Composite load reduction control method for two-blade wind turbine generator under extreme wind condition

Abstract

The invention discloses a composite load reduction control method for a two-blade wind turbine generator under extreme wind conditions, which comprises the following steps of: acquiring wind condition data through a wind direction and wind speed sensor, inputting the acquired original wind condition data into a wind condition prediction model for wind speed and wind direction prediction, judging the wind condition in the prediction time period according to the wind condition prediction data, and outputting the wind speed and wind direction prediction result. And selecting a corresponding strategy according to the wind condition to perform yaw control on the two-blade fan, and performing independent variable pitch control on each blade of the two-blade fan to reduce the load of each blade. According to the method, the accuracy and timeliness of yaw control are improved by predicting the wind speed and the wind direction, and the accuracy of independent pitch control is improved and the stability of the system is improved by designing a brand-new independent pitch control equation applying the two-blade fan.

Description

Technical Field

[0001] The present invention relates to the fields of wind power technology, wind condition prediction technology, yaw control technology and independent pitch control technology, and in particular to a composite load reduction method for a two-blade wind turbine. Background Art

[0002] Under the goal of promoting the "dual carbon" and building a new power system, new energy, mainly wind power, has ushered in rapid development. As a key component of renewable energy, wind energy is making important contributions to a zero-carbon society. With the development of the wind power industry, the country is actively promoting the construction of large-scale wind power and photovoltaic bases with a focus on deserts, Gobi and desert areas. The desert areas have rich wind energy resources, which allows wind turbines to maintain a high power generation time and power generation in these areas; at the same time, the desert areas are far away from human settlements, making these areas ideal places for large-scale wind power development. However, the environment in the desert areas is harsh, the failure rate of wind turbines is high, and the equipment transportation is difficult, and the operation and maintenance costs have increased significantly. Therefore, how to reduce costs and increase efficiency has also become a problem that the wind power industry needs to solve urgently. The research on new large-capacity units with two blades and downwind has successfully broken through the new boundaries of low unit cost per kilowatt-hour, large capacity, high operating reliability, strong environmental adaptability, and low transportation and installation costs.

[0003] Two-blade wind turbines are usually deployed in areas with rich wind energy resources but complex terrain, especially in deserts or wasteland areas, which often face extreme wind conditions, such as strong winds, gusts, and drastic changes in wind speed. Since the blades of two-blade wind turbines are longer and have a larger wind-catching area, strong winds put a greater load on the blades, which can easily lead to blade overload or structural damage. Therefore, it is particularly important for two-blade wind turbines to perform yaw control and independent pitch control. At the same time, since two-blade wind turbines have fewer blades than three-blade wind turbines, the aerodynamic force changes more drastically when the wind conditions suddenly change, which can easily cause the wind turbine structure to bear a greater load. Therefore, a more intelligent wind condition prediction mechanism is needed to optimize the yaw system.

[0004] At present, there are many relevant literatures and patent studies at home and abroad that apply wind condition prediction, yaw control and independent pitch control to wind turbine load reduction, and some effective solutions have been proposed.

[0005] 1. In the article titled "Wind direction prediction for yaw control of wind turbines", the authors Song D et al. proposed a hybrid model that integrates autoregressive moving average (ARIMA) into Kalman filter (KF) to predict wind direction and optimize yaw control. However, the yaw control strategy of this prediction model is mainly aimed at predicting wind direction, achieving the purpose of starting yaw action in advance, but ignoring the prediction of wind speed and not combining the optimization of yaw control parameters. Therefore, the yaw performance in different wind speed ranges needs to be improved.

[0006] 2. P. Gaucher et al. configured a control action signal for the yaw mechanism of the wind turbine according to multiple wind conditions and multiple load conditions in a patent application with application number CN202311029012.2 and titled “Load-dependent autonomous yaw control for wind turbines”. The control system determines the start and stop of the yaw system based on monitoring multiple change values ​​related to the load conditions. However, for a two-blade wind turbine, the yaw system only adjusts the yaw through signal changes, which wastes the wind-facing ability of the two-blade wind turbine as a downwind wind turbine, reduces efficiency and increases the burden on the yaw motor.

