A method and system for optimizing wind farm layout to balance noise and power generation

Abstract

This invention discloses a method and system for optimizing wind farm layout to balance noise and power generation. The method includes: establishing a multi-objective optimization model for wind farm layout; setting reference points and weight vectors based on decision preference information; and generating... N P A wind farm layout is used to form a parent population; the parent population is divided into several subpopulations and searched to generate mutation vectors; local difference operations are performed on the mutation vectors to obtain candidate layout solutions; based on the candidate layout solutions, a double substitution operation is performed on each target vector to generate a offspring population; the offspring population is merged with the parent population, and the optimal layout is selected. N P This invention involves repeatedly performing an optimization process on individual wind farms, outputting optimization results, and determining whether the decision-maker's preference information needs adjustment based on the optimization results. If not, it outputs the wind farm layout. This invention increases power generation while reducing wind farm noise, and can obtain a wind farm layout scheme that matches the decision-maker's preferences.

Description

Technical Field

[0001] This invention mainly relates to the field of wind farm layout optimization, specifically to a method and system for optimizing wind farm layout to balance noise and power generation. Background Technology

[0002] Wind farm noise has a higher modulation frequency than road traffic noise and attenuates poorly over distance and obstacles, causing harm to human physical and mental health. Complaints from residents about wind farm noise have hindered the construction of some wind farms. Because background noise levels are lower in rural areas, the tolerance for wind farm noise is further reduced, making the reduction of wind farm noise an urgent issue.

[0003] After selecting specific wind turbines, wind farm layout optimization can significantly reduce the impact of wind farm noise on residents. During wind power generation, the wake generated by upstream turbines reduces the wind speed of downstream turbines, thus decreasing the power output of the wind turbines. Generally, the further the turbines are from residents, the lower the noise level. However, increasing power generation to reduce wake losses inevitably increases the distance between upstream and downstream turbines. Therefore, there is a conflict between the goals of reducing noise in residential areas and increasing power generation by reducing wake losses. When planning wind farm layout, it is necessary to consider not only the balance between noise and power generation but also the noise limits and power generation requirements of the wind farm location. Traditional layout methods only consider the single goal of reducing wake losses and increasing power generation, while existing multi-objective optimization algorithms for wind farm layout do not consider predetermined objectives, resulting in low efficiency in finding a layout scheme that meets the decision-maker's preferences. Summary of the Invention

[0004] To address the technical problems existing in the prior art, this invention provides a method and system for optimizing wind farm layout to balance noise and power generation. Based on the decision-maker's preferences, a reference point and weight vector are preset, and through interactive adjustment, the power generation is increased while reducing wind farm noise, thereby achieving a balance between noise and power generation.

[0005] To solve the above-mentioned technical problems, the technical solution proposed by this invention is as follows:

[0006] A method for optimizing wind farm layout to balance noise and power generation includes the following steps:

[0007] S1. Obtain wind energy data, wind farm and turbine parameters, and establish a multi-objective optimization model for wind farm layout with the objectives of minimizing noise and maximizing power generation. The variables of the multi-objective optimization model are the coordinates of each turbine, and the outputs are the estimated annual power generation and noise of the wind farm.

[0008] S2. Set a predetermined target value containing a reference point and a weight vector based on the decision preference information; wherein the reference point contains the expected target values ​​of power generation and noise, and the weight vector is used to quantify the relative importance between power generation and noise.

[0009] S3, Randomly generated A total of several wind farm layout schemes that satisfy the boundary constraints and the safe distance constraints between wind turbines constitute the parent population; each wind farm layout scheme is an individual.

[0010] S4. The parent population is randomly divided into several subpopulations. Within each subpopulation, a differential strategy based on neighborhood search is used to search and generate the corresponding mutation vector.

[0011] S5. Perform a local difference operation on the mutation vector to obtain candidate layout solutions;

[0012] S6. Based on the candidate layout solution, perform adaptive replacement and repair replacement operations sequentially on each target vector in the current parent population to generate... Each experimental vector constitutes the offspring population;

[0013] S7. Merge the offspring population with the parent population to obtain a merged population; rate and rank the individuals in the merged population based on a predetermined target value, and select the optimal individuals. Each individual becomes a parent population for the next generation of evolution;

[0014] S8. Repeat steps S4 to S7 until the maximum number of evaluations is reached, and output the optimization results.

