A wind farm micro-siting method based on improved JAYA algorithm

Abstract

This invention discloses a wind farm micro-site selection method based on an improved JAYA algorithm, belonging to the field of wind power technology. Existing technologies suffer from insufficient initial population randomness and a tendency to get trapped in local optima, limiting power generation improvements. This invention includes: generating an initial population based on a Logistic chaotic mapping to enhance population diversity and exploration capabilities; during the iteration process, a historical learning strategy is introduced, using historical population individuals to guide the current solution towards the global optimum; an immediate search strategy and a global search strategy are employed, where the immediate search combines the best, worst, and average solutions of each iteration, introducing upper and lower local attraction points; and the global search strategy incorporates historical population individuals to expand the search space. By optimizing the initial population generation and iteration process, this invention effectively reduces the risk of getting trapped in local optima, outputting optimized wind turbine layout locations, including coordinates and hub height, thereby significantly improving the overall power generation of the wind farm.

Description

Technical Field

[0001] This invention relates to the field of wind power generation technology, and specifically to a wind farm micro-situation method based on an improved JAYA algorithm. Background Technology

[0002] Offshore wind farms offer advantages such as high wind speeds, low turbulence, and minimal environmental impact, and are gradually expanding into deep-sea environments. In these environments, wind shear is significant; increasing hub height can capture higher wind speeds, but it also increases tower costs. Simultaneously, the wake effect between turbines reduces the incoming wind speed to downstream turbines, thus decreasing overall power generation. Therefore, during the micro-site selection phase, it is necessary to simultaneously optimize the horizontal arrangement (coordinates) of the turbines and the hub height to balance construction costs and power generation benefits.

[0003] To address the aforementioned issues, Chinese patent application CN202510936857.2 discloses a wind farm micro-situation optimization method based on a multi-objective genetic algorithm. This method employs a hybrid optimization framework combining a multi-objective algorithm and IHHO (Independent Hybrid Homogeneous Optimization). This method requires setting multiple control parameters, including genetic operators and IHHO parameters, resulting in an exceptionally complex algorithm structure. Furthermore, it utilizes a BP wake model, necessitating pre-training of the neural network, further increasing computational overhead. Each iteration also requires sorting all wind turbines and performing wake superposition calculations, leading to a large overall computational load. This makes it difficult to meet the high-efficiency computational requirements for large-scale wind farm micro-situation, resulting in high engineering practical costs.

[0004] Chinese patent application CN202511281096.8 discloses a micro-site selection optimization method for offshore wind farms considering regular layouts. This method combines mixed integer programming, genetic algorithms, and sparrow search algorithms. However, it requires sequentially constructing and solving multiple sub-models, resulting in a complex algorithm process. Furthermore, it relies on point-by-point calculations using wake models and sum-of-squares superposition, and requires piecewise linearization of power curves, further exacerbating the computational burden. Overall, this method is structurally complex, has numerous parameters, and incurs high computational costs, making it difficult to meet the practical engineering requirements for rapid micro-site selection in large-scale wind farms.

[0005] Chinese patent application CN202510021410.2 discloses a strategy for micro-situation of nearshore wind farms based on an improved wake model. However, it follows traditional micro-situation practices by setting the turbine hub height to a fixed value or excluding it from the optimization variables. In nearshore areas, especially in regions with significant deep-sea wind shear, a fixed turbine height prevents the utilization of high-altitude airflow velocity gains and the active control of wake distribution through height differences. This results in insufficient wind energy capture, severe power losses in downstream units, and low energy utilization, limiting the overall power generation efficiency of the wind farm.

[0006] Chinese patent application CN202111210478.3 discloses a method and apparatus for micro-site selection assessment of wind farms, as well as a computer-readable storage medium method. While relying on CFD models for wind resource assessment and micro-site selection can improve accuracy, the CFD method requires solving a large number of grid cells in the Navier-Stokes equations. In scenarios requiring iterative optimization, this results in enormous computational resource consumption and extremely high engineering costs. Furthermore, this method demands highly skilled operators and involves complex parameter tuning, making it difficult to meet the needs of rapid scheme comparison in the early stages of engineering projects, thus reducing the feasibility of technology promotion.

[0007] In summary, traditional wind turbine layout optimization methods often employ heuristic algorithms such as genetic algorithms and particle swarm optimization, but these suffer from slow convergence speeds and a tendency to get trapped in local optima. Furthermore, most existing methods set the turbine height to a fixed value or optimize height alone, failing to couple the optimization of horizontal position and height. For scenarios with significant wind shear in deep-sea areas, there is a lack of an intelligent layout method that can simultaneously optimize position and height, taking into account the power generation benefits of both the prevailing and sub-prevailing wind directions. Summary of the Invention

[0008] To address the aforementioned problems, this application proposes a wind farm micro-situation method based on an improved JAYA algorithm. This method optimizes the horizontal and vertical arrangement of wind turbines within the wind farm and incorporates a wake effect model, thereby increasing the annual power generation of the wind farm.

[0009] This invention discloses a wind farm micro-situation method based on an improved JAYA algorithm, the method comprising the following steps:

[0010] S1. Generate an initial population based on the Logistic chaotic mapping using the JAYA algorithm. Each individual in the initial population contains the X coordinate, Y coordinate, and hub height of all wind turbines in the wind farm.

[0011] S2. Calculate the wind speed by combining the wind turbine coordinates and hub height, and use the levelized cost per kilowatt-hour in the main wind direction and the secondary main wind direction as the fitness function to calculate the fitness value of each individual in the initial population.

[0012] S3. Set the current iteration number to... =1, taking the initial population generated in step S1 as the current population, and the optimal individual determined by the fitness of each individual in each initial population calculated in step S2 as the current global optimal solution, and start iterative optimization to generate candidate offspring individuals.

[0013] S4. Compare the fitness of the candidate offspring individuals generated in step S3 with the individuals of the previous generation population, perform reflection boundary processing on variables that exceed the search space boundary, and apply local fine-tuning perturbation based on Logistic chaotic mapping to the inferior individuals with lower fitness ranking.

[0014] S5. After the iteration is completed, output the best individual in the current population as the micro-site selection scheme of the wind farm. The best individual includes the optimized X coordinate, optimized Y coordinate and optimized hub height of each wind turbine.