[0007] 3. In the article entitled "Feedback-feedforward individual pitch control for windturbine load reduction", the authors Selvam K et al. use the Coleman transform to convert all signals into a non-rotating reference frame. In the Coleman domain, using the optimal multivariable LQG controller and the feedforward anti-disturbance controller based on the wind speed signal, the LQG control can minimize the rotor tilt and yaw torque, while the feedforward controller achieves further improvement by suppressing the influence of the low-frequency components of the wind on the rotor torque. However, if the basic multi-blade coordinate transformation used for independent pitch control of three-blade wind turbines is applied to two-blade wind turbines, this transformation is singular, which may cause the control system to fail to work properly, causing the system to become unstable, and even affecting the operation and power output of the wind turbine. Summary of the invention

[0008] In view of this, in view of the technical problems that the existing two-blade wind turbines have insufficient adjustment capabilities under extreme wind conditions, the existing adjustment methods are not highly adaptable to two-blade wind turbines, and have slow response speeds, the purpose of the present invention is to provide a composite load reduction control method for a two-blade wind turbine under extreme wind conditions, so as to achieve stable operation of the two-blade wind turbine under extreme wind conditions and improve the power generation efficiency of the two-blade wind turbine.

[0009] The composite load reduction control method of a two-blade wind turbine under extreme wind conditions of the present invention comprises the following steps:

[0010] 1) Collecting wind condition data through wind direction and wind speed sensors, wherein the wind condition data is time series data including wind speed and wind direction;

[0011] 2) inputting the original wind condition data collected in step 1) into the wind condition prediction model to predict wind speed and wind direction;

[0012] 3) Determine the wind conditions within the forecast period based on the wind forecast data, and select the corresponding strategy to perform yaw control on the two-blade wind turbine based on the wind conditions:

[0013] ① Set the maximum wind speed and maximum wind direction change angle within the forecast time period. If the predicted wind speed obtained within the forecast time period is less than the set maximum wind speed, and the predicted wind direction change angle is less than the set maximum wind direction change angle, the wind condition within the forecast time period is normal wind condition, otherwise the wind condition within the forecast time period is extreme wind condition;

[0014] ② If the wind conditions in the forecast period are normal, a passive yaw control strategy is used in the forecast period to automatically align the yaw angle of the wind turbine with the wind direction angle;

[0015] ③ If the wind conditions are extreme during the forecast period, the active yaw control strategy is used to adjust the yaw angle of the wind turbine during the forecast period;

[0016] 4) Performing independent pitch control on each blade of a two-blade wind turbine to reduce the load on each blade, including the following steps:

[0017] A) Calculate the total load of the entire blade;

[0018] B) The calculated total blade load is brought from the rotating reference system to the fixed non-rotating reference system to obtain the matrix expression of the blade load, and then the matrix expression is rewritten into a vector expression, and the collective mode and differential mode of the blade load are obtained from the vector expression equation;

[0019] C) Designing independent pitch control equations for differential mode and independent pitch control equations for collective mode, and applying the pitch control equations for differential mode and collective mode in parallel to reduce blade load.

[0020] Further, the wind condition prediction model described in step 2) is composed of a VMD model, an EEMD model, an LSTM model and an LSSVM model, and the training and wind condition prediction steps of the wind condition prediction model include:

[0021] ① The VMD model is used to decompose the original wind condition data to obtain K subsequences and residuals;

[0022] ② The residual data decomposed in step ① is further decomposed by the EEMD model to obtain N subsequences and residuals;

[0023] ③ Input the K subsequences decomposed in step ① and the N subsequences decomposed in step ② and the residual into the LSTM model, train the LSTM model, and obtain the wind condition prediction data and wind condition prediction error data output by the LSTM model during the training process;

[0024] ④ Input the wind condition prediction error data obtained in step ③ into the LSSVM model, train the LSSVM model, and obtain the wind condition prediction error correction data output by the LSSVM model during the training process;

[0025] ⑤ Add the wind condition prediction data output by the LSTM model in step ③ and the wind condition prediction error correction data output by the LSSVM model in step ④ to obtain the final wind condition prediction data, wherein the wind condition prediction data includes the predicted wind speed and the predicted wind direction.

