronic device, Figure 5 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application. Specifically, the electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the wind turbine generator headroom protection control method based on blade tip deflection prediction disclosed in any of the foregoing embodiments. Furthermore, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0048] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0049] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0050] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the wind turbine generator airspace protection control method based on blade tip deflection prediction executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include computer programs capable of performing other specific tasks.

[0051] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned wind turbine generator headroom protection control method based on blade tip deflection prediction. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.

[0052] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0053] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0054] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0055] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0056] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for airspace protection control of wind turbine generator sets based on blade tip deflection prediction, characterized in that, include: Sample data is constructed based on the operating sequence data of wind turbine generators under different operating conditions and the corresponding true values ​​of blade tip deflection. The total relation matrix corresponding to all operating conditions is determined using the sample data. The total feature value vector corresponding to all operating conditions is determined based on the operating sequence data and preset weight coefficients. Determine the target feature value corresponding to the time series data to be estimated, sort the target feature value based on the total feature value vector, and use the sorting result to determine the working condition weight coefficient corresponding to the time series data to be estimated, and use the total relationship matrix and the working condition weight coefficient to determine the blade tip deflection estimation time series; The residual sequence is determined by the difference between the estimated time series of blade tip deflection and the corresponding true time series of blade tip deflection. The unit root test is performed on the residual sequence based on the augmented Dickey-Fowler test method. The target stationary residual sequence is determined by the obtained unit root test results. A set of candidate ARIMA models is constructed based on the autocorrelation function and partial autocorrelation function of the target stationary residual sequence. Parameter estimation is performed on each model in the candidate ARIMA model set, and Ljung-Box white noise test is performed on the model residual sequence of the estimated model. Based on the test results, the target ARIMA model is determined, and the future trend of the residual sequence is predicted using the target ARIMA model to obtain the corresponding predicted value. The predicted value and the blade tip deflection estimation time series are summed to obtain the blade tip deflection prediction value. When the wind turbine azimuth angle meets the preset range, it is determined whether the difference between the predicted blade tip deflection value and the preset static clearance value is less than the target clearance protection threshold, and corresponding clearance protection control operations are performed based on the obtained judgment result.

2. The wind turbine generator headroom protection control method based on blade tip deflection prediction according to claim 1, characterized in that, The process involves constructing sample data based on the operating sequence data of the wind turbine generator under different operating conditions and the corresponding true values ​​of blade tip deflection. This sample data is then used to determine the overall relational matrix corresponding to all operating conditions. Finally, based on the operating sequence data and preset weighting coefficients, the overall eigenvalue vector corresponding to all operating conditions is determined, including: The system collects operational sequence data of wind turbine generators under different operating conditions. These different operating conditions include varying average wind speeds, turbulence intensities, and wind-related errors. The operational sequence data includes generator parameters, which include rotor speed, rotor azimuth angle, blade pitch angle, blade root flapping torque, blade root oscillation torque, and generator torque. All generator parameters are pairwise linearly uncorrelated. Based on the runtime sequence data under a single operating condition, the Gram-Schmidt orthogonalization method is used to decompose the data to obtain the corresponding relation matrix. The relation matrices under all operating conditions are then merged to obtain the total relation matrix. The mean of the runtime sequence data under a single operating condition is weighted by a preset weight coefficient to obtain the corresponding feature value. The feature values ​​under all operating conditions are then merged to obtain the total feature value vector.

3. The wind turbine generator headroom protection control method based on blade tip deflection prediction according to claim 1, characterized in that, The step of sorting the target feature values ​​based on the total feature value vector, determining the operating condition weight coefficients corresponding to the time series data to be estimated using the obtained sorting results, and determining the blade tip deflection estimation time series using the total relation matrix and the operating condition weight coefficients includes: The target feature values ​​are sorted into the total feature value vector using a sorting algorithm to obtain the location feature value vector. Based on the location feature value vector, the adjacent feature values ​​before and after the target feature value and the target operating conditions are determined. The corresponding working condition weight coefficient is determined based on the adjacent feature values ​​and the target feature value; The blade tip deflection estimation time series is determined based on the overall relationship matrix, the target operating conditions, the operating condition weighting coefficients, and the time series data to be estimated.

4. The wind turbine generator headroom protection control method based on blade tip deflection prediction according to claim 1, characterized in that, The augmented Dickey-Fowler test method is used to perform a unit root test on the residual sequence, and the obtained unit root test results are used to determine the target stationary residual sequence. Based on the autocorrelation function and partial autocorrelation function of the target stationary residual sequence, a set of candidate ARIMA models is constructed, including: The unit root test was performed on the residual sequence based on the augmented Dickey-Fowler test method to obtain the corresponding test statistic; The target stationary residual sequence is determined using the test statistic, and the difference order is determined based on the target stationary residual sequence. Determine the autocorrelation function and partial autocorrelation function of the target stationary residual sequence, and draw the corresponding correlation plot based on the autocorrelation function and the partial autocorrelation function; Based on the truncation or trailing characteristics of the correlation graph and the target information criterion, the autoregressive order and the moving average order are determined, and a set of candidate ARIMA models is constructed using the difference order, the autoregressive order, and the moving average order; the target information criterion includes the Akaike information criterion and the Bayesian information criterion.