[0015] S9. Based on the optimization results, determine whether the decision-maker's preference information needs to be adjusted. If adjustment is needed, update the reference point or weight vector and return to step S2 to execute the optimization process. If no adjustment is needed, output the final wind farm layout scheme.

[0016] Preferably, step S1 specifically includes:

[0017] S11. Acquire and preprocess wind energy data, wind farm and turbine parameters, and establish wind speed probability distribution model, wake model and turbine power curve model.

[0018] S12. Based on the wind speed probability distribution model, wake model and wind turbine power curve model, calculate the expected power output function of each wind turbine, and then sum them to obtain the annual power generation of the wind farm. Take the negative value of the annual power generation as the first objective of the multi-objective optimization model.

[0019] S13. Based on wind farm parameters and acoustic standards, establish a two-dimensional wind farm acoustic model, take a series of preset points at the wind farm boundary and center as noise receivers, calculate the maximum noise value among all noise receivers, and take it as the second objective of the multi-objective optimization model.

[0020] Preferably, the annual power generation in step S12 The calculation formula is:

[0021]

[0022] in For wind turbine i The expected power output value, i =1,…, N， in N This refers to the number of wind turbines in the wind farm. Wind direction The probability density function, For the first i Typhoon generators at wind speed Power output at that time; In the wind direction wind speed The Weibull probability distribution model; Wind direction Scaling parameters considering wake effect, Wind direction The shape parameters below.

[0023] Preferably, in step S4, the difference strategy based on neighborhood search is:

[0024]

[0025] in For the mutation vector, The first vector represents the first vector. One dimension, This is the scaling factor; , Each represents a different individual randomly selected from the current subpopulation. It is the best individual in the current subpopulation.

[0026] Preferably, in step S5, the specific process of performing a local difference operation on the mutation vector to obtain the candidate layout solution is as follows:

[0027] The mutation vector is re-encoded, and the coordinates of each wind turbine are treated as an independent individual, forming a temporary population of size N. The temporary population is then subjected to local differencing, and the differencing population is re-encoded again to obtain candidate layout solutions.

[0028] Preferably, in step S6, the specific process of performing adaptive replacement and repair replacement operations sequentially on each target vector in the current parent population is as follows:

[0029] S61. Adaptive replacement operation: During the algorithm learning cycle, coordinates in the target vector are randomly selected for replacement and the replacement success rate information is recorded; after the learning cycle, based on the accumulated success rate information, a random traversal sampling method is used to determine the index of the replaced coordinates to improve search efficiency.

[0030] S62. Repair and Replacement Operation: Detect the wind turbine coordinates that violate the distance constraint in the test vector obtained after the adaptive replacement operation, and replace them sequentially with the subsequent coordinates in the candidate layout solution until the constraint condition is met or the maximum number of replacements is reached.

[0031] Preferably, in step S7, individuals in the merged population are rated and ranked based on a predetermined target value to select the optimal ones. The specific process for each individual is as follows:

[0032] Calculate the relationship between all individuals in the merged population and the reference point. The weighted Euclidean distance;

[0033] Define the dominance relation r: if Pareto Domination ;or and They are mutually non-Pareto dominants and satisfy... ,but "rdominated" ;in Let be a function defined based on weighted Euclidean distance; where and They are two individuals in the population. For thresholds not dominated by "r";

[0034] The hierarchy of all individuals in the merged population is determined by the dominance relationship of r;

[0035] Non-“r-dominated” individuals at the same level are sorted from largest to smallest by crowding distance;

[0036] Start by selecting all non-dominated layers from the lowest level until a certain layer is added, exceeding the total number. At that time, within this layer, select a subset of individuals based on their crowding distance, from largest to smallest, to fill the gaps. .

[0037] Preferably, The expression is:

[0038]

[0039] in and They are two individuals in the population. and They are weighted Euclidean distances. The maximum and minimum boundary values; Individuals and With reference point The weighted Euclidean distance.