[0015] Furthermore, step S1, which generates the initial population of wind turbines in the wind farm based on the Logistic chaotic mapping, involves the following steps:

[0016] S1-1. Initialize parameters and determine the population size. Number of fans Spatial upper and lower limits of population size, number of iterations Maximum number of iterations ;

[0017] S1-2. Use Logistic chaotic mapping to generate a random sequence for the initial population. and determine the random sequence The internal relationships are as follows: ...... In the formula, For the mapped sequence The value of the step; This is a control parameter, ranging from 0 to 4. For Logistic chaotic mapping function, A random number between 0 and 1;

[0018] S1-3. The initial random population matrix is ​​obtained by combining the logistic chaotic mapping with the wind turbine location and height correlation parameters of the wind farm through the following calculation. :

[0019] Assume the planned number of wind turbines installed in the wind farm is... That is, a single population contains N wind turbines, and each wind turbine needs to optimize three variables: horizontal X-coordinate, horizontal Y-coordinate, and hub height. Therefore, the variable dimension of a single population is... The population size is The total length of the chaotic sequence to be generated is ;

[0020] Therefore, the initial random population matrix For one The matrix has D rows and D columns, with each row representing a population of individuals, and each population representing a wind turbine layout scheme for a wind farm. The column information contains the X coordinates, Y coordinates, and hub heights of N wind turbines; for the i-th individual in the population and the k-th wind turbine, k=1,2,...,N, its X coordinate is... The Y-coordinate is The height value is ;

[0021] The Logistic chaotic sequence generated in S1-2 First, fill in a number of rows in order of priority. The matrix Z, where Z stores normalized chaotic values, is... Representation matrix The Line 1 Column, of which the The line represents the first Individuals in a population, the first The column represents the first individual in this population. The normalized chaotic value corresponding to the dimensional variable. ;

[0022] For the initial random population matrix No. The first individual in the population For a 3D variable, calculate all element values ​​according to the following formula:

[0023] (1) Mapping of the wind turbine's X and Y coordinates:

[0024] matrix The formula for the X-coordinate of the i-th individual is: matrix The formula for the Y-coordinate of the i-th individual is: In the formula, These are the lower and upper boundaries of the wind farm area in the X direction, respectively; These are the lower and upper bounds in the Y direction, respectively;

[0025] (2) Mapping of wind turbine hub height:

[0026] Wheel height is available in only two options: low wheel height. and high wheel hub height For the dimension corresponding to the wheel hub height, a threshold judgment method is used to determine the value of the wheel hub height. Representation matrix The Line 1 If Then the hub height is taken as ;like Then the hub height is taken as Therefore, its formula is: .

[0027] The above calculations yielded the X-coordinate, Y-coordinate, and hub height of the k-th wind turbine for the i-th individual in the population. , that is , , k=1,2,...,N, i=1,2,... That is, to obtain The i-th row of the matrix; the number of individuals in the population is , The matrix has 10 rows. ,calculate This can be obtained. The information of the remaining rows of the matrix is ​​then obtained All element values ​​of the matrix.

[0028] Furthermore, the specific steps of step S2 are as follows:

[0029] S2-1, Based on the wind turbine hub height obtained from S1-3, i.e., the low hub height. Or high wheel hub height In addition, open-source measured wind speed and direction data from the wind measurement towers at the wind farm site were used to calculate the wind speed of each wind turbine under the influence of upstream wind turbines using the wind farm wake model.

[0030] First, calculate the wind speed of the downstream fan affected by a single upstream fan, and then calculate the wind speed of the downstream fan affected by the wake effect of a single upstream fan. Represented as: In the formula, The thrust coefficient is usually obtained from the thrust coefficient curve provided by the manufacturer. β is the blade radius of the wind turbine; β is the wake attenuation coefficient. This refers to the hub height of the wind turbine, with the value being the lower hub height. Or high wheel hub height ; For surface roughness, this invention uses 0.04; Wake flow influencing factors; This represents the shaded area of ​​the front and rear fans; The area of ​​the wake region. Indicates natural wind speed;

[0031] Then, the wind speed of the downstream wind turbine under the influence of the wake effects of multiple upstream wind turbines is calculated: Downstream wind turbines in a wind farm are often affected by the combined wake effects of multiple upstream wind turbines, i.e., the wake superposition effect. The sum of squares of energy deficit is usually used to calculate the composite wake velocity, thus obtaining the wind speed of the downstream wind turbine under the influence of multiple upstream wind turbines. for: In the formula: The wind speed of the downstream wind turbine after being affected by the wake effect of a single upstream wind turbine; wake influence factor. This also corresponds to the s-th wind turbine upstream; Indicates the number of fans;

[0032] S2-2. The wind speed obtained in S2-1 above is the wind speed at the height of the wind tower. First, the wind speed at the height of the wind turbine hub is calculated using the power-law wind shear model. Based on the formula for wind speed under the influence of multiple upstream wind turbines, the wind speed at the hub height of the wind turbine d under the influence of upstream wind turbines is calculated. :

[0033] First, calculate the wind speed at the height of the wind turbine hub. Substituting the wind turbine power curve, the wind power is calculated. The conversion of wind speed at different heights is calculated using a power-law wind shear model, the specific formula of which is as follows: In the formula: Wind turbine hub height The actual wind speed at the location; For reference height The wind speed at the location; γ is the wind shear index, which is 0.14 in this invention;

[0034] Then, using the formula for wind speed under the influence of multiple upstream wind turbines, the wind speed at the hub height of the wind turbine d under the influence of multiple upstream wind turbines is calculated. :

[0035] The variable dimension of the i-th individual in the population is D=3N. For the k-th wind turbine, k=1,2,...,N, its X-coordinate is... The Y-coordinate is The height value is Let the X-coordinate, Y-coordinate, and hub height of the upstream wind turbine s be... The X-coordinate, Y-coordinate, and hub height of the downstream wind turbine d are: ;

[0036] For ease of calculation, the coordinate axis with north as 0° is rotated counterclockwise. To determine the angle, first rotate the coordinates of fan s and fan d. , For the fans s and d to rotate counterclockwise The X coordinate below, For the fans s and d to rotate counterclockwise The Y-coordinate is given by the following formula: In the formula Predominant wind angle;

[0037] Calculate the relative X and Y distances of upstream and downstream wind turbines s and d. , :

[0038] Combined with the wind speed of downstream wind turbines affected by multiple upstream wind turbines Formula and wind speed at wind turbine hub height The formula yields the wind speed at the hub height of wind turbine d under the influence of multiple upstream wind turbines. The formula is:

[0039] S2-3. Based on the wind speed at the hub height of each wind turbine under the influence of multiple upstream wind turbines calculated in S2-2, and combined with the power curve of the wind turbine, the cut-in wind speed is determined. Rated wind speed Cut-off wind speed and rated power Defined, the power of each wind turbine under various wind conditions is obtained using the standard piecewise interpolation method. The formula for the power dP of the wind turbine is: .

