A wind turbine independent pitch control method based on SVM partition weight coefficient allocation

Abstract

The present invention relates to an independent variable pitch control method for a wind turbine based on SVM partition weight coefficient allocation, the method comprising: performing independent variable pitch control partitioning according to blade azimuth angle θ; when the angle between the current position to which the blade rotates and the horizontal plane is less than a preset angle, independent variable pitch control based on azimuth weight system allocation is used to perform variable pitch adjustment; when the angle between the current position to which the blade rotates and the horizontal plane is greater than a preset angle, independent variable pitch control based on SVM fuzzy weight coefficient is used to perform variable pitch adjustment. Compared with the prior art, the present invention realizes that the wind turbine can alleviate the unbalanced load of the wind wheel while maintaining the stability of the output power.

Description

Technical Field

[0001] The invention relates to the field of wind power generation, and in particular to a wind turbine independent pitch control method based on SVM partition weight coefficient allocation. Background Art

[0002] Independent pitch control is mainly divided into two categories. One is the control based on force analysis. This strategy aims to suppress aerodynamic loads. It first measures the load and then independently controls the blades based on a certain control algorithm. The other is the control based on the change of blade azimuth angle. This strategy is less dependent on the wind turbine load model. It only needs to measure the blade azimuth angle and then reduce the load by periodically changing the blade pitch angle.

[0003] At present, for the pitch angle control of wind turbine independent variable pitch, most of them use traditional PI control or compound PI control, fuzzy anti-disturbance control, LQG control, sliding mode variable structure control and neural network control; however, wind energy has the characteristics of randomness and instability, and wind shear, tower shadow effect, turbulence and gusts will be encountered in the actual operation process. In addition, the characteristics of wind turbines will change with time and space, which often makes the traditional control system deviate from the nominal characteristics based on the design, making it difficult to achieve good control performance and stability.

[0004] Corresponding to the sliding mode variable structure control, when the system state trajectory reaches the sliding surface, it traverses back and forth on both sides of the sliding surface instead of strictly sliding along the sliding surface to the equilibrium point, thus forming a buffeting. Considering the limited actual wind power generation variable pitch control system, the sliding film gain causes high-frequency buffeting, which aggravates the fatigue load of the wind turbine. However, the neural network has the problem of building a high-dimensional data model under limited sample conditions, and it is easy to have multiple local optimal situations. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a wind turbine independent pitch control method based on SVM partition weight coefficient allocation. This method can alleviate the unbalanced load of the wind wheel while maintaining the stability of the wind turbine output power.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] The present invention provides a wind turbine independent pitch control method based on SVM partition weight coefficient allocation, the method is:

[0008] Independent pitch control partitioning is performed according to the blade azimuth angle θ;

[0009] When the angle between the blade azimuth angle θ and the horizontal plane is less than the preset angle, independent pitch control based on azimuth weight coefficient allocation is used for independent pitch adjustment; otherwise, independent pitch control based on SVM fuzzy weight coefficient is used for independent pitch adjustment.

[0010] Preferably, the preset angle is 30°.

[0011] Preferably, the independent pitch control based on azimuth weight coefficient allocation is specifically:

[0012] First, the rated power given value is subtracted from the output power as the input of the power controller, and the power controller outputs a uniform pitch angle Δβ;

[0013] Then, the uniform pitch angle Δβ is input into the weight coefficient distributor to redistribute the pitch angle change Δβ of each blade i (i=1,2,3), thereby completing independent pitch adjustment.

[0014] Among them, the pitch angle change of each blade Δβ i The expression is: Δβ i =K i ×Δβ,K i (i=1,2,3) is the azimuth weight coefficient.

[0015] Preferably, the azimuth weight coefficient K i The expression is:

[0016]

[0017] Wherein, n is a set positive integer; R is the blade length, H0 is the height from the ground; θ is the blade azimuth angle.

[0018] Preferably, n is set to 2.