[0040] The present invention also discloses a computer-readable storage medium having a computer program stored thereon, the computer program performing the steps of the method described above when run by a processor.

[0041] The present invention further discloses a wind farm layout optimization system that balances noise and power generation, including a memory and a processor connected to each other. The memory stores a computer program, which, when run by the processor, executes the steps of the method described above.

[0042] Compared with the prior art, the advantages of the present invention are as follows:

[0043] The wind farm layout optimization method of the present invention, which balances noise and power generation with predetermined target values, first acquires wind energy data, wind farm and turbine parameters, and establishes a wind farm layout optimization model with the objectives of minimizing noise and maximizing power generation. The target values ​​are pre-set for the algorithm, i.e., the reference points and weight vectors of the algorithm are set according to the decision-maker's preferences for noise and power generation. Then, a wind farm layout scheme is randomly generated, constituting a system that includes… The parent population of each individual is randomly divided into several subpopulations. A subpopulation search is performed to obtain mutation vectors. Local differencing is applied to the mutation vectors to obtain candidate layout solutions. Based on these candidate layout solutions, [the following is a process / method / approach]... The target vectors are obtained by performing a double substitution operation. Each experimental vector is used to form a progeny population. Finally, the progeny population and the parent population are merged into one population, ranked, and the best progeny is selected. Each individual wind farm is preserved as a parent population for the next evolutionary iteration. The above steps are repeated for this population until the maximum number of evaluations is reached. The optimization results are then saved. The decision-maker uses these results to determine if adjustments to their preferences are needed. If adjustments are required, the weight vector of the algorithm's reference points is adjusted, and optimization is repeated until a final wind farm layout that meets the decision-maker's preferences is obtained. This invention optimizes wind farm layouts by balancing the goals of wind farm noise and power generation, and efficiently finds an optimal layout that meets the decision-maker's preferences. It reduces wind farm noise and increases power generation by minimizing wake effects compared to traditional layouts.

[0044] This invention uses a Gaussian function that closely matches actual wind turbine power curve data to represent the power curve, establishes an annual power generation model, and is based on actual wind farm parameters and ISO. The 9613-2:2024 standard defines wind farm layout optimization models that simulate real-world wind farm conditions. These models define areas susceptible to noise, such as residential areas, small industrial zones, or bird sanctuaries, as a series of discrete points placed at regular intervals along the wind farm boundary, plus a point at the wind farm center. The model aims to optimize wind farm layout by minimizing noise and power generation. A robust, efficient, and practical multi-objective evolutionary algorithm, which allows for setting predetermined target values ​​based on decision-maker preferences, is used for wind farm layout optimization. Considering the "curse of dimensionality" that can occur with increasing turbine numbers during optimization, a dual-replacement strategy is employed to obtain the experimental vectors. The adaptive replacement operation maintains global search capability while leveraging prior learning experience to find better solutions with the same number of evaluations, thus improving the algorithm's convergence speed and stability. Furthermore, considering the strict constraints of wind farm layout problems, the repair replacement operation accelerates computation and increases the quantity and quality of feasible solutions. A "pair-sorting" method is proposed to determine whether the experimental vectors obtained after the adaptive replacement operation are better, recording information on whether the experimental vectors are superior. Alternatively, the r-nondominated sorting method is used to sort the offspring population and the parent population, and the better individuals are selected to form a new parent population to continue the evolution. This can efficiently obtain a wind farm layout scheme that meets the predetermined target value, reduce noise and wake effect while increasing power generation. Attached Figure Description

[0045] Figure 1 The flowchart is shown in an embodiment of the wind farm layout optimization method of the present invention.

[0046] Figure 2 This is a schematic diagram of the acoustic model of a two-dimensional wind farm in an embodiment of the present invention; (a) corresponds to the first wind farm, N=16, power generation=140.847GWh, noise=56.169dB; (b) corresponds to the second wind farm, N=24, power generation=207.604GWh, noise=52.345dB.

[0047] Figure 3 This is a schematic diagram of the double substitution operation in an embodiment of the present invention.

[0048] Figure 4 This is a diagram showing the frontier solution obtained from the first optimization in this invention and its corresponding reference point.