[0040] S2-4. Substitute the calculated power of each wind turbine in the annual power generation calculation to obtain the theoretical annual power generation of the wind farm. The formula for calculating the annual power generation of the wind turbine is: In the formula: This refers to the number of wind turbines. Hours per year (8760h); The output power of the fan at the specified wind speed is determined by S2-3; Let the probability density function of the wind speed at the hub height be a two-parameter Weibull distribution: In the formula: For shape parameters, The scale parameter is obtained by fitting site wind measurement data;

[0041] S2-5 Calculate the life cycle cost of a wind farm Its formula is: In the formula: For initial investment costs, For operation and maintenance costs, This represents the cost of decommissioning and dismantling. The components in the formula are as follows: In the formula: The cost of foundation construction for a single wind turbine. Cost of offshore booster stations Cost of the on-site power collection lines.

[0042] In the formula: The design life of a wind farm is typically 20-25 years. The maintenance cost for year t (including routine maintenance, replacement of major components, insurance, etc.); is the discount rate.

[0043] In the formula: For equipment dismantling costs, The residual value of recycled materials.

[0044] S2-6. Minimizing the levelized cost of electricity (LCOE) is the optimization objective and is used as the fitness function of the algorithm. The formula is: In the formula: This represents the total annual power generation of the wind farm in the prevailing wind direction. This represents the total annual power generation of the wind farm in the secondary prevailing wind direction.

[0045] Furthermore, step S3, which uses the initial population as the current population and the best individual in the initial population as the current global optimal solution, begins iterative optimization to generate candidate offspring individuals. The steps are as follows:

[0046] S3-1. Obtain the initial population generated in step S1 and the fitness of each initial individual calculated in step S2, and set the current iteration number as... =1, using the initial population as the current population, the best individual in the initial population as the current global optimal solution, and setting a probability threshold that decreases linearly with the number of iterations. Then begin iterative optimization;

[0047] S3-2. Determine the probability sequence based on the strategy generated by the number of iterations. Calculate the fitness of each individual in the population and increment the iteration count by one;

[0048] S3-3, Determining Probability Sequences Current value size, if If yes, then execute S3-4; otherwise, execute S3-5.

[0049] S3-4. Perform a global search based on the historical learning strategy, and generate candidate offspring individuals using the difference vector between historical individuals and the current individual;

[0050] S3-5. Perform local search based on the upper and lower local attraction points, and construct the upper and lower local attraction points using the current global optimal solution, the current global worst solution, and the population average solution to generate candidate offspring individuals.

[0051] Furthermore, the probability threshold that decreases linearly with the number of iterations... The calculation formula is: In the formula: and These are the preset maximum and minimum probability thresholds (in this invention, the thresholds are...). , ), This represents the current iteration number. This represents the maximum number of iterations.

[0052] Furthermore, the global search steps described in S3-4 are as follows:

[0053] S3-4-1. Calculate the difference vector between the historical solution and the current solution. The historical solution is generated by the following formula: In the formula: A sequence formed by rearranging historical solutions; It is a sequence of random values;

[0054] S3-4-2. Calculate the next generation of individuals. The calculation formula for the next generation of individuals is as follows: In the formula: A random number uniformly distributed between [0,1].

[0055] Furthermore, the local search step described in S3-5 is as follows:

[0056] S3-5-1 Calculate the average solution Local attraction points and lower local attraction points Their values ​​are as follows: In the formula: , , , A random number uniformly distributed between [0,1]. and These are the current global optimal solution and the worst solution, respectively. The average solution is obtained by weighting all individuals in the initial population S1 if it is the first iteration; otherwise, it is obtained by weighting all individuals in the current population.

[0057] S3-5-2. Based on the definitions of the average solution, the upper local attraction point, and the lower local attraction point, the next generation of individuals is calculated. The calculation formula for the next generation of individuals is as follows: In the formula: , A random number uniformly distributed between [0,1]. This is the first iteration update. The value of an individual; For the next iteration update The value of an individual.

[0058] Furthermore, the reflection boundary processing strategy steps described in S4 are as follows:

[0059] S4-1-1 Record the number of reflections =1, Determine the individual values ​​in matrix Z With search space boundary , Size, if corresponding Representing the X coordinate, then , for ;

[0060] S4-1-2, if corresponding Representing the Y coordinate, then , for , Dangruo If it is below the lower bound, execute S4-1-3; if If it is above the upper limit, execute S4-1-4;

[0061] S4-1-3, will Perform the following calculation to reflect it back to the inside of the lower boundary: = 2* Execute S4-1-5;

[0062] S4-1-4, will Perform the following calculation to reflect it back to the inside of the upper bound: = 2* Execute S4-1-5;

[0063] S4-1-5. Increment the reflection count by one and determine the current iteration count. and Size relationship, if Then determine the current individual value. Does it meet the requirements? If the condition is not met, execute S4-1-2; if it is met, end the reflection processing step. > Then re-initialize:

[0064] , Represents a random number between 0 and 1. From 0 to A random number between [a certain number of points].