[0019] Preferably, the independent pitch control based on SVM fuzzy weight coefficient is specifically:

[0020] Fuzzy power control module: The difference between the rated power given value and the output power is used as the input of the fuzzy power controller, and the fuzzy power controller outputs a uniform pitch angle Δβ;

[0021] SVM weight coefficient independent variable pitch control module: the azimuth weight coefficient K output by the weight coefficient distribution unit i (i=1,2,3) and the azimuth weight coefficient compensation value ΔK output by the SVM optimization compensation unit i (i=1,2,3) is input to the pitch angle dynamic coefficient adjustment calculation unit to obtain the dynamic weighting coefficient δ i(i=1,2,3); and the pitch angle variation of each blade Δβ is calculated by combining the uniform pitch angle Δβ output by the fuzzy power controller i , and then complete independent pitch adjustment.

[0022] Preferably, the solution process of the SVM weight coefficient independent pitch control module specifically includes the following steps:

[0023] Step S1, obtaining a new blade azimuth angle θ i (i=1,2,3) and blade axial aerodynamic force F i (i=1,2,3), and use it as a new training sample;

[0024] Step S2: determine whether the training sample is valid according to the set evaluation period. If it is valid, add it to the training sample set; otherwise, discard the data and go to step S1;

[0025] Step S3, according to the KKT condition and the Lagrange multiplier, obtain the boundary support vector set B, the error support vector set E and the retained sample set R;

[0026] Step S4: The SVM optimization compensation unit outputs the azimuth weight coefficient compensation value ΔK i (i=1,2,3);

[0027] Step S5: Calculate the azimuth weight coefficient compensation value ΔK i (i=1,2,3) output as the reference value, and the dynamic weight coefficient δ obtained by inputting the fuzzy power controller, weight coefficient allocation unit and SVM optimization compensation unit i =(i=1,2,3) Generate pitch angle reference value Δβ i (i=1,2,3), control the variable pitch object;

[0028] Step S6: Use a decreasing algorithm to delete the samples that need to be forgotten in the training sample set to obtain a new support vector set BD, an error support vector set ED, and a retained sample set RD;

[0029] Step S7, wait for data to be updated online, and go to step S1.

[0030] Preferably, the dynamic weighting coefficient δ i (i=1,2,3) the expression is:

[0031]

[0032] Among them, K i =(i=1,2,3) is the azimuth weight coefficient, ΔK i =(i=1,2,3) is the azimuth angle weight coefficient compensation value.

[0033] According to a second aspect of the present invention, there is provided an electronic device, comprising a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the program, any one of the methods described above is implemented.

[0034] According to a third aspect of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, any one of the methods described above is implemented.

[0035] Compared with the prior art, the present invention has the following advantages:

[0036] 1) The present invention designs a wind turbine independent pitch control method based on SVM partition weight coefficient allocation. The method determines the control algorithm according to the degree of influence of the axial unbalanced aerodynamic force borne by the blades and the wind rotor. If the influence is small, the independent pitch control based on the blade azimuth angle weight coefficient allocation is selected. If the influence is large, the independent pitch control based on the SVM fuzzy weight coefficient is selected. Through the above-mentioned partition control, the pitch angle offset is made more reasonable and accurate, and the partitioned independent pitch of the blades is realized. While maintaining the stability of the output power, the unbalanced load of the wind rotor is alleviated;

[0037] 2) The partition angle of the present invention is obtained by analyzing the change of wind speed at the blade tip at different times and heights under the effect of wind shear, taking into account the axial unbalanced aerodynamic force borne by the blades and the wind wheel of the wind turbine, and the partition is more reasonable;

[0038] 3) When the influence of axial unbalanced aerodynamic force is small, independent variable pitch control based on blade azimuth angle weight coefficient distribution is used for variable pitch adjustment. The control structure is simple, but it can meet the stability of output power while alleviating the unbalanced load of the wind wheel;

[0039] 4) When the influence of axial unbalanced aerodynamic force is large, independent variable pitch control based on SVM fuzzy weight coefficient is adopted, and SVM algorithm based on incremental online learning is used to realize adaptive adjustment of dynamic weight coefficient; it has the advantages of good stability and high robustness;

[0040] 5) Compared with neural networks, the support vector machine (SVM) used in the present invention can effectively solve the problem of building a high-dimensional data model under limited sample conditions, and has the advantages of strong generalization ability, convergence to the global optimum, and dimension insensitivity. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is the partition diagram of independent pitch control;

[0042] Figure 2 This is the independent pitch partition control flow chart.