[0026] Further, the active control strategy described in step 3) includes the steps of:

[0027] a) Calculate the average value of the predicted wind speed within the prediction period, and set different yaw error angle thresholds T according to the average value of the predicted wind speed angle and the yaw error duration threshold T time ;

[0028] b) The yaw error angle is obtained by the predicted average wind direction every 5 seconds and the wind turbine cabin angle θ With duration T, when the yaw error angle Greater than the preset yaw error angle threshold T angle , the duration T is greater than the preset yaw error duration threshold T time When the control cabin deflects at a speed of 0.6 (°) / s, the yaw error after m seconds is The yaw adjustment time is obtained as

[0029] c) Start the yaw to wind action (mt) seconds in advance until θ = real-time wind direction angle;

[0030] d) Finally, when the time exceeds the current forecast time period, the average forecast wind speed in the next forecast time period is recalculated.

[0031] Further, the step A) of calculating the total load of the entire blade includes:

[0032] Divide the blade into n segments and calculate the relative wind speed v of each blade segment rel

[0033]

[0034] Where v is the predicted wind speed, Ω is the blade angular velocity, and r is the radius of rotation of the blade segment midpoint;

[0035] Calculate the lift and drag per unit area on the blade segment,

[0036]

[0037] Where L is the lift, C L is the lift coefficient, D is the drag, C D is the drag coefficient, dr is the arc length of the blade segment, and ρ is the air density;

[0038] The formula for calculating the aerodynamic torque on each blade segment is:

[0039] q · =L·r·sin(β)-D·r·cos(β) (3)

[0040] Among them, q · is the aerodynamic torque, β is the windward angle of the blade;

[0041] Integrating the moment for each blade segment gives the total load for the entire blade:

[0042]

[0043] Among them, q (b) is the total load, is the moment of the ith blade segment.

[0044] Furthermore, in step B), the calculated total blade load is brought from the rotating reference system to the fixed non-rotating reference system to obtain a matrix expression of the blade load, including:

[0045] According to the following signal conversion relationship:

[0046]

[0047] Where: B is the total number of leaves, b is the leaf number, q (b) To rotate the signal, we can express formulas (1) and (2) in matrix form until:

[0048]

[0049] In the formula, q is replaced by M, M y,1 and M y,B Represents each blade load respectively;

[0050] Rewriting formula (3) into vector notation yields

[0051]

[0052] Among them, M0 is the collective mode and M1 is the differential mode.

[0053] Further, in step C): the independent pitch control equation for the differential mode is designed as follows:

[0054]

[0055] Where: K 1p is the proportional gain; s is the Laplace operator.

[0056] NF inv,1P (s) is the inverse notch filter set at the 1P frequency, and its calculation formula is as follows:

[0057]

[0058] Where ω 1P is the inverse notch filter 1P frequency, ζ 1P and β 1P are the inverse notch filter 1P parameters respectively;

[0059] is a high-pass filter with a cutoff frequency of 0.4Hz. Its calculation formula is as follows:

[0060]

[0061] In the formula are the high-pass filter parameters, and are the high-pass filter frequencies, respectively;

[0062] It is a low-pass filter with a cutoff frequency of 2Hz. Its calculation formula is as follows:

[0063]

[0064] In the formula is the low-pass filter frequency, is the low-pass filter parameter;

[0065] NF 3P (s) is a standard notch filter set at the 3P frequency, and its calculation formula is as follows:

[0066]

[0067] Where ω 3P is the notch filter 3P frequency, ζ 3P and β 3P They are the 3P parameters of the notch filter respectively;

[0068] The independent pitch control equations designed for collective mode are as follows:

[0069]

[0070] Where: K 2P is the proportional gain,

[0071] NF inv,2P (s) is the inverse notch filter set at the 2P frequency, and its calculation formula is as follows:

[0072]

[0073] Where ω 2P is the inverse notch filter 2P frequency, ζ 2P and β 2P They are the 2P parameters of the inverse notch filter respectively;

[0074] It is a high-pass filter with a cutoff frequency of 3Hz. Its calculation formula is as follows:

[0075]

[0076] In the formula are the high-pass filter parameters, and are the high-pass filter frequencies, respectively;

[0077] It is a low-pass filter with a cutoff frequency of 1.3Hz. Its calculation formula is as follows:

[0078]

[0079] In the formula is the low-pass filter frequency, are the low-pass filter parameters.