[0065] Furthermore, the steps of the local fine-tuning perturbation are as follows:

[0066] S4-2-1. Filtering out inferior individuals: Sort all individuals in the current population according to their fitness values ​​from smallest to largest, and select those with the lowest fitness ranking. % of individuals were selected as the subjects of fine-tuning. The value range is 5 to 20, and the present invention takes... =10;

[0067] S4-2-2, Triggering condition judgment: Only when the current iteration number is reached... Preset interval Fine-tuning is triggered when the value is an integer multiple of the target value. The value range is 3 to 10, and the present invention takes... =5; If the current iteration count does not meet the triggering condition, skip this strategy and proceed directly to S7;

[0068] S4-2-3, Adaptive Chaotic Fine-tuning: For the selected disadvantaged individual, apply a small perturbation based on Logistic chaotic mapping near its current wind turbine layout coordinates. The perturbation formula is:

[0069] In the formula, for The new value after perturbation (perturbation is applied only to the X and Y coordinates, and the hub height dimension is not involved in the perturbation); The chaotic values ​​generated by the Logistic chaotic mapping in step S1-2, The adaptive perturbation amplitude coefficient decreases linearly with the number of iterations, and its formula is:

[0070] In the formula: The value range is 0.05 to 0.2. The value range is 0.005~0.02. This represents the current iteration number. To determine the maximum number of iterations, this invention takes... =0.1, =0.01.

[0071] S4-2-4 Boundary Constraint Handling: If an individual after fine-tuning exceeds the wind farm planning boundary, it shall be handled in accordance with the reflection boundary strategy described above.

[0072] The present invention also provides a wind farm micro-situation optimization system, comprising:

[0073] The data input module is used to acquire wind resource data, wind turbine parameters, and wind farm planning boundary data at the wind farm site.

[0074] The population initialization module is used to generate the initial population for the wind turbine layout of the wind farm based on the Logistic chaotic mapping.

[0075] The fitness calculation module is used to calculate the fitness of individuals in the population using the unit cost in the prevailing wind direction and the sub-prevailing wind direction as the fitness function.

[0076] The iterative optimization module is used to perform population iterative optimization of wind turbine layout in wind farms based on historical learning strategies. During the iteration process, the allocation ratio of global search and local search is dynamically adjusted by using a linearly decreasing probability threshold. For individuals that exceed the boundary, a reflection boundary processing strategy is used, and for individuals with low fitness ranking, a local fine-tuning perturbation strategy is used.

[0077] The results output module is used to output the optimized wind turbine layout coordinates and hub height.

[0078] Compared with the prior art, the present invention has the following advantages:

[0079] 1. Chaotic initialization: The initial population is generated using a Logistic sequence to avoid the clustering phenomenon of random initialization and improve the global convergence ability.

[0080] 2. Height-location joint optimization: By using the wind turbine hub height as an optimization variable and participating in the layout iteration, it can adapt to the deep-sea wind shear environment and increase annual power generation by 5% to 12% compared with the fixed height scheme.

[0081] 3. Dual wind direction adaptability: It not only considers the power generation of the main wind direction, but also takes into account the secondary main wind direction, so that the final layout is more favorable to both the main and secondary wind directions, reducing the drastic fluctuations in power generation caused by changes in wind direction.

[0082] 4. Historical learning strategy: Introducing historical solution difference vectors into the global search can accelerate convergence; constructing up and down attractors in the local search can avoid getting trapped in local optima.

[0083] 5. Engineering cost coupling: The cost of tower height is linearly incorporated into the objective function, avoiding blindly increasing the height of the tower and achieving a balance between economic efficiency and power generation. Attached Figure Description

[0084] Figure 1 This is an overall flowchart of the method of the present invention;

[0085] Figure 2 This is a schematic diagram of the wake effect in this invention;

[0086] Figure 3 This is a diagram illustrating the wake superposition.

[0087] Figure 4 This is the iteration diagram of the JAYA algorithm;

[0088] Figure 5 This is a layout diagram of a wind farm. Detailed Implementation

[0089] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. This embodiment presents a wind farm micro-site selection method based on an improved JAYA algorithm. Taking a deep-sea wind farm with a total installed capacity of 426 MW as an example, it plans to install 71 wind turbines with a single unit capacity of 6MW. The prevailing wind direction is North-Northeast (22.5°), and the secondary prevailing wind direction is North-Northwest (337.5°). The wind shear index is taken as 0.14, the surface roughness is 0.04 (sea surface), and the wake attenuation coefficient k=0.04. Figure 1 As shown, the specific steps include the following:

[0090] S1: Generate the initial population for the JAYA algorithm based on Logistic chaotic mapping.

[0091] S1-1, Perform parameter initialization

[0092] Wind turbine parameters: cut-in wind speed 3 m / s, rated wind speed 12 m / s, cut-out wind speed 25 m / s, rated power 6 MW, rotor diameter 77 m, hub height adjustable range [103.5 m, 163.5 m].

[0093] Cost parameters: The cost of a single foundation is 3.81 million yuan, and the cost of a tower per unit height is 36,400 yuan.

[0094] Optimize algorithm parameters: Set the population size to 80 and the maximum number of iterations to 100.

[0095] Wind resource data: based on the annual average wind speed and direction data of a certain deep-sea wind measurement tower.

[0096] The wind farm is planned to install n=71 wind turbines. The variables to be optimized for each turbine include: horizontal X-coordinate (in meters), horizontal Y-coordinate (in meters), and hub height (in meters). Therefore, the variable dimension of a single turbine is D=71×2=142, with height being a zero-one variable. The population size is set to... =80, meaning each generation contains 80 candidate layout schemes.

[0097] The search space boundaries are defined as follows:

[0098] The range of the X coordinate is m, Y coordinate range is The wheel hub height ranges from 103.5m to 163.5m.

[0099] Set the control parameter for the Logistic chaotic mapping to μ = 3.99, with an initial value of... =0.372. Set the total length of the chaotic sequence to be generated to L = 80 × 142 = 11360.

[0100] S1-2: Generate a Logistic chaotic sequence

[0101] 11360 chaotic values ​​are generated iteratively according to the following formula:

[0102] The generated partial values ​​are shown in Table 1.

[0103] Table 1: Steps Chaos value 1 0.3720 2 0.9326 3 0.2507 4 0.7503 5 0.7485 ... ...

[0104] The sequence exhibits seemingly random but completely deterministic chaotic traversal characteristics, with no obvious periodicity or clustering.

[0105] S1-3: Mapping to wind turbine layout variables

[0106] Fill the chaotic sequence into a... matrix Here, the i-th row represents the i-th individual, and the j-th column represents the normalized chaotic value corresponding to the j-th dimension variable of that individual. Then, it is mapped to the actual physical space using the following formula:

[0107]

[0108] In the formula These are the lower and upper bounds of the j-th dimension variable, respectively.

[0109] Taking the first two dimensions of the first individual as an example, the calculation is as follows: .

