Design method for low-noise blade of wind turbine
Through the combination of neural network model and sparrow algorithm, the sawtooth structure on the blades of the wind turbine is optimized, which solves the problem of insufficient noise reduction of the wind turbine, and achieves an efficient design process and optimized noise effect.
Patent Information
- Application Number
- CN202411572376.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-06
- Publication Date
- 2025-02-11
AI Technical Summary
The existing wind turbine blade design has challenges in reducing noise, resulting in less noise reduction and a long design process, which may not be able to achieve the optimal solution.
The neural network model is used instead of CFD software, and the structural parameters of the serrated structure are iteratively optimized through the sparrow algorithm to design the optimal serrated structure to achieve low noise effect.
It significantly reduces the noise of the wind turbine blades, and greatly reduces the time consumption during the design process, ensuring the optimization of the design plan.
Smart Images
Figure CN119416703A_ABST
Abstract
Description
Technical Field
[0001] The present invention generally relates to the technical field of wind turbines, and in particular to a method for designing low-noise blades for wind turbines. Background Art
[0002] There are many factors that need to be considered in the design process of existing wind turbine blades, but most of them are optimized to increase torque and reduce thrust. However, the noise problem of wind turbines is equally important. If the noise problem is not considered during the blade design process, it will affect the quality of life of surrounding residents and the living environment of wild animals.
[0003] To solve the noise problem when wind turbines are working, people have come up with the idea of the feather arrangement of owl wings in bionics. The feathers on the trailing edge of an owl's wings have special edges that force the air on both sides of the wing to vibrate half a cycle apart, that is, in opposite directions. Then, after the wing passes, the air on both sides mixes in the gap between the two feathers on the trailing edge of the wing, and the amplitudes cancel each other out, thereby reducing the loudness of the surrounding air flow.
[0004] However, to achieve the above effect, it is not enough to just install a structure similar to the feather arrangement of an owl's wing on the wind turbine blade. A lot of experiments are required to repeatedly test the various parameters of the sawtooth structure. In the specific design process, it is necessary to design the structure and parameters in advance, and then input CFD (Computational Fluid Dynamics, fluid dynamics simulation software) for simulation to test whether the various aerodynamic performances of the wind turbine meet the low noise requirements under the current structural design and parameter settings.
[0005] However, the CFD software simulation process is cumbersome, and the simulation calculation process is computationally intensive and time-consuming. When designing low-noise wind turbine blades, it is necessary to repeatedly adjust the structural design and parameter settings to achieve the ideal low-noise effect. This results in a long total time consumption for the blade design process using CFD software, and in most cases, the final structural design is likely not the optimal solution based on time cost considerations. This results in the installation of wind turbines designed and produced in this way with no significant noise reduction. Summary of the invention
[0006] In view of the above-mentioned defects or deficiencies in the prior art, it is desired to provide a method for designing low-noise blades for a wind turbine.
[0007] The present invention provides a method for designing a low-noise blade for a wind turbine, comprising: One end of the blade is connected to the rotating shaft of the wind turbine, and a sawtooth structure is installed on the leeward side; Methods include: Acquire a neural network model; the neural network model is used to input the structural parameters of the sawtooth structure to predict the aerodynamic performance data of the wind turbine; the structural parameters include: the length, width, thickness, tooth spacing of the sawtooth structure along the extension direction of the blade and the distance from one end of the sawtooth structure away from the rotation axis to the rotation axis; the aerodynamic performance data include: the torque and thrust of the wind turbine and the average sound pressure of the blade; constructing a fitness function according to the aerodynamic performance data; Obtaining the total length of the blade to be optimized; Setting the search range of the structural parameters and the iteration parameters of the sparrow algorithm; According to the total length of the blade to be optimized, the search range, the iteration parameters, the neural network model and the fitness function, the structural parameters are iteratively optimized using the sparrow algorithm to obtain the optimal structural parameters; The sawtooth structure is designed according to the optimal structural parameters.
[0008] According to the technical solution provided by the present invention, the steps of iteratively optimizing the structural parameters using the sparrow algorithm according to the total length of the blade to be optimized, the search range, the iteration parameters, the neural network model and the fitness function, and obtaining the optimal structural parameters include: S21: Calculate the i-th structural parameter corresponding to the sparrow in the sparrow algorithm according to the search range and the iteration parameter; the initial value of i is set to 1; S22: inputting the total length of the blade to be optimized and the i-th structural parameter into the neural network model to predict and obtain the i-th aerodynamic performance data; S23: Substituting the i-th aerodynamic performance data into the fitness function to calculate the i-th fitness; S24: According to the i-th fitness, use the sparrow algorithm to iteratively optimize the i-th structural parameter once to obtain the i+1-th structural parameter; S25: judging whether the end condition is reached according to the i-th fitness, and when the end condition is reached, proceeding to step S26; otherwise, adding 1 to the value of i, and repeating steps S22-S25; S26: taking the structural parameters corresponding to the maximum fitness in the multiple iterative optimization processes as the optimal structural parameters.