[0049] Figure 5 This is a convergence curve of the power generation AEP value and noise SPL value of the layout scheme corresponding to the maximum power generation in this invention as the number of evaluations increases.

[0050] Figure 6 This is a diagram showing the frontier solution and its corresponding reference point obtained after adjusting the preference information in this invention.

[0051] Figure 7 The diagram shows the optimized layout scheme obtained in the embodiment of the present invention; (a) corresponds to the first wind farm, N=16, power generation=146.600GWh, noise=43.229dB; (b) corresponds to the second wind farm, N=24, power generation=213.886GWh, noise=46.876dB. Detailed Implementation

[0052] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0053] like Figure 1 As shown in the figure, the wind farm layout optimization method for balancing noise and power generation provided in this embodiment of the invention specifically includes the following steps:

[0054] S1. Obtain wind energy data, wind farm and wind turbine parameters from the wind farm and perform preprocessing. Based on the wind speed probability distribution model, wake model and wind turbine power curve model, derive the annual power generation of the wind farm and take the negative value of the annual power generation as the first objective of the multi-objective optimization model.

[0055] Based on wind farm parameters and the ISO 9613-2:2024 standard, a two-dimensional acoustic model of the wind farm is established, as follows: Figure 2 As shown: The noise receiver is set as a series of discrete points placed at certain intervals along the boundary of the wind farm and a point at the center of the wind farm, representing areas that are susceptible to noise, such as residential areas, small industrial areas or bird sanctuaries. The maximum noise value in the noise receiver is used as the second objective of the multi-objective optimization model.

[0056] The wind energy data includes wind speed, wind direction, wind energy density, wind speed probability distribution, air temperature, air pressure, and humidity; wind farm parameters include wind farm location, scale, area, and layout; wind turbine parameters include wind turbine rated power, cut-in wind speed, cut-out wind speed, wind turbine height, and rotor diameter.

[0057] Specifically, the process of step S1 above is as follows:

[0058] S11. Obtain wind farm and wind turbine parameters, as well as wind direction and speed data measured by the wind farm's meteorological tower for 12 consecutive months; the wind farm has N There are 10 wind turbines, and the index of each wind turbine is: i , i = 1, … , N ;

[0059] Preprocessing of wind speed data in wind energy data: The wind speed probability distribution model is represented by a two-parameter Weibull distribution;

[0060] A wake model is established using the Jensen model;

[0061] The Gaussian function is used to represent the wind turbine power curve model, as shown in the following formula:

[0062]

[0063] in For wind turbine i At wind speed Power at that time This refers to the rated power of the fan; For wind speed, , and These are the wind turbine's cut-in velocity, rated velocity, and cut-out velocity, respectively. , and The values ​​are 3 m / s, 11.5 m / s, and 25 m / s, respectively. a , b and c These are the parameters obtained by fitting the wind turbine power curve data using the Gaussian function. a , b and c Specifically, the values ​​are 2959, 12.85, and 4.934;

[0064] S12. The annual power generation function of the wind farm is established by combining the two-parameter Weibull wind speed probability distribution model, the Jensen wake model, and the wind turbine power curve model. Annual power generation... The formula is as follows:

[0065]

[0066] in For wind turbine i The expected power output value, i =1,…, N， in N This refers to the number of wind turbines in the wind farm. Wind direction The probability density function, For the first i Typhoon generators at wind speed Power output at that time; In the wind direction wind speed The Weibull probability distribution model; Wind direction Scaling parameters considering wake effect, Wind direction The shape parameters below;

[0067] Will The negative value is used as the first objective of the multi-objective optimization model;

[0068] S13. Based on wind farm parameters and the ISO 9613-2:2024 acoustic standard, establish a two-dimensional wind farm acoustic model. Specifically, set the noise receiver as a series of discrete points placed at certain intervals along the boundary of the wind farm and a point at the center of the wind farm, representing areas that are susceptible to noise, such as residential areas, small industrial areas, or bird sanctuaries.

[0069] The noise calculation formula for a noise receiver is as follows:

[0070]

[0071] in The A-weighted equivalent continuous downwind sound pressure level is used to measure the intensity of the sound field. For the first Typhoon machine One equivalent continuous downwind octave band sound pressure level. For the first A weighted average of standard A;

[0072] The maximum noise value in the noise receiver is used as the second objective of the multi-objective optimization model.