[0110] The initial population was obtained by calculating each of the eighty individuals.

[0111] S2: Combining coordinates and hub height, calculate the fitness function using the levelized cost of electricity in the primary and secondary wind directions.

[0112] This embodiment details the specific calculation process of the fitness function described in step S2. For ease of understanding, an intermediate individual (i.e., a wind turbine layout scheme) is selected as an example, containing 5 wind turbines. The calculation logic for 71 wind turbines in the actual algorithm is completely consistent with this, only the amount of data increases and the conditions become more complex.

[0113] The coordinates and hub height settings of the five wind turbines in this individual scheme are shown in Table 2.

[0114] Table 2: Fan number X-coordinate (m) Y coordinate (m) Wheel hub height (m) 1 500 500 110 2 800 500 120 3 1100 500 130 4 800 900 115 5 1100 900 125

[0115] S2-1: Calculate the wind speed of each wind turbine under the influence of the upstream wind turbine:

[0116] Taking a prevailing wind direction of 45° as an example, a coordinate system along the wind direction is established. Rotating the wind turbine coordinate system by 45° yields the downwind distance x′ and the crosswind distance y′. The formula for the wind turbine speed under the influence of the wake is as follows: The downwind position (m) of each fan can be obtained: Fan 1: =500×0.7071+500×0.7071=707.1 Fan 2: =800×0.7071+500×0.7071=919.2 Fan 3: =1100×0.7071+500×0.7071=1131.4 Fan 4: =800×0.7071+900×0.7071=1202.1 Fan 5: =1100×0.7071+900×0.7071=1414.2

[0117] The impact of fan 1 on fan 2 is calculated below.

[0118] Difference between the two wind turbines in the downwind direction =919.2-707.1=212.1m

[0119] Side wind distance difference between the two wind turbines =0 - (-212.13) = 212.13m

[0120] wake radius is The value is greater than the wake radius, therefore fan 2 is not affected by fan 1.

[0121] For fan 2 and fan 3 The value is less than the wake radius. This is the projection of the upstream wind turbine wake. for The area overlapping with the downstream wind turbine, then = < 1, such as Figure 3 As shown.

[0122] Assuming that fans 2 and 3 are perfectly aligned in the wind direction, = <1, such as Figure 2 As shown, the effect of the wake is calculated below.

[0123]

[0124] If affected by multiple wind turbines simultaneously, the sum-of-squares rule should be applied: .

[0125] S2-2: Calculate the actual incoming air velocity at the height of the wind turbine hub, such as Figures 2-3 As shown.

[0126] Using the wind shear formula, the wind speed of each wind turbine under the influence of the upstream wind turbine is calculated: .

[0127] The calculated wind speeds of each fan under the influence of the upstream fan are shown in Table 3.

[0128] Table 3: serial number wind speed 1 9.1215 2 9.2340 3 9.3402 4 9.1782 5 9.2880

[0129] S2-3: Calculate the fan power

[0130] Taking fan 1 as an example, Then the power of fan 1 is: .

[0131] Similarly, other wind turbines can be calculated, yielding a total power of 12.74MW.

[0132] S2-4: Calculate the power generation under this wind direction.

[0133] Assuming this wind condition occurs for 120 hours per year, the corresponding power generation of the wind farm under this condition would be: .

[0134] S2-5: Calculate the total cost of the wind farm .

[0135] S2-6: Calculate the fitness function

[0136] If the wind energy from the secondary prevailing wind direction is 1 GWh under this wind condition, then the cost, i.e., the fitness function, is: .

[0137] This value represents the fitness of the candidate arrangement. The optimization algorithm will use this as its objective, continuously updating the population and seeking the maximum value of the fitness function.

[0138] S3: Algorithm iteration.

[0139] An improved JAYA algorithm is used iteratively, with the initial population generated in step S1 as the current population, and the individual with the best fitness obtained in step S2 as the current global optimum. A probability threshold that decreases linearly with the number of iterations is set. The probability threshold is used to dynamically switch between global search and local search. S4 is executed followed by S5, and the flowchart is as follows: Figure 1 As shown;

[0140] This embodiment details the population iterative optimization process based on the historical learning strategy described in step S3, including the setting of continuation parameters. Taking the 30th iteration as an example, it demonstrates the complete process of population update in a single iteration.

[0141] S3-1: Read S1 and S2, and initialize the current population and fitness:

[0142] At the end of the twenty-ninth iteration, the population The fitness of each individual Given. Wherein:

[0143] Global optimal solution fitness =0.000395 (corresponding to individual number 45);

[0144] Global worst solution fitness =0.000332 (corresponding to individual number 12);

[0145] S3-2: For each individual i in the population, generate a uniformly random number. Generate probability judgment sequence And begin iterating, as shown in Table 4.

[0146] Table 4:

[0147] S3-3: Determine the probability sequence Current value size, If the value is greater than 0.5, then individual 1 will proceed to the global search step. 3876 < 0.5.

[0148] S3-4: Global search steps (taking individual 1 as an example)

[0149] S3-4-1: Calculate the difference vector between the historical solution and the current solution:

[0150] The current value of individual 1 is Select global search. This will search the previous generation population. All individuals are randomly shuffled in order, resulting in After shuffling, the paired individual corresponding to individual 1 is the 52nd individual in the original population. .

[0151] Calculate the difference vector:

[0152] In actual calculations, each dimension of the variable is calculated separately.

[0153] S3-4-2: The formula for generating the next generation of individuals is: set up The value of the middle dimension is , ,but: .

[0154] S3-5: Local search steps (taking individual 2 as an example)

[0155] Individual 2's current value is Choose local search.

[0156] The globally optimal value remains individual number 45, and the globally worst value remains individual number 12. Taking the first wind turbine as an example:

[0157]

[0158] S3-5-1: Calculate the average solution, upper attraction point, and lower attraction point:

[0159] S3-5-2: Calculate the value of the next generation of new individuals:

[0160] Update the population and compare fitness

[0161] For newly generated individuals Wait, call the fitness calculation function to get And so on, comparing the fitness of the corresponding individuals in the previous generation. If Then keep Otherwise, leave it as is. .

[0162] S4: Variable reflection boundary handling strategy.

[0163] In the updated population, some individuals may have certain variables that exceed the search space boundary. The following example illustrates the complete process of reflection boundary handling.