[0009] According to the technical solution provided by the present invention, the i-th aerodynamic performance data includes: the i-th predicted thrust, the i-th predicted torque and the i-th predicted sound pressure average value; The fitness function is shown in Formula 1: Formula 1; in, f i represents the i-th fitness,w 1. w 2. w 3 represents weight, T i represents the ith predicted thrust, N i represents the i-th predicted torque, Y i represents the average value of the predicted sound pressure of the ith T 0 represents the thrust of the wind turbine at rated wind speed, N 0 represents the torque of the wind turbine at rated wind speed, Y 0 represents the average sound pressure of the wind turbine blades at rated wind speed.
[0010] According to the technical solution provided by the present invention, the iteration parameters include: The maximum number of iterations for iterative optimization; The number of sparrows in the sparrow algorithm and the initial positions of multiple sparrows; The number of predators in the sparrow algorithm and the initial positions of multiple predators; The sparrows belonging to the discoverer and the sparrows belonging to the joiner in the sparrow algorithm; The first set ratio in the sparrow algorithm; the first set ratio represents the ratio of sparrows belonging to the discoverer among all sparrows.
[0011] According to the technical solution provided by the present invention, the initial positions of the plurality of sparrows are set as follows: The position of the equivalent structural parameter within the search range is taken as the mean and obeys the normal distribution; the equivalent structural parameter represents the structural parameter equivalent to the aerodynamic performance data of the wind turbine when the sawtooth structure is not installed and at the rated wind speed; The initial positions of the predators are set in such a manner that one of the predators is located at a position of equivalent structural parameters within the search range, and the remaining predators are evenly distributed within the search range; The termination condition includes: the maximum value among the multiple i-th fitness values is optimized for a set number of consecutive iterations and no longer increases, or the maximum number of iterations is reached.
[0012] According to the technical solution provided by the present invention, the first set ratio is initially set to a first set value; During the iterative optimization process, when the maximum value among the multiple i-th fitness values does not increase, the first setting ratio is reduced by the first setting value times the second setting value.
[0013] According to the technical solution provided by the present invention, the product of the second set value and the set number of times is equal to 1; When the number of times that the maximum value among the plurality of i-th fitness values does not increase does not reach the set number of times, the first set ratio is reset to the first set value.
[0014] According to the technical solution provided by the present invention, the step of iteratively optimizing the i-th structural parameter once using the sparrow algorithm according to the i-th fitness to obtain the i+1-th structural parameter includes: Get the specific value of the current first set ratio; The sparrows of the first set proportion with smaller fitness are re-considered as joiners, and the remaining sparrows are re-considered as discoverers; Using the positions corresponding to the plurality of i-th structural parameters within the search range as the current positions of the plurality of sparrows, to obtain a plurality of i-th positions; Obtain the current positions of multiple predators, and obtain multiple i-th predator positions; Calculate the distances from the multiple i-th positions to the multiple i-th predator positions to obtain the distance from each sparrow to each predator; The second set proportion of sparrows closest to each i-th predator position are used as sentinels; Updates the positions of multiple predators; Obtain the discoverer position update formula, joiner position update formula and sentinel position update formula; When the sparrow belongs to the discoverer, the i-th fitness and the i-th position corresponding to the sparrow are substituted into the discoverer position update formula to obtain the i+1-th positions corresponding to multiple discoverers; When a sparrow belongs to a joiner, the i-th fitness and the i-th position corresponding to the sparrow are substituted into the joiner position update formula to obtain the i+1-th positions corresponding to multiple joiners; When the sparrow belongs to a sentinel, the i-th fitness and the i-th position corresponding to the sparrow are substituted into the sentinel position update formula to obtain the i+1-th positions corresponding to multiple sentinels; The multiple i+1th structural parameters are calculated according to the i+1th positions corresponding to the multiple discoverers, the i+1th positions corresponding to the multiple joiners, and the i+1th positions corresponding to the multiple alerters.
[0015] According to the technical solution provided by the present invention, the method of updating the position of the predator is: Move to the i-th position corresponding to the sentinel closest to the predator.
[0016] According to the technical solution provided by the present invention, the step of obtaining the neural network model includes: Obtain multiple groups of initial structural parameters of the sawtooth structure; Obtain the total length of various blades; Inputting the various total lengths of the blades and the multiple groups of initial structural parameters into CFD software respectively, and calculating multiple groups of initial aerodynamic performance data of the wind turbine; Get the initial neural network model; Taking the total length of the blades and the initial aerodynamic performance data as input, the multiple groups of initial structural parameters as label data, and the length, width, thickness, tooth spacing of the sawtooth structure along the extension direction of the blade and the distance from one end of the sawtooth structure away from the rotation axis to the rotation axis as output, the initial neural network model is pre-trained to obtain a pre-trained neural network model; With multiple total lengths of the blades and multiple groups of initial structural parameters as input, the corresponding multiple groups of initial aerodynamic performance data as label data, the torque, thrust of the wind turbine and the average sound pressure of the blades as output, the pre-trained neural network model is trained again to obtain the neural network model.