[0073] The objective function of the multi-objective optimization model for wind farm layout is expressed as follows:

[0074]

[0075] in It is the highest A-weighted equivalent continuous downwind sound pressure level among all noise receivers in the wind farm;

[0076] The variables in the wind farm layout optimization model are the two-dimensional coordinates of each wind turbine, represented as [ x i1 , x i2 ], i =1,…, ; The coordinates of each wind turbine constitute an individual in the population. That is, a layout solution , , and These represent the first objective value and the second objective value of the multi-objective optimization model, respectively. In this example... and 24;

[0077] The boundary constraints of the wind farm layout optimization model are: the wind turbines are located within the shape of the wind farm, such as a rectangular wind farm with a horizontal range of 2500m and a vertical range of 4000m.

[0078] The distance constraint condition of the wind farm layout optimization model is: the safe distance between two wind turbines is m times the rotor diameter of the wind turbine; where m takes the value of 3-5; in this embodiment, it is taken as 4 times.

[0079] S2. Pre-set a target value containing reference points and weight vectors based on the decision-maker's preference information; the specific steps for setting the reference points of the algorithm are as follows:

[0080] (1) Reference point express, The first target value of the reference point; The formula for calculating (corresponding power generation) is as follows:

[0081] =

[0082] in The wind farm layout scheme is obtained by traditional wind farm layout methods or by performing a certain number of iterations of the algorithm. The power generation corresponding to this layout scheme; coefficient In this example ;

[0083] (2) The second target value of the reference point These are noise limits determined based on the geographical location and environmental conditions of the wind farm.

[0084] (3) The weight vector is used to quantify the relative importance between power generation and noise. In this embodiment, the weight vector is set to [0.5, 0.5].

[0085] For two different numbers of wind turbines (N=16, 24), assuming the wind farms with these two turbine numbers are located in rural areas and residential areas near industrial zones, respectively, what are the noise limit values ​​for each? Set to 45dB and 50dB.

[0086] For example, if the algorithm is executed 200 times, and the negative AEP value of the resulting layout scheme is multiplied by 1.05, and used as the first target value for the reference point, the values ​​are -138.229 GWh and -207.350 GWh respectively. Then, the reference points for wind farms with the two different numbers of turbines are respectively... and ;

[0087] S3, the number of wind turbines in the wind farm is NUnder conditions such as 16 and 24, randomly generate Five (e.g., 50) wind farm layout schemes that meet the constraints constitute the parent population. P Each individual represents a layout scheme, and each individual... ;

[0088] S4. Parental population P Randomly divided into Groups (e.g., 5 groups) of subpopulations; perform a subpopulation search within each subpopulation to obtain the mutation vector; the subpopulation search uses the "NSDE / best / 1" neighborhood difference method:

[0089]

[0090] in For the mutation vector, The first vector represents the first vector. One dimension, ; The scaling factor is obtained through a neighborhood search strategy; , Each of these represents a different individual randomly selected from the current subpopulation; The best individual in the current subpopulation is obtained by sorting the subpopulation by non-dominated sorting and crowding distance sorting.

[0091] S5. Obtain candidate layout solutions by performing local difference based on mutation vectors:

[0092] The mutation vectors obtained in step S4 are re-encoded, treating the coordinates of each wind turbine as an independent entity, forming a scale of N A temporary population Q is obtained; a local difference operation is performed on population Q, and then the offspring population after the difference is re-encoded to obtain candidate layout solutions, corresponding to... Figure 3 Candidate solutions in: ( u 11 , u 12 , u 21 , u 22 , u 31 , u 32 ,…, u N1 , u N2 ).

[0093] S6. Based on the candidate layout solutions, perform a double substitution operation on the target vector to obtain the test vector. Composition includes The offspring population of an individual, such as Figure 3 As shown, the specific process is as follows:

[0094] S61. Execute the adaptive replacement operator. The specific process is as follows:

[0095] (1) The learning cycle is L If p is 150 times, and the number of evaluations is less than the learning period of the adaptive replacement operation, the coordinates of the first wind turbine in the candidate layout solution are used to randomly replace a coordinate in the target vector. The index of the replaced coordinate is... , The experimental vector is obtained. The "Success and Failure Memory" stores information about random replacements, including the following:

[0096] Replaced test vector If the constraints are met, record "State" = 1; otherwise, record "State" = 0.