[0043] Figure 3 It is a schematic diagram of the weight coefficient allocation model based on azimuth;

[0044] Figure 4 It is a fuzzy weight coefficient allocation model based on SVM; DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0046] Example

[0047] like Figure 2 As shown, this embodiment provides a wind turbine independent pitch control method based on SVM partition weight coefficient allocation, the method is:

[0048] Independent pitch control partitioning is performed according to the blade azimuth angle θ;

[0049] When the angle between the blade azimuth angle θ and the horizontal plane is less than the preset angle, independent pitch control based on azimuth weight coefficient allocation is used for independent pitch adjustment; otherwise, independent pitch control based on SVM fuzzy weight coefficient is used for independent pitch adjustment.

[0050] The independent pitch control based on azimuth weight coefficient allocation is specifically as follows:

[0051] First, the rated power given value is subtracted from the output power as the input of the power controller, and the power controller outputs a uniform pitch angle Δβ;

[0052] Then, the uniform pitch angle Δβ is input into the weight coefficient distributor to redistribute the pitch angle change Δβ of each blade i (i=1,2,3), thereby completing independent pitch adjustment.

[0053] Among them, the pitch angle change of each blade Δβ i The expression is: Δβ i =K i ×Δβ,K i (i=1,2,3) is the azimuth weight coefficient.

[0054] The azimuth weight coefficient K i The expression is:

[0055]

[0056] Wherein, n is a set positive integer; R is the blade length, H0 is the height from the ground; θ is the blade azimuth angle.

[0057] The n is set to 2.

[0058] The independent pitch control based on SVM fuzzy weight coefficient is specifically as follows:

[0059] Fuzzy power control module: The difference between the rated power given value and the output power is used as the input of the fuzzy power controller, and the fuzzy power controller outputs a uniform pitch angle Δβ;

[0060] SVM weight coefficient independent variable pitch control module: the azimuth weight coefficient K output by the weight coefficient distribution unit i (i=1,2,3) and the azimuth weight coefficient compensation value ΔK output by the SVM optimization compensation unit i (i=1,2,3) is input to the pitch angle dynamic coefficient adjustment calculation unit to obtain the dynamic weighting coefficient δ i (i=1,2,3); and the pitch angle variation of each blade Δβ is calculated by combining the uniform pitch angle Δβ output by the fuzzy power controller i , and then complete independent pitch adjustment.

[0061] The specific implementation is as follows:

[0062] 1. Pitch control related characteristics

[0063] 1.1 Analysis of wind speed characteristics

[0064] The wind shear effect refers to the phenomenon that due to the presence of rough elements on the ground, a certain amount of gas kinetic energy is consumed when air flows through, causing the wind speed to be reduced to a certain extent. Its effect decreases as the height from the ground increases.

[0065] Assuming the ground is used as the reference surface, the general formula for wind shear effect is:

[0066]

[0067] Where: V H V is the wind speed at the height H above the ground; o Height from the ground H o The wind speed at the location; n is the wind shear coefficient.

[0068] 1.2 Aerodynamic analysis

[0069] Assuming that the wind rotor rotates at an angular velocity ω, the force dF acting on the blade element dm can be decomposed into the force dF along the tangent direction of the wind rotor rotation: n and the force dF along the rotor axis t , respectively as follows:

[0070]

[0071] Where: ρ is air density; V is wind speed; dS is blade area of ​​blade element; C L , C D are the blade drag coefficient and the blade lift coefficient respectively.

[0072] 1.3 Average effective wind speed model of blades

[0073] Assuming that the average wind speed point on each blade is located at the center of the blade, the average effective wind speed model for each blade is:

[0074]

[0075] Where: i is the blade number, i = 1, 2, 3; R is the blade length; θ is the azimuth angle of blade i, v o is the wind speed at the center of the wind rotor.

[0076] 2. Principle of independent pitch partitioning

[0077] Under the effect of wind shear, the change of wind speed at the blade tip with different time and height is analyzed to obtain the independent variable pitch control partition diagram, such as Figure 1 shown.