[0017] The beneficial effects of the present invention are: A sawtooth structure similar to the arrangement of owl wing feathers is installed on the wind turbine blade. During the rotation of the wind turbine, the vibration of the air on both sides of the blade differs by half a cycle, so that the amplitudes cancel each other out and achieve a noise reduction effect. When designing the sawtooth structure, a neural network model is used to replace the CFD software. The neural network model is trained using existing data, and the structural parameters of the sawtooth structure are input to predict the aerodynamic performance data of the wind turbine. In the process of blade design, the sparrow algorithm is used to optimize the structural parameters. Each iteration of the optimization inputs the obtained structural parameters into the neural network model, so that the corresponding aerodynamic performance data can be obtained more quickly. The sparrow algorithm can optimize the structural parameters according to the fitness function constructed by the aerodynamic performance data. The optimal structural parameters obtained are the solution with the lowest noise that can be designed under the current circumstances. This method greatly reduces the time consumption in the design process while ensuring that the final design solution can significantly reduce the blade noise. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings: Figure 1 It is a flow chart of a method for designing low-noise blades for wind turbines; Figure 2 A top view of a wind turbine blade; Figure 3 is the front view of the wind turbine blade; Among them: 1. Blades; 2. Rotating axis; 3. Sawtooth structure. DETAILED DESCRIPTION
[0019] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant inventions, rather than to limit the invention. It should also be noted that, for ease of description, only the parts related to the invention are shown in the accompanying drawings.
[0020] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.
[0021] Please refer to Figure 1 The present invention provides a method for designing a low-noise blade for a wind turbine, wherein one end of a blade 1 is connected to a rotating shaft 2 of the wind turbine, and a sawtooth structure 3 is installed on the leeward side; Methods include: S1: obtaining a neural network model; the neural network model is used to input the structural parameters of the sawtooth structure 3 to predict the aerodynamic performance data of the wind turbine; S2: constructing a fitness function according to the aerodynamic performance data; S3: Obtaining the total length of the blade to be optimized; S4: setting the search range of the structural parameters and the iteration parameters of the sparrow algorithm; S5: according to the total length of the blade to be optimized, the search range, the iteration parameters, the neural network model and the fitness function, the sparrow algorithm is used to iteratively optimize the structural parameters to obtain the optimal structural parameters; S6: Design the sawtooth structure 3 according to the optimal structural parameters.
[0022] Specifically, when blade 1 rotates clockwise, Figure 2 The upper side is the windward side, and the lower side is the leeward side. In this embodiment, a sawtooth structure 3 is installed on the leeward side. The sawtooth structure 3 includes a plurality of feather-like elliptical parts, forming a tooth shape, and the distance between the midlines of the plurality of elliptical parts is the tooth spacing d.
[0023] The structural parameters of the sawtooth structure 3 include: the length, width, thickness, tooth spacing of the sawtooth structure 3 along the extension direction of the blade 1 and the distance from one end of the sawtooth structure 3 away from the rotating shaft 2 to the rotating shaft 2 .
[0024] The aerodynamic performance data include: the torque and thrust of the wind turbine and the average sound pressure of the blade 1 .
[0025] exist Figure 2 and Figure 3In the figure, the length of the sawtooth structure 3 is a%*L, which means that in this embodiment, the total length L of blades to be optimized can be adapted to different sizes of blades to be optimized by calculating the ratio a% of the length of the sawtooth structure 3 to the total length L of the blade to be optimized. The width of the sawtooth structure 3 is w, the thickness of the sawtooth structure 3 is h, and the tooth spacing is d; The distance from the end of the sawtooth structure 3 away from the rotation axis 2 to the rotation axis 2 is b%*L, which means that in this embodiment, the calculation is performed according to the ratio b% of the distance from the end of the sawtooth structure 3 away from the rotation axis 2 to the rotation axis 2 to the total length L of the blade to be optimized.
[0026] Specifically, a sawtooth structure 3 with a shape similar to the arrangement of feathers on an owl's wing is installed on the wind turbine blade. During the rotation of the wind turbine, the vibration of the air on both sides of the blade differs by half a cycle, so that the amplitudes cancel each other out and achieve a noise reduction effect.
[0027] When designing a sawtooth structure, a neural network model is used to replace CFD software. The neural network model is trained using existing data, and the structural parameters of the sawtooth structure are input to predict the aerodynamic performance data of the wind turbine.