[0097] For the replaced test vector Perform the "better-sort" operator operation: if after performing non-dominated sorting and crowding distance sorting with the corresponding parent individual, the parent individual can be Pareto dominated, or the parent individual is not dominated by the parent individual, then record "Better"=1; otherwise, "Better"=0.

[0098] (2) When the number of evaluations reaches or exceeds the learning period of the adaptive replacement operation, adaptive replacement is used to obtain the test vector. The adaptive replacement process is as follows:

[0099] Update the adaptive replacement information in the "success and failure memory" and calculate the trial vector. The success probability corresponding to all wind turbine coordinate indices is calculated, and then a random traversal sampling method is used to select the wind turbine coordinate index to be replaced based on the success probability information. , , i = 1, … , N The coordinates of the first wind turbine in the candidate layout solution are used to replace the index in the target vector. k The corresponding coordinates yield the experimental vector. ;

[0100] S62. Execute the repair and replacement operation operator. The specific process is as follows: find the test vector. =( x 11 , x 12 ,…, u 21 ,u 22 ,…, x i1 , x i2 ,…, x N1 , x N2 Coordinates that violate constraints in ) x i1 , x i2 ], and use candidate layout solutions ( u 11 , u 12 , u 21 , u 22 , u 31 , u 32 ,… , u N1 , u N2 The next coordinate in ) u 21 , u 22 Replace the test vector with ] Coordinates that violate constraints in [ x i1 , x i2 ];

[0101] Repeat the above process until the replacement is reached. p The coordinates are used to obtain the repaired test vector. =( x 11 , x 12 , …, u 21 , u 22 , …, u p1 , u p2 ,… , x N1 , x N2 ); p The value is N / 2.

[0102] S7. Update the population. The specific process is as follows:

[0103] Merge the offspring population and the parent population into one with The merged population of individuals Perform r-nondominated sorting, specifically as follows:

[0104] (1) Calculate the merged population All individuals With reference point Weighted Euclidean distance :

[0105] and

[0106] in It is the first The weights of each objective, M represents the number of optimization objectives. In this example... ; , For individuals In the The actual value on each target; For reference point In the The value on each target, and The first The maximum and minimum boundary values ​​of each target;

[0107] (2) Define the dominance relation of r: if any of the following conditions are true, then "rdominated" :

[0108] 1. Pareto Domination ;

[0109] 2. and They are mutually non-Pareto dominants and satisfy... ;

[0110]

[0111] in and They are two individuals in the population. Individuals and With reference point The weighted Euclidean distance; and They are respectively Maximum and minimum boundary values, For non-“r-dominated” thresholds; in this example The value is defined as: as the number of evaluations increases, The value decreases linearly from 1 to 0.1;

[0112] The hierarchy (stratification) of all individuals in a population is determined by the "r dominance" relationship.

[0113] (3) For non-“r-dominated” individuals at the same level, determine the ranking of each individual at the same level by sorting them by crowding distance;

[0114] (4) Lower non-“r-dominated” levels and greater crowding distances Each individual is preserved as a parent population for the next evolution. Specifically, starting from the lowest-level non-r-dominant layer, all layers are selected until a certain layer exceeds the total number after being added. At that time, within this layer, select a subset of individuals based on their crowding distance, from largest to smallest, to fill the gaps. .

[0115] S8. Continue the S4-S7 process on the population until the maximum number of evaluations (e.g., 150,000 times) is reached, and save the optimization results.

[0116] S9. The decision-maker determines whether to adjust the decision-maker's preference information based on the optimization result. If adjustment is needed, the reference point and / or weight vector are adjusted, and then the process returns to step S2 to start a new round of optimization. If no adjustment is needed, no iteration is required, and a wind farm layout scheme that conforms to the decision-maker's preference information is output.