[0080] Beneficial effects of the present invention:

[0081] 1. The composite load reduction control method for two-blade wind turbines under extreme wind conditions of the present invention proposes a wind condition prediction model based on the combination of variational mode decomposition (VMD)-ensemble empirical mode decomposition (EEMD)-long short-term memory (LSTM)-least squares support vector machine (LSSVM), which completes the accurate prediction of short-term wind speed and wind direction, so that the wind turbine can perform active yaw control more accurately and quickly to cope with extreme wind conditions, and can reduce the impact force on the blades and yaw motors under extreme wind conditions, which is beneficial to improving the stability of the wind turbine system. And accurate yaw control also lays a good foundation for independent pitch control of blades.

[0082] 2. The composite load reduction control method for two-blade wind turbines under extreme wind conditions of the present invention combines the passive yaw and active yaw of the two-blade fan. Under normal circumstances, the two-blade fan is used as a downwind fan to automatically yaw according to the wind's ability, which can reduce the energy consumption of yaw control and improve the overall performance of the fan.

[0083] 3. The composite load reduction control method for a two-blade wind turbine under extreme wind conditions of the present invention provides a linear non-singular coordinate transformation applied to a two-blade wind turbine, and on this basis designs a new independent pitch control equation applied to a two-blade wind turbine, thereby improving the accuracy of independent pitch control and improving the stability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 Schematic diagram of the wind condition prediction model structure.

[0085] Figure 2 This is the flow chart of the yaw control strategy.

[0086] Figure 3 This is the composite load reduction flow chart for a two-blade fan. DETAILED DESCRIPTION

[0087] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0088] The composite load reduction control method for a two-blade wind turbine under extreme wind conditions in this embodiment includes the following steps:

[0089] 1) Collect wind condition data through wind direction and wind speed sensors, where the wind condition data is time series data including wind speed and wind direction.

[0090] 2) Inputting the original wind condition data collected in step 1) into the wind condition prediction model to predict wind speed and wind direction. The wind condition prediction model described in this step is composed of a VMD model, an EEMD model, an LSTM model and an LSSVM model. The training and wind condition prediction steps of the wind condition prediction model include:

[0091] ① The VMD model is used to decompose the original wind data to obtain K subsequences and residuals. This step can effectively extract the characteristics of wind data based on the signal decomposition method and improve the accuracy of the prediction model.

[0092] ② The residual data decomposed in step ① is further decomposed by the EEMD model to obtain N subsequences and residuals. Although VMD performs well in signal decomposition and feature extraction, the residual information remaining after VMD decomposition is discarded in the existing method, resulting in the loss of some information. This embodiment further extracts the residual information through the EEMD model, which helps to further improve the prediction accuracy of the model.

[0093] ③ Input the K subsequences decomposed from step ① and the N subsequences decomposed from step ② and the residual into the LSTM model, train the LSTM model, and obtain the wind condition prediction data and wind condition prediction error data output by the LSTM model during the training process. The LSTM network can extract the complex feature relationship of long time series and short time series, and the wind condition change process is closely related to time. In this embodiment, the complex feature relationship of wind condition time series data is extracted by the LSTM network, which can significantly improve the wind condition prediction accuracy.

[0094] ④ Input the wind condition prediction error data obtained in step ③ into the LSSVM model, train the LSSVM model, and obtain the wind condition prediction error correction data output by the LSSVM model during the training process;

[0095] LSSVM is a small batch data modeling algorithm that transfers the linear inseparable problem in low-dimensional space to high-dimensional space for solution. It has good fitting ability and can further improve the prediction accuracy of the model by correcting the prediction error through the LSSVM model.

[0096] ⑤ Add the wind condition prediction data output by the LSTM model in step ③ and the wind condition prediction error correction data output by the LSSVM model in step ④ to obtain the final wind condition prediction data, wherein the wind condition prediction data includes the predicted wind speed and the predicted wind direction.