[0078] The angle range of zone I is 30°~150°, the angle range of zone II is 150°~210°, the angle range of zone III is 210°~330°, and the angle range of zone IV is 330°~396°. When the blades are in zones I and III, the axial unbalanced aerodynamic force borne by the blades and the wind rotor of the wind turbine is large, and the independent variable pitch control based on the SVM fuzzy weight coefficient considering the axial aerodynamic force of the blade is adopted; when the blades are in zones II and IV, the axial unbalanced aerodynamic force borne by the blades and the wind rotor of the wind turbine is small, and the independent variable pitch control based on the azimuth weight system allocation is adopted.

[0079] 3. Two control methods based on partition weight coefficient allocation

[0080] 3.1 Weight coefficient allocation model based on azimuth

[0081] like Figure 3 As shown in the figure, the rated power given value is subtracted from the output power as the input of the power controller, and the output of the power controller is a uniform pitch angle, θ1, θ2, and θ3 are the azimuth angles of the three blades. When the wind turbine is operating at a wind speed higher than the rated speed, in order to stabilize the output power, the power controller first generates a uniform pitch angle Δβ, and then the weight coefficient distributor redistributes the pitch angle change Δβ of each blade. i (i=1, 2, 3), thereby completing independent pitch change.

[0082] The pitch angle changes of each blade are as follows:

[0083]

[0084] Where: K i is the weight coefficient based on the azimuth angle.

[0085] The weight coefficient allocation model is as follows:

[0086]

[0087] Here n is 2. When n=2, a stable fan output power can be obtained.

[0088] 3.2 Fuzzy weight coefficient allocation model based on SVM

[0089] like Figure 4 As shown in the figure, the SVM fuzzy weight coefficient allocation model is divided into two parts. The first part is the power controller, which uses a fuzzy controller to control the system. The second part is the SVM weight coefficient independent variable pitch controller, which includes a weight coefficient allocation unit, an SVM optimization compensation and a weight coefficient adjustment operation unit. The weight coefficient allocation unit obtains the weight coefficient of the optimized design. The SVM optimization compensation combines the blade azimuth angle and the blade axial aerodynamic force to form a training sample of the SVM fuzzy weight coefficient controller for training. The SVM weight coefficient controller uses the support vector machine learning algorithm to learn the sample data offline according to the objective function to obtain the structure and preliminary parameters of the controller. Then, online learning correction is performed, and the weight coefficient adjustment operation unit explores and synthesizes the independent variable pitch control output according to the output of the first two parts.

[0090] An incremental online learning algorithm is used, and the KKT condition is used as the basis. When new samples are added, the structure and parameters of the SVM pitch controller are updated by modifying the Lagrange multiplier and the bias value. The following is its implementation process:

[0091] Step S1, obtaining a new blade azimuth angle θ i (i=1,2,3) and blade axial aerodynamic force F i (i=1,2,3), and use it as a new training sample;

[0092] Step S2: determine whether the training sample is valid according to the set evaluation period. If it is valid, add it to the training sample set; otherwise, discard the data and go to step S1;

[0093] Step S3, according to the KKT condition and the Lagrange multiplier, obtain the boundary support vector set B, the error support vector set E and the retained sample set R;

[0094] Step S4: The SVM optimization compensation unit outputs the azimuth weight coefficient compensation value ΔK i (i=1,2,3);

[0095] Step S5: Calculate the azimuth weight coefficient compensation value ΔK i (i=1,2,3) output as the reference value, and the dynamic weight coefficient δ obtained by inputting the fuzzy power controller, weight coefficient allocation unit and SVM optimization compensation unit i =(i=1,2,3) Generate pitch angle reference value Δβ i (i=1,2,3), control the variable pitch object;

[0096] Step S6: Use a decreasing algorithm to delete the samples that need to be forgotten in the training sample set to obtain a new support vector set BD, an error support vector set ED, and a retained sample set RD;

[0097] Step S7, wait for data to be updated online, and go to step S1.

[0098] The above algorithm selectively adds new data to the support vector machine training set, and adjusts the SVM fuzzy weight coefficient controller parameters online by learning the samples added in each iteration to achieve continuous optimization of the system. The online learning method can use the results of the previous operation to reduce the amount of calculation in the online training process while ensuring the nonlinear approximation ability of the online support vector. This can not only shorten the training time, but also improve the adaptability of the controller.