[0028] In the process of designing the structural parameters of the sawtooth structure 3, the sparrow algorithm is used to optimize the structural parameters. Each time the optimization is iterated, the obtained structural parameters are input into the neural network model, so that the corresponding aerodynamic performance data can be obtained more quickly. The sparrow algorithm can optimize the structural parameters according to the fitness function constructed by the aerodynamic performance data, and the obtained optimal structural parameters are the solution with the least noise that can be designed under the current situation.
[0029] This approach greatly reduces the time spent in the design process while ensuring that the final design can significantly reduce blade noise.
[0030] Further, according to the total length of the blade to be optimized, the search range, the iteration parameter, the neural network model and the fitness function, the sparrow algorithm is used to iteratively optimize the structural parameters to obtain the optimal structural parameters, which includes: S21: Calculate the i-th structural parameter corresponding to the sparrow in the sparrow algorithm according to the search range and the iteration parameter; the initial value of i is set to 1; S22: inputting the total length of the blade to be optimized and the i-th structural parameter into the neural network model to predict and obtain the i-th aerodynamic performance data; S23: Substituting the i-th aerodynamic performance data into the fitness function to calculate the i-th fitness; S24: According to the i-th fitness, use the sparrow algorithm to iteratively optimize the i-th structural parameter once to obtain the i+1-th structural parameter; S25: judging whether the end condition is reached according to the i-th fitness, and when the end condition is reached, proceeding to step S26; otherwise, adding 1 to the value of i, and repeating steps S22-S25; S26: taking the structural parameters corresponding to the maximum fitness in the multiple iterative optimization processes as the optimal structural parameters.
[0031] Specifically, the search scope includes: The ratio of the length of the sawtooth structure 3 to the total length L of the blade to be optimized is a%∈[1%,30%]; The width of the sawtooth structure 3 is w∈[1,10], in millimeters; The thickness of the sawtooth structure 3 is h∈[2,8], in millimeters; Tooth spacing d∈[2,10], unit: mm; The ratio b%∈[50%, 100%] of the distance from one end of the sawtooth structure 3 away from the rotation axis 2 to the rotation axis 2 to the total length L of the blade to be optimized.
[0032] Further, the i-th aerodynamic performance data includes: the i-th predicted thrust, the i-th predicted torque and the i-th predicted sound pressure average value; The fitness function is shown in Formula 1: Formula 1; in, f i represents the i-th fitness, w 1. w 2. w 3 represents weight, T i represents the ith predicted thrust, N i represents the i-th predicted torque, Y i represents the average value of the predicted sound pressure of the ith T 0 represents the thrust of the wind turbine at rated wind speed, N 0 represents the torque of the wind turbine at rated wind speed, Y 0 represents the average sound pressure of wind turbine blade 1 at rated wind speed.
[0033] Specifically, in this embodiment, the greater the fitness, the closer it is to the ideal solution. The sum of multiple weights is equal to 100%; the specific weight value can be adjusted according to actual design requirements. The thrust, torque and sound pressure average values of wind turbines of the same model at rated wind speed are all known fixed values.
[0034] In this embodiment, the setting method of the fitness function can amplify the amplitude of the sound pressure value reduction, making it easier to screen out solutions for reducing the sound pressure value. w 1.w 2. w 3 are set to 20%, 10%, and 70% respectively; this method can screen out solutions with greater sound pressure reduction. At the same time, the constraints ensure that the screened solutions must be solutions with lower sound pressure values and greater torque, and the fitness of the remaining solutions is set to negative infinity and discarded during the algorithm iteration process.
[0035] The above-mentioned fitness function setting method can filter out interference solutions, which is more conducive to the algorithm finding the optimal solution.
[0036] Furthermore, the iteration parameters include: The maximum number of iterations for iterative optimization; The number of sparrows in the sparrow algorithm and the initial positions of multiple sparrows; The number of predators in the sparrow algorithm and the initial positions of multiple predators; The sparrows belonging to the discoverer and the sparrows belonging to the joiner in the sparrow algorithm; The first set ratio in the sparrow algorithm; the first set ratio represents the ratio of sparrows belonging to the discoverer among all sparrows.
[0037] In this embodiment, the maximum number of iterations is set to 200, the number of sparrows is set to 100, and the number of predators is set to 4-8.
[0038] Furthermore, the initial positions of the plurality of sparrows are set as follows: The position of the equivalent structural parameter within the search range is taken as the mean and is subject to normal distribution; The initial positions of the predators are set in such a manner that one of the predators is located at a position of equivalent structural parameters within the search range, and the remaining predators are evenly distributed within the search range; The termination condition includes: the maximum value among the multiple i-th fitness values is iterated and optimized for a set number of consecutive times (10 times in this embodiment) and no longer increases, or the maximum number of iterations is reached.