[0097] 3) Determine the wind conditions within the forecast period based on the wind forecast data, and select the corresponding strategy to perform yaw control on the two-blade wind turbine based on the wind conditions:

[0098] ① Set the maximum wind speed and maximum wind direction change angle within the forecast time period. If the predicted wind speed obtained within the forecast time period is less than the set maximum wind speed, and the predicted wind direction change angle is less than the set maximum wind direction change angle, the wind condition within the forecast time period is normal wind condition, otherwise the wind condition within the forecast time period is extreme wind condition.

[0099] ② If the wind conditions within the forecast period are normal, a passive yaw control strategy is used within the forecast period to automatically align the yaw angle of the wind turbine with the wind direction angle.

[0100] Under normal wind conditions, the aerodynamic torque of the fan can be approximately calculated by the windward area of ​​the fan blades and the wind speed. Assuming that the yaw angle of the fan is θ yaw is the deviation angle relative to the ideal direction of the wind, and the aerodynamic moment M yaw It can be expressed as:

[0101] M yaw =C·A·V 2 ·sin(θ yaw )

[0102] Among them, C is the aerodynamic coefficient related to the fan blade design, which reflects the ability of the fan blade to generate aerodynamic torque. A is the effective frontal area of ​​the fan blade. V is the real-time wind speed.

[0103] According to the generated aerodynamic torque, the fan will adjust its yaw angle θ yaw , until it is aligned as closely as possible with the wind direction (ideal yaw angle). The yaw angle of the wind turbine changes with wind speed and direction, and the rate of change of the yaw angle can be described as follows:

[0104]

[0105] in, is the rate of change of the yaw angle, I yaw is the yaw moment of inertia of the wind turbine.

[0106] Combining the above two equations, the yaw angle change rate of the two-blade fan in the passive yaw mode is obtained:

[0107]

[0108] ③ If the wind conditions are extreme during the forecast period, the wind turbine needs more precise adjustment to ensure optimal operating efficiency. Therefore, an active yaw control strategy is used to adjust the yaw angle of the wind turbine during the forecast period. The active control strategy includes the following steps:

[0109] a) Calculate the average value of the predicted wind speed within the prediction period, and set different yaw error angle thresholds T according to the average value of the predicted wind speed angle and the yaw error duration threshold Ttime The smaller the predicted average wind speed, the smaller the impact of wind on the system. The yaw error angle threshold T is set angle and the yaw error duration threshold T time The larger the predicted average wind speed is, the more significant the impact of wind on the system will be. The yaw error angle threshold T is set angle and the yaw error duration threshold T time Specifically, the wind speed can be divided into different intervals, such as: a low wind speed interval with a wind speed less than 5m / s, a medium wind speed interval with a wind speed between 5m / s and 10m / s, and a high wind speed interval with a wind speed greater than 10m / s. Each wind speed interval corresponds to a yaw error angle threshold T. angle and the yaw error duration threshold T time In this way, the corresponding yaw error angle threshold T can be obtained according to the wind speed interval described by the predicted wind speed average value angle and the yaw error duration threshold T time ; The finer the wind speed interval is divided, the higher the yaw error angle threshold T is set. angle and the yaw error duration threshold T time The more suitable.

[0110] In this embodiment, T angle The optimization range is (5°, 6°, 7°, ..., 30°), T time The optimization range is (5, 10, 15, ..., 60s). The average predicted wind speed of 30 minutes is used as the basis for setting the yaw control parameters, and the predicted average wind direction every 5 seconds is used as the basis for judging whether the yaw action should be taken in advance. Selecting the average wind direction can smooth the small-scale and irregular fluctuations of the second-level wind direction data, which can effectively improve the accuracy of the prediction model and the accuracy of the yaw action judgment.

[0111] b) The yaw error angle is obtained by the predicted average wind direction every 5 seconds and the wind turbine cabin angle θ With duration T, when the yaw error angle Greater than the preset yaw error angle threshold T angle , the duration T is greater than the preset yaw error duration threshold T time When the control cabin deflects at a speed of 0.6 (°) / s, the yaw error after m seconds is The yaw adjustment time is obtained as

[0112] c) Start the yaw to wind action (mt) seconds in advance until θ = real-time wind direction angle.

[0113] d) Finally, when the time exceeds the current forecast time period, the average forecast wind speed in the next forecast time period is recalculated.