[0117] Specifically, decision-makers analyze the optimization results of step S8 based on the following three aspects to determine whether it is necessary to return to step S2 to adjust preference information:

[0118] 1. The decision-maker compares the Pareto front solution in the optimization results with the preset reference point in step S2; if the difference between the two is greater than the preset threshold, it indicates that the current preference setting (expected value) conflicts with the feasible region of the actual problem, and the reference point needs to be adjusted so that the expectation is more in line with the actual optimization potential.

[0119] 2. Based on the convergence curve characteristics of each objective function value during the optimization process, analyze the optimization potential and convergence state of each objective;

[0120] 3. Based on the distribution pattern and range of Pareto front solutions in the target space, evaluate the diversity and trade-off characteristics of the solution set;

[0121] Based on the comprehensive analysis results of the above three aspects, determine the degree of matching between the preference direction defined by the current reference point and weight vector and the optimization result; if the degree of matching does not meet expectations, adjust the reference point and / or weight vector and return to step S2; otherwise, output the optimization result as the final wind farm layout.

[0122] like Figure 4As shown, after the first optimization layout of wind farms with two different numbers of wind turbines (N=16, 24), the target value of the resulting layout scheme is... and The obtained Pareto front solution and its corresponding reference point are as follows: Figure 4 As shown, the convergence curves of the power generation AEP and noise SPL values ​​of the layout scheme corresponding to the maximum power generation as the number of evaluations increases are as follows: Figure 5 As shown. By Figure 4 It can be seen that the reference point is set relatively high compared to the leading edge solution, and the target value of the reference point is set relatively high. Therefore, the reference point is adjusted to [-144.726, 43.380] and [-212.741, 47.500]. Figure 5 It can be seen that the power generation AEP value remains basically unchanged in the later stage, while the noise SPL value continues to decrease. Therefore, the weight of the noise target value is increased, and the weight vector is adjusted to [0.3,0.7] for the second optimization.

[0123] The target value of the layout scheme obtained in the second optimization is and The final frontier solution obtained is as follows Figure 6 As shown, the decision-maker believes that their preferred requirements have been met, and outputs the final layout scheme as follows. Figure 7 As shown.

[0124] The layout scheme obtained from the second optimization (e.g.) Figure 7 (As shown) Compared to the layout scheme obtained by the traditional layout method that only aims to maximize power generation (corresponding to) Figure 2 ):

[0125] 1. For example Figure 7 As shown in (a), relative to Figure 2 In (a), where The first wind farm with a noise level of 16 reduced noise by 23.04% and increased power generation by 4.09%, and within the pre-set reference point range, it met the decision-makers' preferred requirements.

[0126] 2. For example Figure 7 As shown in (b), relative to Figure 2 (b) The second wind farm with a noise level of 24 reduced noise by 10.45% and increased power generation by 3.03%, all within the pre-set reference range, meeting the decision-makers' preferred requirements.

[0127] This invention provides a wind farm layout optimization method that can interact with decision-makers to balance noise and power generation. First, wind energy data, wind farm and turbine parameters are acquired. A wind farm layout optimization model is established with the objectives of minimizing noise and maximizing power generation. Reference points and weight vectors for the algorithm are set according to the decision-maker's preferences for noise and power generation. Then, wind farm layout schemes are randomly generated, forming a system that includes… The parent population of each individual is randomly divided into several subpopulations. A subpopulation search is performed to obtain mutation vectors. Local differencing is applied to the mutation vectors to obtain candidate layout solutions. Based on these candidate layout solutions, [the following is a process / method / approach]... The target vectors are obtained by performing a double substitution operation. Each experimental vector is used to form a progeny population. Finally, the progeny population and the parent population are merged into one population, ranked, and the best progeny is selected. Each individual wind farm is preserved as a parent population for the next evolutionary iteration. The above steps are repeated for this population until the maximum number of evaluations is reached. The optimization results are then saved. The decision-maker uses these results to determine if adjustments to their preferences are needed. If adjustments are required, the weight vector of the algorithm's reference points is adjusted, and optimization is repeated until a final wind farm layout that meets the decision-maker's preferences is obtained. This invention optimizes wind farm layouts by balancing the goals of wind farm noise and power generation, and efficiently finds an optimal layout that meets the decision-maker's preferences. It reduces wind farm noise and increases power generation by minimizing wake effects compared to traditional layouts.