[0039] Specifically, the equivalent structural parameters represent structural parameters equivalent to the aerodynamic performance data of the wind turbine when the sawtooth structure 3 is not installed and at the rated wind speed.
[0040] During the optimization process, it is necessary to ensure that the adaptability after the sawtooth structure 3 is installed is higher than before the installation. In order to ensure that the above conditions are met after optimization, the concept of the present invention is: First, an equivalent situation is found, that is, the fitness is equal when the sawtooth structure 3 is not installed and when the sawtooth structure 3 is installed. Then, further optimization is performed on this basis. This method can reduce the probability of the algorithm fully searching the search range, reduce the time consumed by the algorithm exploration, and improve the convergence rate of the algorithm, and finally reduce the time consumption during the optimization design.
[0041] The step of obtaining equivalent structural parameters includes: inputting the total length of the blade to be optimized and the aerodynamic performance data of the wind turbine at the rated wind speed into the neural network model to obtain the equivalent corresponding structural parameters.
[0042] Specifically, in the prior art, predators are randomly distributed within the search range; affected by the randomness, the algorithm convergence rate fluctuates greatly.
[0043] Compared with the random distribution, the setting method of this embodiment can prevent sparrows from gathering at the equivalent structural parameter position at the first time, thereby avoiding falling into a local optimal solution in the early stage of algorithm iteration.
[0044] Therefore, by combining these two methods, in most cases the algorithm converges faster than random distribution, and the convergence rate is less volatile and more stable.
[0045] Further, the first set ratio is initially set to a first set value; During the iterative optimization process, when the maximum value among the multiple i-th fitness values does not increase, the first setting ratio is reduced by the first setting value times the second setting value.
[0046] Further, the product of the second set value and the set number of times is equal to 1; When the number of times that the maximum value among the plurality of i-th fitness values does not increase does not reach the set number of times, the first set ratio is reset to the first set value.
[0047] Specifically, the first setting value is 80% and the second setting value is 0.1.
[0048] This method allows the sparrow, acting as the discoverer, to move more quickly to the position with the highest current fitness. Once the algorithm finds the global optimal solution, it can converge quickly. At the same time, in order to avoid quickly converging to the local optimal solution, a method for resetting the first set ratio is also set. When a position with higher fitness is found, the algorithm can restore its search capability and avoid falling into the local optimal solution again.
[0049] Among them, the number of all discoverers is rounded, and the rest are joiners. There is at least one discoverer in the entire iterative optimization process of the algorithm.
[0050] Specifically, the product of the second set value and the set number of times is equal to 1, so that the maximum value of the fitness does not increase for several iterations. However, when it does not increase for 10 consecutive times, the first set ratio is prevented from being reduced to 0, causing an algorithm error.
[0051] Further, according to the i-th fitness, the step of iteratively optimizing the i-th structural parameter once using the sparrow algorithm to obtain the i+1-th structural parameter includes: Get the specific value of the current first set ratio; The sparrows of the first set proportion with smaller fitness are re-considered as joiners, and the remaining sparrows are re-considered as discoverers; Using the positions corresponding to the plurality of i-th structural parameters within the search range as the current positions of the plurality of sparrows, to obtain a plurality of i-th positions; Obtain the current positions of multiple predators, and obtain multiple i-th predator positions; Calculate the distances from the multiple i-th positions to the multiple i-th predator positions to obtain the distance from each sparrow to each predator; The second set proportion of sparrows closest to each i-th predator position are used as sentinels; Updates the positions of multiple predators; Obtain the discoverer position update formula, joiner position update formula and sentinel position update formula; When the sparrow belongs to the discoverer, the i-th fitness and the i-th position corresponding to the sparrow are substituted into the discoverer position update formula to obtain the i+1-th positions corresponding to multiple discoverers; When a sparrow belongs to a joiner, the i-th fitness and the i-th position corresponding to the sparrow are substituted into the joiner position update formula to obtain the i+1-th positions corresponding to multiple joiners; When the sparrow belongs to a sentinel, the i-th fitness and the i-th position corresponding to the sparrow are substituted into the sentinel position update formula to obtain the i+1-th positions corresponding to multiple sentinels; The multiple i+1th structural parameters are calculated according to the i+1th positions corresponding to the multiple discoverers, the i+1th positions corresponding to the multiple joiners, and the i+1th positions corresponding to the multiple alerters.
[0052] Furthermore, the position of the predator is updated as follows: Move to the i-th position corresponding to the sentinel closest to the predator.
[0053] Specifically, the second set ratio can be set to any value between 10% and 20%; then multiplied by the number of sparrows and rounded to the integer, the number of sentinels is obtained.
[0054] When a sparrow is within the second set ratio of the distance from multiple predators, it also acts as a sentinel, the same as other sentinels.