[0114] 4) Performing independent pitch control on each blade of a two-blade wind turbine to reduce the load on each blade, including the following steps:

[0115] A) Calculate the total load on the entire blade, including:

[0116] Divide the blade into n segments and calculate the relative wind speed v of each blade segment rel

[0117]

[0118] Where v is the predicted wind speed, Ω is the blade angular velocity, and r is the radius of rotation of the blade segment midpoint.

[0119] Calculate the lift and drag per unit area on the blade segment,

[0120]

[0121] Where L is the lift, C L is the lift coefficient, D is the drag, C D is the drag coefficient, dr is the arc length of the blade segment, and ρ is the air density.

[0122] The formula for calculating the aerodynamic torque on each blade segment is:

[0123] q · =L·r·sin(β)-D·r·cos(β) (3)

[0124] Among them, q · is the aerodynamic torque and β is the windward angle of the blade.

[0125] Integrating the moment for each blade segment gives the total load for the entire blade:

[0126]

[0127] Among them, q (b) is the total load, is the moment of the ith blade segment.

[0128] B) The calculated total blade load is brought from the rotating reference system to the fixed non-rotating reference system to obtain the matrix expression of the blade load, and then the matrix expression is rewritten into a vector expression to obtain the collective mode and differential mode of the blade load from the vector expression equation. This step specifically includes:

[0129] According to the following signal conversion relationship:

[0130]

[0131] Where: B is the total number of leaves, b is the leaf number, q (b) To rotate the signal, we can express formulas (1) and (2) in matrix form until:

[0132]

[0133] In the formula, q is replaced by M, M y,1 and M y,B Represents each blade load respectively;

[0134] Rewriting formula (3) into vector notation yields

[0135]

[0136] Among them, M0 is the collective mode and M1 is the differential mode;

[0137] The corresponding inverse transform is given by:

[0138]

[0139] Among them, θ1 and θ2 are the rotation angles of the two blade rotation domains, and are the rotation angles of the two blades in the non-rotating domain, respectively.

[0140] C) Designing independent pitch control equations for differential mode and independent pitch control equations for collective mode, and applying the pitch control equations for differential mode and collective mode in parallel to reduce blade load. This step specifically includes:

[0141] The independent pitch control equations designed for differential mode are as follows:

[0142]

[0143] Where: K 1p is the proportional gain; s is the Laplace operator.

[0144] NF inv,1P (s) is the inverse notch filter set at the 1P frequency, and its calculation formula is as follows:

[0145]

[0146] Where ω 1P is the inverse notch filter 1P frequency, ζ 1P and β 1P are the inverse notch filter 1P parameters; in this embodiment, ω 1P Set to 0.695Hz, 1P Set to 0.9, β 1P Set to 0.009.

[0147] is a high-pass filter with a cutoff frequency of 0.4Hz. Its calculation formula is as follows:

[0148]

[0149] In the formula are the high-pass filter parameters, and are respectively the high-pass filter frequencies; in this embodiment, Set to 0.7, Set to 0.4*2Πrad / s; Set to 0.0004.2Πrad / sec.

[0150] It is a low-pass filter with a cutoff frequency of 2Hz. Its calculation formula is as follows:

[0151]

[0152] In the formula is the low-pass filter frequency, is the low-pass filter parameter; in this embodiment, Set to 4Πrad / s, Set to 0.7.

[0153] NF 3P (s) is a standard notch filter set at the 3P frequency, and its calculation formula is as follows:

[0154]

[0155] Where ω 3P is the notch filter 3P frequency, ζ 3P and β 3P are the 3P parameters of the notch filter respectively; in this embodiment, ω 3P Set to 2.085Hz, 3P Set to 0.3, β 3P Set to 0.006.

[0156] The independent pitch control equations designed for collective mode are as follows:

[0157]

[0158] Where: K 2P is the proportional gain,

[0159] NF inv,2P (s) is the inverse notch filter set at the 2P frequency, and its calculation formula is as follows:

[0160]

[0161] Where ω 2P is the inverse notch filter 2P frequency, ζ 2P and β 2P are the 2P parameters of the inverse notch filter respectively; in this embodiment, ω 2P Set to 1.39HZ, 2P Set to 0.5, β 2P Set to 0.005.