[0164] S4-1-1: Initialize parameters

[0165] Record the number of reflections Maximum number of reflections . judge Corresponding to the X coordinate, therefore , .

[0166] S4-1-2: Determine the direction of boundary crossing

[0167] If the value is below the lower bound, execute S4-1-3.

[0168] If it is higher than the upper limit, execute S4-1-4.

[0169] S4-1-3: Reflected back to the inner side of the lower boundary m,

[0170] at this time m has fallen within the range of [0, 3000], and the reflection was successful.

[0171] S4-1-4: Reflected back to the inner side of the upper boundary 2880m

[0172] at this time 2880m, which falls within the range of [0, 3000], reflection successful.

[0173] S4-1-5: Check the reflection results

[0174] Number of reflections Determine if the current value satisfies Once the conditions are met, the reflection process ends.

[0175] Example 5: Local fine-tuning perturbation

[0176] S4-2-1: Screening for disadvantaged individuals

[0177] Sort the current population of 80 individuals by fitness value from smallest to largest (lower fitness is better), and select the individuals ranked lower. Individuals numbered 73 to 80 (a total of 8) were selected as the subjects for fine-tuning, as shown in Table 5.

[0178] Table 5: Ranking Individual ID fitness value Fine-tuning? 1 (Optimal) 45 0.000395 no ... ... ... no 72 67 0.000478 no 73 12 0.000512 yes 74 34 0.000523 yes 75 56 0.000534 yes 76 23 0.000548 yes 77 78 0.000562 yes 78 41 0.000581 yes 79 19 0.000603 yes 80 (Worst) 66 0.000625 yes

[0179] S4-2-2: Triggering condition judgment

[0180] Current iteration number Preset interval Determine if 30 is a multiple of 5: If the remainder is 0, the trigger condition is met, and fine-tuning is performed.

[0181] S4-2-3: Adaptive Chaotic Fine-tuning

[0182] Taking the X-coordinate of the first wind turbine of the disadvantaged individual 12 as an example, the current value m.

[0183] In the 30th iteration, the adaptive perturbation amplitude coefficient is:

[0184] Take the chaotic value in step S1-2 The fine-tuning amount is: m

[0185] New value: m

[0186] The above perturbation is applied to all X and Y coordinates of the inferior individual (the hub height dimension is not included in the perturbation) to obtain a new individual after fine-tuning.

[0187] S4-2-4: Boundary Constraint Handling

[0188] If the finely adjusted individual exceeds the wind farm planning boundary (e.g.) or The reflection boundary processing strategy is applied. Let's assume a certain X-coordinate after fine-tuning... m, exceeding the upper bound:

[0189] First reflection: m (falls within the valid range)

[0190] S5: Determine the termination condition:

[0191] Following step S4, after updating 80 individuals, a new generation of the population is obtained. The current iteration count is 30, which is less than the preset iteration count, so we return to S2 and continue with the 31st iteration. After 100 iterations, the final optimized solution for the wind turbine can be obtained. The algorithm iteration curve is shown below. Figure 4 As shown, the final optimized scheme for the wind turbine is as follows: Figure 5 As shown.

[0192] The method of this invention can be used as a software module for early-stage micro-site selection optimization of offshore wind farms. After inputting wind resource data and turbine parameters, it directly outputs a list of optimized turbine coordinates and heights, applicable to various floating and fixed deep-sea wind farms. This method can also be extended to onshore wind farms with complex terrain.

[0193] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, any modifications, equivalent substitutions, or improvements made without departing from the principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A wind farm micro-siting method based on an improved JAYA algorithm, characterized in that, The method includes the following steps: S1. Generate an initial population based on the Logistic chaotic mapping using the JAYA algorithm. Each individual in the initial population contains the X coordinate, Y coordinate, and hub height of all wind turbines in the wind farm. S2. Calculate the wind speed by combining the wind turbine coordinates and hub height, and use the levelized cost per kilowatt-hour in the main wind direction and the secondary main wind direction as the fitness function to calculate the fitness value of each individual in the initial population. S3, setting the current iteration number as =1, using the initial population generated in step S1 as the current population, using the optimal individual determined by the fitness of each individual in each initial population calculated in step S2 as the current global optimal solution, starting the iteration optimization to generate candidate offspring individuals; S4. Compare the fitness of the candidate offspring individuals generated in step S3 with the individuals of the previous generation population, perform reflection boundary processing on variables that exceed the search space boundary, and apply local fine-tuning perturbation based on Logistic chaotic mapping to the inferior individuals with lower fitness ranking. S5. After the iteration is completed, output the best individual in the current population as the micro-site selection scheme of the wind farm. The best individual includes the optimized X coordinate, optimized Y coordinate and optimized hub height of each wind turbine.

2. The wind farm micro-situation method based on the improved JAYA algorithm according to claim 1, characterized in that, Step S1 involves generating the initial population for wind turbine arrangement in a wind farm based on Logistic chaotic mapping. The specific steps are as follows: S1-1. Initialize parameters and determine the population size. Number of fans Spatial upper and lower limits of population size, number of iterations Maximum number of iterations ; S1-2. Use Logistic chaotic mapping to generate a random sequence for the initial population. and determine the random sequence The internal relationships are as follows: ...... In the formula, For the mapped sequence The value of the step; This is a control parameter, ranging from 0 to 4. For Logistic chaotic mapping function, A random number between 0 and 1; S1-3. The initial random population matrix is ​​obtained by combining the logistic chaotic mapping with the wind turbine location and height correlation parameters of the wind farm through the following calculation. : Assume the planned number of wind turbines installed in the wind farm is... That is, a single population contains N wind turbines, and each wind turbine needs to optimize three variables: horizontal X-coordinate, horizontal Y-coordinate, and hub height. Therefore, the variable dimension of a single population is... The population size is The total length of the chaotic sequence to be generated is ; Therefore, the initial random population matrix For one The matrix has D rows and D columns, with each row representing a population of individuals, and each population representing a wind turbine layout scheme for a wind farm. The column information contains the X coordinates, Y coordinates, and hub heights of N wind turbines; for the i-th individual in the population and the k-th wind turbine, k=1,2,...,N, its X coordinate is... The Y-coordinate is The height value is ; The Logistic chaotic sequence generated in S1-2 First, fill in a number of rows in order of priority. The matrix Z, where Z stores normalized chaotic values, is... Representation matrix The Line 1 Column, of which the The line represents the first Individuals in a population, the first The column represents the first individual in this population. The normalized chaotic value corresponding to the dimensional variable. ; For the initial random population matrix No. The first individual in the population For a 3D variable, calculate all element values ​​according to the following formula: (1) Mapping of the wind turbine's X and Y coordinates: matrix The formula for the X-coordinate of the i-th individual is: matrix The formula for the Y-coordinate of the i-th individual is: In the formula, These are the lower and upper boundaries of the wind farm area in the X direction, respectively; These are the lower and upper bounds in the Y direction, respectively; (2) Mapping of wind turbine hub height: Wheel height is available in only two options: low wheel height. and high wheel hub height For the dimension corresponding to the wheel hub height, a threshold judgment method is used to determine the value of the wheel hub height. Representation matrix The Line 1 If Then the hub height is taken as ;like Then the hub height is taken as Therefore, its formula is: The above calculations yielded the X-coordinate, Y-coordinate, and hub height of the k-th wind turbine for the i-th individual in the population. , that is , , k=1,2,...,N, i=1,2,... That is, to obtain The i-th row of the matrix; the number of individuals in the population is , The matrix has 10 rows. ,calculate This can be obtained. The information of the remaining rows of the matrix is ​​then obtained All element values ​​of the matrix.