5. The wind turbine generator headroom protection control method based on blade tip deflection prediction according to claim 4, characterized in that, The process of determining the target stationary residual sequence using the test statistic includes: If the test statistic is not greater than the preset critical value, then the residual sequence is determined to be the target stationary residual sequence; If the test statistic is greater than the preset critical value, the residual sequence is determined to be a non-stationary residual sequence, and a difference operation is performed on the non-stationary residual sequence to obtain a new residual sequence. Then, the process jumps to the step of performing a unit root test on the residual sequence based on the augmented Dickey-Fowler test method until the test statistic is not greater than the preset critical value, so as to obtain the target stationary residual sequence.

6. The wind turbine generator headroom protection control method based on blade tip deflection prediction according to claim 4, characterized in that, The process of estimating parameters for each model in the candidate ARIMA model set, performing Ljung-Box white noise testing on the model residual sequences of the estimated models, and determining the target ARIMA model based on the test results includes: The parameters of each model in the set of candidate ARIMA models are estimated using the maximum likelihood estimation method or the conditional least squares method to obtain the autoregressive coefficients and moving average coefficients. An estimated model is constructed based on the autoregressive coefficients and the moving average coefficients. The Ljung-Box white noise test is performed on the model residual sequence of the estimated model to obtain the corresponding test value. If the test value is not greater than the preset significance level, the difference order, the autoregression order, and the moving average order are adjusted to obtain new difference order, autoregression order, and moving average order, and then the process jumps to the step of constructing a set of candidate ARIMA models using the difference order, the autoregression order, and the moving average order, until the tested model is obtained. If the test value is greater than the preset significance level, the test is considered passed, and the estimated model that passed the test is determined as the tested model. The target ARIMA model is determined based on the goodness of fit, parameter significance level, and information criterion value of the tested model.

7. The wind turbine generator headroom protection control method based on blade tip deflection prediction according to any one of claims 1 to 6, characterized in that, When the wind turbine azimuth angle meets the preset range, it is determined whether the difference between the predicted blade tip deflection value and the preset static clearance value is less than the target clearance protection threshold, and corresponding clearance protection control operations are performed based on the obtained determination result, including: When the wind turbine azimuth angle meets the preset range, it is determined whether the difference between the predicted blade tip deflection value and the preset static clearance value is not greater than the target clearance protection threshold. If the value is not greater than the target airspace protection threshold, then the original pitch angle in the time series data to be estimated is superimposed with a compensation value, and the protection window timer is started. During the protection window timer, the real-time runtime sequence data of the next sampling cycle is determined as the new time sequence data to be estimated, and then the process jumps to the step of determining the target feature value corresponding to the time sequence data to be estimated, so as to obtain a new blade tip deflection prediction value, until the wind turbine azimuth angle does not meet the preset range, so as to complete the corresponding airspace protection control operation.

8. A wind turbine generator headroom protection control device based on blade tip deflection prediction, characterized in that, include: The eigenvalue vector determination module is used to construct sample data based on the operating sequence data of the wind turbine generator under different operating conditions and the corresponding true values ​​of the blade tip deflection, use the sample data to determine the total relation matrix corresponding to all operating conditions, and determine the total eigenvalue vector corresponding to all operating conditions based on the operating sequence data and preset weight coefficients. The deflection estimation time series determination module is used to determine the target feature value corresponding to the time series data to be estimated, sort the target feature value based on the total feature value vector, and use the sorting result to determine the working condition weight coefficient corresponding to the time series data to be estimated, and use the total relationship matrix and the working condition weight coefficient to determine the blade tip deflection estimation time series. The model set construction module is used to determine the residual sequence by using the difference between the estimated time series of leaf tip deflection and the corresponding true time series of leaf tip deflection, perform a unit root test on the residual sequence based on the augmented Dickey-Fowler test method, determine the target stationary residual sequence using the obtained unit root test results, and construct a candidate ARIMA model set based on the autocorrelation function and partial autocorrelation function of the target stationary residual sequence. The deflection prediction value determination module is used to estimate the parameters of each model in the candidate ARIMA model set, perform Ljung-Box white noise test on the model residual sequence of the estimated model, determine the target ARIMA model based on the test results, predict the future trend of the residual sequence using the target ARIMA model to obtain the corresponding predicted value, and sum the predicted value and the blade tip deflection estimation time series to obtain the blade tip deflection prediction value. The airspace protection control module is used to determine whether the difference between the predicted blade tip deflection value and the preset static airspace value is less than the target airspace protection threshold when the wind turbine azimuth angle meets the preset range, and to perform corresponding airspace protection control operations based on the obtained judgment result.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor is configured to execute the computer program to implement the wind turbine generator headroom protection control method based on blade tip deflection prediction as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store a computer program, wherein the computer program, when executed by a processor, implements the wind turbine generator headroom protection control method based on blade tip deflection prediction as described in any one of claims 1 to 7.

Images