[0055] In this embodiment, the second set ratio is set to 10%. Therefore, the method of using the second set ratio of sparrows closest to one of the i-th predators as guards is as follows: Calculate the distance from all sparrows to the predator, and use the 10% of sparrows with the smallest distance as sentinels. Since the number of sparrows is 100, the 10 sparrows closest to the predator will eventually be used as sentinels.
[0056] Specifically, the distance from the sparrow to the predator is calculated as follows: The squares of the differences of all types of parameters in the structural parameters corresponding to the current positions of the sparrow and the predator are calculated respectively; then the square of the distance is equal to the sum of the squares of the differences of all types of parameters; finally, the distance from the sparrow to the predator can be calculated.
[0057] In some embodiments, the predator is used to drive away the sparrows. When the sparrows are located at a local optimal solution, it is difficult to jump out of the local optimal solution without the predator. Therefore, setting the predator can enable the algorithm to have the ability to jump out of the local optimal solution and increase the probability of finding the global optimal solution.
[0058] In this embodiment, the movement mode of the predator is set to move to the i-th position corresponding to the sentinel closest to the predator, so that when multiple sparrows gather together, the predator can move to the vicinity of the gathering position to drive away the gathered sparrows.
[0059] If the clustering location is the location of the local optimal solution, this method can quickly prompt the algorithm to jump out of the location of the local optimal solution.
[0060] In some implementations, the predator may be removed when the number of iterations exceeds 0.8 times the maximum number of iterations. This approach can prevent predators from gathering around the global optimal solution and hindering the sparrows from moving to the location of the global optimal solution.
[0061] Specifically, the discoverer location update formula is expressed by Formula 2: Formula 2: in, represents the position of the jth sparrow in the gth dimension in the i+1th iteration after the update, represents the position of the j-th sparrow in the g-th dimension in the i-th iteration before the update; e represents the natural logarithm, a0 represents a random number in the interval (0,1], and the value of a0 is randomly selected before each position update. Dmax represents the maximum number of iterations, Q represents a random number that follows a normal distribution. Represents a matrix with 1 row and 2 columns, where all elements are 1, such as [1,1]; R 2 represents the warning value, which is in the interval [0,1] and is randomly assigned according to the distances from multiple predators. The larger the sum of the distances, the smaller the warning value. S T It means that the safety value is in the interval [0.5,1] and the safety value is a constant in one iteration.
[0062] In this embodiment, i and j are both natural numbers; i represents the number of iterations, with an initial value of 1 and an increment of 1 for each iteration; j represents the number of sparrows, with a minimum value of 1 and a maximum value equal to the number S of sparrows.
[0063] The dimension g is equal to 5, which represents the parameters in the structural parameters in turn, so to When combined into coordinates, they correspond to the i+1th position corresponding to the jth sparrow within the search range.
[0064] Specifically, the joiner position update formula is expressed by Formula 3: Formula 3; in, Indicates the position of the sparrow with the lowest fitness among all the sparrows in the previous i iterations. Indicates the position of the sparrow with the highest fitness among all the sparrows in the i+1th iteration. represents a 1-row, 2-column matrix where elements are randomly assigned values of 1 or -1, for example , S Indicates the number of sparrows.
[0065] Specifically, the sentinel position update formula is expressed by Formula 4: Formula 4; in, Indicates the position of the sparrow with the highest fitness among all the sparrows in the previous i iterations; b0 represents the step size control parameter, which is a random number that follows a normal distribution with a mean of 0 and a variance of 1; K Represents a random number in the interval [-1,1]; represents the fitness of the jth sparrow in the i-th iteration, f g represents the maximum fitness during the previous i iterations, f w represents the minimum fitness during the previous i iterations, is a constant to avoid the denominator being equal to zero.
[0066] Specifically, the maximum number of iterations is set to 200.
[0067] When the maximum value of the fitness corresponding to multiple sparrows does not increase for 10 consecutive iterations, it means that the sparrow algorithm has found the optimal structural parameters, and the iteration can be terminated. This method can save the total amount of calculation and optimize the optimal structural parameters more quickly.
[0068] In some cases, if the Sparrow Algorithm has not reached the maximum value of the multiple i-th fitness values for 200 iterations and has not increased for many consecutive times, it means that the optimal solution currently being sought is very close to the global optimal solution and the error between them can be ignored; at this time, the iteration should be terminated in time to save computing time.
[0069] Furthermore, the step of obtaining the neural network model includes: Obtain multiple groups of initial structural parameters of the sawtooth structure 3; Obtain the total length of various blades; Inputting the various total lengths of the blades and the multiple groups of initial structural parameters into CFD software respectively, and calculating multiple groups of initial aerodynamic performance data of the wind turbine; Get the initial neural network model; Taking the total length of the blades and the initial aerodynamic performance data as input, the multiple groups of initial structural parameters as label data, and the length, width, thickness, tooth spacing of the sawtooth structure 3 along the extension direction of the blade 1 and the distance from the end of the sawtooth structure 3 away from the rotation axis 2 to the rotation axis 2 as output, the initial neural network model is pre-trained to obtain a pre-trained neural network model; With multiple total blade lengths and multiple groups of initial structural parameters as input, the corresponding multiple groups of initial aerodynamic performance data as label data, the torque, thrust and average sound pressure of the wind turbine 1 as output, the pre-trained neural network model is trained again to obtain the neural network model.