3. The wind farm micro-situation method based on the improved JAYA algorithm according to claim 2, characterized in that, The specific steps of step S2 are as follows: S2-1, Based on the wind turbine hub height obtained from S1-3, i.e., the low hub height. Or high wheel hub height In addition, open-source measured wind speed and direction data from the wind measurement towers at the wind farm site were used to calculate the wind speed of each wind turbine under the influence of upstream wind turbines using the wind farm wake model. First, calculate the wind speed of the downstream fan affected by a single upstream fan, and then calculate the wind speed of the downstream fan affected by the wake effect of a single upstream fan. Represented as: In the formula, The thrust coefficient is usually obtained from the thrust coefficient curve provided by the manufacturer. β is the blade radius of the wind turbine; β is the wake attenuation coefficient. This refers to the hub height of the wind turbine, with the value being the lower hub height. Or high wheel hub height ; For surface roughness; Wake flow influencing factors; This represents the shaded area of ​​the front and rear fans; The area of ​​the wake region. Indicates natural wind speed; Then, the wind speed of the downstream wind turbine under the influence of the wake effects of multiple upstream wind turbines is calculated: Downstream wind turbines in a wind farm are often affected by the combined wake effects of multiple upstream wind turbines, i.e., the wake superposition effect. The sum of squares of energy deficit is usually used to calculate the composite wake velocity, thus obtaining the wind speed of the downstream wind turbine under the influence of multiple upstream wind turbines. for: In the formula: The wind speed of the downstream fan after being affected by the wake effect of a single upstream fan; Wake Influence Factor This also corresponds to the s-th wind turbine upstream; Indicates the number of fans; S2-2. The wind speed obtained in S2-1 above is the wind speed at the height of the wind tower. First, the wind speed at the height of the wind turbine hub is calculated using the power-law wind shear model. Based on the formula for wind speed under the influence of multiple upstream wind turbines, the wind speed at the hub height of the wind turbine d under the influence of upstream wind turbines is calculated. : First, calculate the wind speed at the height of the wind turbine hub. Substituting the wind turbine power curve, the wind power is calculated. The conversion of wind speed at different heights is calculated using a power-law wind shear model, the specific formula of which is as follows: In the formula: Wind turbine hub height The actual wind speed at the location; For reference height Wind speed at the location; γ is the wind shear index; Then, using the formula for wind speed under the influence of multiple upstream wind turbines, the wind speed at the hub height of the wind turbine d under the influence of multiple upstream wind turbines is calculated. : The variable dimension of the i-th individual in the population is D=3N. For the k-th wind turbine, k=1,2,...,N, its X-coordinate is... The Y-coordinate is The height value is Let the X-coordinate, Y-coordinate, and hub height of the upstream wind turbine s be... The X-coordinate, Y-coordinate, and hub height of the downstream wind turbine d are: ; For ease of calculation, the coordinate axis with north as 0° is rotated counterclockwise. To determine the angle, first rotate the coordinates of fan s and fan d. , For the fans s and d to rotate counterclockwise The X coordinate below, For the fans s and d to rotate counterclockwise The Y-coordinate is given by the following formula: In the formula Predominant wind angle; Calculate the relative X and Y distances of upstream and downstream wind turbines s and d. , : Combined with the wind speed of downstream wind turbines affected by multiple upstream wind turbines Formula and wind speed at wind turbine hub height The formula yields the wind speed at the hub height of wind turbine d under the influence of multiple upstream wind turbines. The formula is: S2-3. Based on the wind speed at the hub height of each wind turbine under the influence of multiple upstream wind turbines calculated in S2-2, and combined with the power curve of the wind turbine, the cut-in wind speed is determined. Rated wind speed Cut-off wind speed and rated power Defined, the power of each wind turbine under various wind conditions is obtained using the standard piecewise interpolation method. The formula for the power dP of the wind turbine is: S2-4. Substitute the calculated power of each wind turbine in the annual power generation calculation to obtain the theoretical annual power generation of the wind farm. The formula for calculating the annual power generation of the wind turbine is: In the formula: This refers to the number of wind turbines. Hours per year (8760h); The output power of the fan at the specified wind speed is determined by S2-3; Let the probability density function of the wind speed at the hub height be a two-parameter Weibull distribution: In the formula: For shape parameters, The scale parameter is obtained by fitting site wind measurement data; S2-5 Calculate the life cycle cost of a wind farm Its formula is: In the formula: For initial investment costs, For operation and maintenance costs, The cost of decommissioning and dismantling; the components in the formula are as follows: In the formula: The cost of foundation construction for a single wind turbine. Cost of offshore booster stations Cost of the on-site power collection lines; In the formula: The design life of a wind farm is typically 20-25 years. The maintenance cost for year t is... The discount rate; In the formula: For equipment dismantling costs, The residual value of recycled materials; S2-6. Minimizing the levelized cost of electricity (LCOE) is the optimization objective and is used as the fitness function of the algorithm. The formula is: In the formula: This represents the total annual power generation of the wind farm in the prevailing wind direction. This represents the total annual power generation of the wind farm in the secondary prevailing wind direction.