[0162] It is a high-pass filter with a cutoff frequency of 3Hz. Its calculation formula is as follows:

[0163]

[0164] In the formula are the high-pass filter parameters, and are respectively the high-pass filter frequencies; in this embodiment, Set to 0.7, Set to 2.6Πrad / s, Set to 0.0008Πrad / s.

[0165] It is a low-pass filter with a cutoff frequency of 1.3Hz. Its calculation formula is as follows:

[0166]

[0167] In the formula is the low-pass filter frequency, is the low-pass filter parameter; in this embodiment Set to 1.39HZ, Set to 0.7.

[0168] In the process of controlling the pitch of the wind turbine, by jointly executing the control equations (10) and (15) described in the embodiment, the 1P and 2P loads in the blade root bending moment can be removed, and the execution of the control equation (10) also removes the 1P rotating hub moment, reducing the load transfer from the blade to the shaft. The execution of the control equation (15) further reduces the 2P load in the tower base moment.

[0169] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A composite load reduction control method for a two-blade wind turbine under extreme wind conditions, characterized in that: The following steps are involved: 1) Collecting wind condition data through wind direction and wind speed sensors, wherein the wind condition data is time series data including wind speed and wind direction; 2) inputting the original wind condition data collected in step 1) into a wind condition prediction model to predict wind speed and wind direction; 3) Determine the wind conditions within the forecast period based on the wind forecast data, and select the corresponding strategy to perform yaw control on the two-blade wind turbine based on the wind conditions: ① Set the maximum wind speed and maximum wind direction change angle within the forecast time period. If the predicted wind speed obtained within the forecast time period is less than the set maximum wind speed, and the predicted wind direction change angle is less than the set maximum wind direction change angle, the wind condition within the forecast time period is normal wind condition, otherwise the wind condition within the forecast time period is extreme wind condition; ② If the wind conditions in the forecast period are normal, a passive yaw control strategy is used in the forecast period to automatically align the yaw angle of the wind turbine with the wind direction angle; ③ If the wind conditions are extreme during the forecast period, the active yaw control strategy is used to adjust the yaw angle of the wind turbine during the forecast period; 4) Performing independent pitch control on each blade of a two-blade wind turbine to reduce the load on each blade includes the following steps: A) Calculate the total load of the entire blade; B) The calculated total blade load is brought from the rotating reference system to the fixed non-rotating reference system to obtain a matrix expression of the blade load, and then the matrix expression is rewritten into a vector expression, and the collective mode and differential mode of the blade load are obtained from the vector expression equation; C) Designing independent pitch control equations for differential mode and independent pitch control equations for collective mode, and applying the pitch control equations for differential mode and collective mode in parallel to reduce blade load.

2. The composite load reduction control method for a two-blade wind turbine under extreme wind conditions according to claim 1 is characterized in that: The wind condition prediction model described in step 2) is composed of a VMD model, an EEMD model, an LSTM model and an LSSVM model. The training of the wind condition prediction model and the wind condition prediction steps include: ① The VMD model is used to decompose the original wind condition data to obtain K subsequences and residuals; ② The residual data decomposed in step ① is further decomposed by the EEMD model to obtain N subsequences and residuals; ③ Input the K subsequences decomposed in step ① and the N subsequences decomposed in step ② and the residual into the LSTM model, train the LSTM model, and obtain the wind condition prediction data and wind condition prediction error data output by the LSTM model during the training process; ④ Input the wind condition prediction error data obtained in step ③ into the LSSVM model, train the LSSVM model, and obtain the wind condition prediction error correction data output by the LSSVM model during the training process; ⑤ Add the wind condition prediction data output by the LSTM model in step ③ and the wind condition prediction error correction data output by the LSSVM model in step ④ to obtain the final wind condition prediction data, wherein the wind condition prediction data includes the predicted wind speed and the predicted wind direction.