[0070] Specifically, the method of obtaining multiple groups of initial structural parameters of the sawtooth structure 3 is: randomly selecting multiple groups of structural parameters within the search range, and then inputting them into the CFD software for simulation together with the total length of the blade to obtain multiple groups of initial aerodynamic performance data.
[0071] The plurality of total blade lengths are selected from total blade lengths of all types of existing wind turbines.
[0072] In some cases, existing historical data can be directly used as initial structural parameters and initial aerodynamic performance data.
[0073] Specifically, the input dimension of the initial neural network model is 6, and the output dimension is 4.
[0074] The six input dimensions correspond to: the total length of the blade, the length, width, thickness, tooth spacing of the sawtooth structure 3 , and the distance from one end of the sawtooth structure 3 away from the rotation axis 2 to the rotation axis 2 .
[0075] The four output dimensions correspond to: total blade length, wind turbine torque, thrust, and average sound pressure of blade 1.
[0076] In the two processes of pre-training and re-training, in one round of training, the total length of the leaves as input and the total length of the leaves as output are the same. The output dimension adds an additional dimension of the total length of the leaves, which can not only facilitate the pre-training process of back propagation, but also predict the equivalent structural parameters.
[0077] When predicting equivalent structural parameters, the aerodynamic performance data of the wind turbine when the sawtooth structure is not installed and at the rated wind speed is reversely input into the four output dimensions of the neural network model, and then the total length of the blade and the equivalent corresponding structural parameters can be obtained in the six input dimensions of the neural network model.
[0078] The above description is only a preferred embodiment of the present invention and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present invention is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present invention (but not limited to) to form a technical solution.
Claims
1. A method for designing low-noise blades for a wind turbine, characterized in that: One end of the blade is connected to the rotating shaft of the wind turbine, and a sawtooth structure is installed on the leeward side; Methods include: Acquire a neural network model; the neural network model is used to input the structural parameters of the sawtooth structure to predict the aerodynamic performance data of the wind turbine; the structural parameters include: the length, width, thickness, tooth spacing of the sawtooth structure along the extension direction of the blade and the distance from one end of the sawtooth structure away from the rotation axis to the rotation axis; the aerodynamic performance data include: the torque and thrust of the wind turbine and the average sound pressure of the blade; constructing a fitness function according to the aerodynamic performance data; Obtaining the total length of the blade to be optimized; Setting the search range of the structural parameters and the iteration parameters of the sparrow algorithm; According to the total length of the blade to be optimized, the search range, the iteration parameters, the neural network model and the fitness function, the structural parameters are iteratively optimized using the sparrow algorithm to obtain the optimal structural parameters; The sawtooth structure is designed according to the optimal structural parameters.
2. A method for designing low-noise blades for a wind turbine according to claim 1, characterized in that: According to the total length of the blade to be optimized, the search range, the iteration parameters, the neural network model and the fitness function, the sparrow algorithm is used to iteratively optimize the structural parameters to obtain the optimal structural parameters, which comprises: S21: Calculate the i-th structural parameter corresponding to the sparrow in the sparrow algorithm according to the search range and the iteration parameter; the initial value of i is set to 1; S22: inputting the total length of the blade to be optimized and the i-th structural parameter into the neural network model to predict and obtain the i-th aerodynamic performance data; S23: Substituting the i-th aerodynamic performance data into the fitness function to calculate the i-th fitness; S24: According to the i-th fitness, use the sparrow algorithm to iteratively optimize the i-th structural parameter once to obtain the i+1-th structural parameter; S25: judging whether the end condition is reached according to the i-th fitness, and when the end condition is reached, proceeding to step S26; otherwise, adding 1 to the value of i, and repeating steps S22-S25; S26: taking the structural parameters corresponding to the maximum fitness in the multiple iterative optimization processes as the optimal structural parameters.
3. A method for designing low-noise blades for a wind turbine according to claim 2, characterized in that: The i-th aerodynamic performance data includes: the i-th predicted thrust, the i-th predicted torque and the i-th predicted sound pressure average value; The fitness function is shown in Formula 1: Formula 1; in, f i represents the i-th fitness, w 1. w 2. w 3 represents weight, T i represents the ith predicted thrust, N i represents the i-th predicted torque, Y i represents the average value of the predicted sound pressure of the ith T 0 represents the thrust of the wind turbine at rated wind speed, N 0 represents the torque of the wind turbine at rated wind speed, Y 0 represents the average sound pressure of the wind turbine blades at rated wind speed.