4. The wind farm micro-situation method based on the improved JAYA algorithm according to claim 1, characterized in that, Step S3, using the initial population as the current population and the best individual in the initial population as the current global optimal solution, begins iterative optimization to generate candidate offspring individuals. The steps are as follows: S3-1. Obtain the initial population generated in step S1 and the fitness of each initial individual calculated in step S2, and set the current iteration number as... =1, using the initial population as the current population, the best individual in the initial population as the current global optimal solution, and setting a probability threshold that decreases linearly with the number of iterations. Then begin iterative optimization; S3-2. Determine the probability sequence based on the strategy generated by the number of iterations. Calculate the fitness of each individual in the population and increment the iteration count by one; S3-3, Determining Probability Sequences Current value size, if If yes, then execute S3-4; otherwise, execute S3-5. S3-4. Perform a global search based on the historical learning strategy, and generate candidate offspring individuals using the difference vector between historical individuals and the current individual; S3-5. Perform local search based on the upper and lower local attraction points, and construct the upper and lower local attraction points using the current global optimal solution, the current global worst solution, and the population average solution to generate candidate offspring individuals.

5. A wind farm micro-situation method based on an improved JAYA algorithm according to claim 4, characterized in that, The probability threshold that decreases linearly with the number of iterations The calculation formula is: In the formula: and These are the preset maximum and minimum probability thresholds, respectively. This represents the current iteration number. This represents the maximum number of iterations.

6. The wind farm micro-situation method based on the improved JAYA algorithm according to claim 4, characterized in that, The global search steps described in S3-4 are as follows: S3-4-1. Calculate the difference vector between the historical solution and the current solution. The historical solution is generated by the following formula: In the formula: A sequence formed by rearranging historical solutions; It is a sequence of random values; S3-4-2. Calculate the next generation of individuals. The calculation formula for the next generation of individuals is as follows: In the formula: It is a random number uniformly distributed between [0,1].

7. A wind farm micro-situation method based on an improved JAYA algorithm according to claim 4, characterized in that, The local search steps described in S3-5 are as follows: S3-5-1 Calculate the average solution Local attraction points and lower local attraction points Their values ​​are as follows: In the formula: , , , A random number uniformly distributed between [0,1]. and These are the current global optimal solution and the worst solution, respectively. The average solution is obtained by weighting all individuals in the initial population S1 if it is the first iteration; otherwise, it is obtained by weighting all individuals in the current population. S3-5-2. Based on the definitions of the average solution, the upper local attraction point, and the lower local attraction point, the next generation of individuals is calculated. The calculation formula for the next generation of individuals is as follows: In the formula: , A random number uniformly distributed between [0,1]. This is the first iteration update. The value of an individual; For the next iteration update The value of an individual.

8. The wind farm micro-situation method based on the improved JAYA algorithm according to claim 1, characterized in that, The steps of the reflection boundary processing strategy described in S4 are as follows: S4-1-1 Record the number of reflections =1, Determine the individual values ​​in matrix Z With search space boundary , Size, if corresponding Representing the X coordinate, then , for ; S4-1-2, if corresponding Representing the Y coordinate, then , for , Dangruo If it is below the lower bound, execute S4-1-3; if If it is above the upper limit, execute S4-1-4; S4-1-3, will Perform the following calculation to reflect it back to the inside of the lower boundary: = 2* Execute S4-1-5; S4-1-4, will Perform the following calculation to reflect it back to the inside of the upper bound: = 2* Execute S4-1-5; S4-1-5. Increment the reflection count by one and determine the current iteration count. and Size relationship, if Then determine the current individual value. Does it meet the requirements? If the condition is not met, execute S4-1-2; if it is met, end the reflection processing step. > Then re-initialize: , Represents a random number between 0 and 1. From 0 to A random number between [a certain number of points].

9. A wind farm micro-situation method based on an improved JAYA algorithm according to claim 8, characterized in that, The steps for the local fine-tuning perturbation described in S4 are as follows: S4-2-1. Filtering out inferior individuals: Sort all individuals in the current population according to their fitness values ​​from smallest to largest, and select those with the lowest fitness ranking. % of individuals were selected as the subjects of fine-tuning. The value range is 5 to 20; S4-2-2, Triggering condition judgment: Only when the current iteration number is reached... Preset interval Fine-tuning is triggered when the value is an integer multiple of the target value. The value range is 3 to 10; if the current iteration number does not meet the triggering condition, skip this strategy and proceed directly to S7; S4-2-3, Adaptive Chaotic Fine-tuning: For the selected disadvantaged individual, apply a small perturbation based on Logistic chaotic mapping near its current wind turbine layout coordinates. The perturbation formula is: In the formula, for The new value after the perturbation; The chaotic values ​​generated by the Logistic chaotic mapping in step S1-2, The adaptive perturbation amplitude coefficient decreases linearly with the number of iterations, and its formula is: In the formula: The value range is 0.05 to 0.

2. The value range is 0.005~0.

02. This represents the current iteration number. This represents the maximum number of iterations. S4-2-4 Boundary Constraint Handling: If an individual after fine-tuning exceeds the wind farm planning boundary, it shall be handled in accordance with the reflection boundary handling strategy described above.

10. A wind farm micro-situation optimization system, characterized in that, include: The data input module is used to acquire wind resource data, wind turbine parameters, and wind farm planning boundary data at the wind farm site. The population initialization module is used to generate the initial population for the wind turbine layout of the wind farm based on the Logistic chaotic mapping. The fitness calculation module is used to calculate the fitness of individuals in the population using the unit cost in the prevailing wind direction and the sub-prevailing wind direction as the fitness function. The iterative optimization module is used to perform population iterative optimization of wind turbine layout in wind farms based on historical learning strategies. During the iteration process, the allocation ratio of global search and local search is dynamically adjusted by using a linearly decreasing probability threshold. For individuals that exceed the boundary, a reflection boundary processing strategy is used, and for individuals with low fitness ranking, a local fine-tuning perturbation strategy is used. The results output module is used to output the optimized wind turbine layout coordinates and hub height.

Images