3. The composite load reduction control method for a two-blade wind turbine under extreme wind conditions according to claim 2 is characterized in that: The active control strategy described in step 3) comprises the steps of: a) Calculate the average value of the predicted wind speed within the prediction period, and set different yaw error angle thresholds T according to the average value of the predicted wind speed angle and the yaw error duration threshold T time ; b) The yaw error angle is obtained by the predicted average wind direction every 5 seconds and the wind turbine cabin angle θ With duration T, when the yaw error angle Greater than the preset yaw error angle threshold T angle , the duration T is greater than the preset yaw error duration threshold T time When the control cabin deflects at a speed of 0.6 (°) / s, the yaw error after m seconds is The yaw adjustment time is obtained as c) Start the yaw to wind action (mt) seconds in advance until θ = real-time wind direction angle; d) Finally, when the time exceeds the current forecast time period, the average forecast wind speed in the next forecast time period is recalculated.

4. The composite load reduction control method for a two-blade wind turbine under extreme wind conditions according to claim 3 is characterized in that: The step A) of calculating the total load of the entire blade comprises: Divide the blade into n segments and calculate the relative wind speed v of each blade segment rel Where v is the predicted wind speed, Ω is the blade angular velocity, and r is the radius of rotation of the blade segment midpoint; Calculate the lift and drag per unit area on the blade segment, Where L is the lift, C L is the lift coefficient, D is the drag, C D is the drag coefficient, dr is the arc length of the blade segment, and ρ is the air density; The formula for calculating the aerodynamic torque on each blade segment is: q · =L·r·sin(β)-D·r·cos(β) (3) Among them, q · is the aerodynamic torque, β is the windward angle of the blade; Integrating the moment for each blade segment gives the total load for the entire blade: Among them, q (b) is the total load, is the moment of the ith blade segment.

5. The composite load reduction control method for a two-blade wind turbine under extreme wind conditions according to claim 4 is characterized in that: In step B), the calculated total blade load is brought from the rotating reference system to the fixed non-rotating reference system to obtain a matrix expression of the blade load, including: According to the following signal conversion relationship: Where: B is the total number of leaves, b is the leaf number, q (b) To rotate the signal, we can express formulas (1) and (2) in matrix form until: In the formula, M is used to replace q, M y,1 and M y,B Represents each blade load respectively; Rewriting formula (3) into vector notation yields Among them, M0 is the collective mode and M1 is the differential mode.

6. The composite load reduction control method for a two-blade wind turbine under extreme wind conditions according to claim 5 is characterized in that: In step C): the independent pitch control equations for differential mode are designed as follows: C IPC-1P (s)=K 1P ×NF inv,1P (s)×H 1P (s)×L 1P (s)×NF 3P (s) (9) Where: K 1p is the proportional gain; s is the Laplace operator. NF inv,1P (s) is the inverse notch filter set at the 1P frequency, and its calculation formula is as follows: Where ω 1P is the inverse notch filter 1P frequency, ζ 1P and β 1P are the inverse notch filter 1P parameters respectively; H 1P (s) is a high-pass filter with a cutoff frequency of 0.4 Hz, and its calculation formula is as follows: In the formula are the high-pass filter parameters, and are the high-pass filter frequencies, respectively; L 1P (s) is a low-pass filter with a cutoff frequency of 2 Hz, and its calculation formula is as follows: In the formula is the low-pass filter frequency, is the low-pass filter parameter; NF 3P (s) is a standard notch filter set at the 3P frequency, and its calculation formula is as follows: Where ω 3P is the notch filter 3P frequency, ζ 3P and β 3P They are the 3P parameters of the notch filter respectively; The independent pitch control equations designed for collective mode are as follows: C IPC-2P (s)=K 2P ×NF inv,2P (s)×H 2P ×L 2P (14) Where: K 2P is the proportional gain, NF inv,2P (s) is the inverse notch filter set at the 2P frequency, and its calculation formula is as follows: Where ω 2P is the inverse notch filter 2P frequency, ζ 2P and β 2P They are the 2P parameters of the inverse notch filter respectively; H 2P It is a high-pass filter with a cutoff frequency of 3Hz. Its calculation formula is as follows: In the formula are the high-pass filter parameters, and are the high-pass filter frequencies, respectively; L 2P It is a low-pass filter with a cutoff frequency of 1.3Hz. Its calculation formula is as follows: In the formula is the low-pass filter frequency, are the low-pass filter parameters.

Images