4. A method for designing low-noise blades for a wind turbine according to claim 2, characterized in that: The iteration parameters include: The maximum number of iterations for iterative optimization; The number of sparrows in the sparrow algorithm and the initial positions of multiple sparrows; The number of predators in the sparrow algorithm and the initial positions of multiple predators; The sparrows belonging to the discoverer and the sparrows belonging to the joiner in the sparrow algorithm; The first set ratio in the sparrow algorithm; the first set ratio represents the ratio of sparrows belonging to the discoverer among all sparrows.
5. A method for designing low-noise blades for a wind turbine according to claim 4, characterized in that: The initial positions of the plurality of sparrows are set as follows: The position of the equivalent structural parameter within the search range is taken as the mean and obeys the normal distribution; the equivalent structural parameter represents the structural parameter equivalent to the aerodynamic performance data of the wind turbine when the sawtooth structure is not installed and at the rated wind speed; The initial positions of the predators are set in such a manner that one of the predators is located at a position of equivalent structural parameters within the search range, and the remaining predators are evenly distributed within the search range; The termination condition includes: the maximum value among the multiple i-th fitness values is optimized for a set number of consecutive iterations and no longer increases, or the maximum number of iterations is reached.
6. A method for designing low-noise blades for a wind turbine according to claim 5, characterized in that: The first set ratio is initially set to a first set value; During the iterative optimization process, when the maximum value among the multiple i-th fitness values does not increase, the first setting ratio is reduced by the first setting value times the second setting value.
7. A method for designing low-noise blades for a wind turbine according to claim 6, characterized in that: The product of the second set value and the set number of times is equal to 1; When the number of times that the maximum value among the plurality of i-th fitness values does not increase does not reach the set number of times, the first set ratio is reset to the first set value.
8. A method for designing low-noise blades for a wind turbine according to any one of claims 4 to 7, characterized in that: According to the i-th fitness, the step of iteratively optimizing the i-th structural parameter once using the sparrow algorithm to obtain the i+1-th structural parameter comprises: Get the specific value of the current first set ratio; The sparrows of the first set proportion with smaller fitness are re-considered as joiners, and the remaining sparrows are re-considered as discoverers; Using the positions corresponding to the plurality of i-th structural parameters within the search range as the current positions of the plurality of sparrows, to obtain a plurality of i-th positions; Obtain the current positions of multiple predators, and obtain multiple i-th predator positions; Calculate the distances from the multiple i-th positions to the multiple i-th predator positions to obtain the distance from each sparrow to each predator; The second set proportion of sparrows closest to each i-th predator position are used as sentinels; Updates the positions of multiple predators; Obtain the discoverer position update formula, joiner position update formula and sentinel position update formula; When the sparrow belongs to the discoverer, the i-th fitness and the i-th position corresponding to the sparrow are substituted into the discoverer position update formula to obtain the i+1-th positions corresponding to multiple discoverers; When a sparrow belongs to a joiner, the i-th fitness and the i-th position corresponding to the sparrow are substituted into the joiner position update formula to obtain the i+1-th positions corresponding to multiple joiners; When the sparrow belongs to a sentinel, the i-th fitness and the i-th position corresponding to the sparrow are substituted into the sentinel position update formula to obtain the i+1-th positions corresponding to multiple sentinels; The multiple i+1th structural parameters are calculated according to the i+1th positions corresponding to the multiple discoverers, the i+1th positions corresponding to the multiple joiners, and the i+1th positions corresponding to the multiple alerters.
9. A method for designing low-noise blades for a wind turbine according to claim 8, characterized in that: The way to update the predator's position is: Move to the i-th position corresponding to the sentinel closest to the predator.
10. A method for designing low-noise blades for a wind turbine according to claim 1, characterized in that: The steps of obtaining the neural network model include: Obtain multiple groups of initial structural parameters of the sawtooth structure; Obtain the total length of various blades; Inputting the various total lengths of the blades and the multiple groups of initial structural parameters into CFD software respectively, and calculating multiple groups of initial aerodynamic performance data of the wind turbine; Get the initial neural network model; Taking the total length of the blades and the initial aerodynamic performance data as input, the multiple groups of initial structural parameters as label data, and the length, width, thickness, tooth spacing of the sawtooth structure along the extension direction of the blade and the distance from one end of the sawtooth structure away from the rotation axis to the rotation axis as output, the initial neural network model is pre-trained to obtain a pre-trained neural network model; With multiple total lengths of the blades and multiple groups of initial structural parameters as input, the corresponding multiple groups of initial aerodynamic performance data as label data, the torque, thrust of the wind turbine and the average sound pressure of the blades as output, the pre-trained neural network model is trained again to obtain the neural network model.