[0099] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A wind turbine independent pitch control method based on SVM partition weight coefficient allocation, characterized in that: The method is: Independent pitch control partitioning is performed according to the blade azimuth angle θ; the preset angle is 30°; When the angle between the blade azimuth angle θ and the horizontal plane is less than the preset angle, independent pitch control based on azimuth weight coefficient allocation is used for independent pitch adjustment; otherwise, independent pitch control based on SVM fuzzy weight coefficient is used for independent pitch adjustment. The independent pitch control based on SVM fuzzy weight coefficient is specifically as follows: Fuzzy power control process: The rated power given value is subtracted from the output power as the input of the fuzzy power controller, and the fuzzy power controller outputs a uniform pitch angle Δβ; SVM weight coefficient independent pitch control process: The azimuth weight coefficient K output by the weight coefficient distribution unit is i (i=1,2,3) and the azimuth weight coefficient compensation value ΔK output by the SVM optimization compensation unit i (i=1,2,3) is input to the pitch angle dynamic coefficient adjustment calculation unit to obtain the dynamic weighting coefficient δ i (i=1,2,3); and the pitch angle variation of each blade Δβ is calculated by combining the uniform pitch angle Δβ output by the fuzzy power controller i , and then complete the independent pitch adjustment; the SVM weight coefficient independent pitch control process specifically includes the following steps: Step S1, obtaining a new blade azimuth angle θ i (i=1,2,3) and blade axial aerodynamic force F i (i=1,2,3), and use it as a new training sample; Step S2: determine whether the training sample is valid according to the set evaluation period. If it is valid, add it to the training sample set; otherwise, discard the data and go to step S1; Step S3, according to the KKT condition and the Lagrange multiplier, obtain the boundary support vector set B, the error support vector set E and the retained sample set R; Step S4: The SVM optimization compensation unit outputs the azimuth weight coefficient compensation value ΔK i (i=1,2,3); Step S5: Calculate the azimuth weight coefficient compensation value ΔK i (i=1,2,3) output as the reference value, and the dynamic weight coefficient δ obtained by inputting the fuzzy power controller, weight coefficient allocation unit and SVM optimization compensation unit i =(i=1,2,3) Generate pitch angle reference value Δβ i (i=1,2,3), control the variable pitch object; Step S6: Use a decreasing algorithm to delete the samples that need to be forgotten in the training sample set to obtain a new support vector set BD, an error support vector set ED, and a retained sample set RD; Step S7, wait for data to be updated online, and go to step S1.

2. The wind turbine independent pitch control method based on SVM partition weight coefficient allocation according to claim 1 is characterized in that: The independent pitch control based on azimuth weight coefficient allocation is specifically as follows: First, the rated power given value is subtracted from the output power as the input of the power controller, and the power controller outputs a uniform pitch angle Δβ; Then, the uniform pitch angle Δβ is input into the weight coefficient distributor to redistribute the pitch angle change Δβ of each blade i (i=1,2,3), and then complete independent pitch adjustment; Among them, the pitch angle change of each blade Δβ i The expression is: Δβ i =K i ×Δβ,K i (i=1,2,3) is the azimuth weight coefficient.

3. The wind turbine independent pitch control method based on SVM partition weight coefficient allocation according to claim 2 is characterized in that: The azimuth weight coefficient K i The expression is: Wherein, n is a set positive integer; R is the blade length, H0 is the height from the ground; θ is the blade azimuth angle.

4. The wind turbine independent pitch control method based on SVM partition weight coefficient allocation according to claim 3 is characterized in that: The n is set to 2.

5. The wind turbine independent pitch control method based on SVM partition weight coefficient allocation according to claim 1 is characterized in that: The dynamic weighting coefficient δ i (i=1,2,3) the expression is: Among them, K i =(i=1,2,3) is the azimuth weight coefficient, ΔK i =(i=1,2,3) is the azimuth angle weight coefficient compensation value.

6. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the program, the method according to any one of claims 1 to 5 is implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images