[0128] This invention uses a Gaussian function that closely matches actual wind turbine power curve data to represent the power curve, establishes an annual power generation model, and is based on actual wind farm parameters and ISO. The 9613-2:2024 standard defines wind farm layout optimization models that simulate real-world wind farm conditions. These models define areas susceptible to noise, such as residential areas, small industrial zones, or bird sanctuaries, as a series of discrete points placed at regular intervals along the wind farm boundary, plus a point at the wind farm center. The model aims to optimize wind farm layout by minimizing noise and power generation. A robust, efficient, and practical multi-objective evolutionary algorithm, which allows for setting predetermined target values ​​based on decision-maker preferences, is used for wind farm layout optimization. Considering the "curse of dimensionality" that can occur with increasing turbine numbers during optimization, a dual-replacement strategy is employed to obtain the experimental vectors. The adaptive replacement operation maintains global search capability while leveraging prior learning experience to find better solutions with the same number of evaluations, thus improving the algorithm's convergence speed and stability. Furthermore, considering the strict constraints of wind farm layout problems, the repair replacement operation accelerates computation and increases the quantity and quality of feasible solutions. A "pair-sorting" method is proposed to determine whether the experimental vectors obtained after the adaptive replacement operation are better, recording information on whether the experimental vectors are superior. Alternatively, the r-nondominated sorting method is used to sort the offspring population and the parent population, and the better individuals are selected to form a new parent population to continue the evolution. This can efficiently obtain a wind farm layout scheme that meets the predetermined target value, reduce noise and wake effect while increasing power generation.

[0129] The present invention also discloses a computer-readable storage medium having a computer program stored thereon, the computer program performing the steps of the method described above when run by a processor.

[0130] The present invention further discloses a wind farm layout optimization system that balances noise and power generation, including a memory and a processor connected to each other. The memory stores a computer program, which, when run by the processor, executes the steps of the method described above.

[0131] The medium and system of the present invention, corresponding to the methods described above, also have the advantages described above.

[0132] The present invention can implement all or part of the processes in the methods of the above embodiments, or it can be implemented by hardware related to computer program instructions. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of the above method embodiments. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable storage medium includes: any entity or device capable of carrying computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. The memory is used to store computer programs and / or modules. The processor implements various functions by running or executing the computer programs and / or modules stored in the memory, and by calling data stored in the memory. The memory may include high-speed random access memory, as well as non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart media cards (SMC), secure digital (SD) cards, flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0133] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A method for optimizing wind farm layout to balance noise and power generation, characterized in that, Including the following steps: S1. Obtain wind energy data, wind farm and turbine parameters, and establish a multi-objective optimization model for wind farm layout with the objectives of minimizing noise and maximizing power generation. The variables of the multi-objective optimization model are the coordinates of each turbine, and the outputs are the estimated annual power generation and noise of the wind farm. S2. Set a predetermined target value containing a reference point and a weight vector based on the decision preference information; wherein the reference point contains the expected target values ​​of power generation and noise, and the weight vector is used to quantify the relative importance between power generation and noise. S3, Randomly generated A total of several wind farm layout schemes that satisfy the boundary constraints and the safe distance constraints between wind turbines constitute the parent population; each wind farm layout scheme is an individual. S4. The parent population is randomly divided into several subpopulations. Within each subpopulation, a differential strategy based on neighborhood search is used to search and generate the corresponding mutation vector. S5. Perform a local difference operation on the mutation vector to obtain candidate layout solutions; S6. Based on the candidate layout solution, perform adaptive replacement and repair replacement operations sequentially on each target vector in the current parent population to generate... Each experimental vector constitutes the offspring population; S7. Merge the offspring population with the parent population to obtain a merged population; rate and rank the individuals in the merged population based on a predetermined target value, and select the optimal individuals. Each individual becomes a parent population for the next generation of evolution; S8. Repeat steps S4 to S7 until the maximum number of evaluations is reached, and output the optimization results. S9. Determine whether the decision-maker's preference information needs to be adjusted based on the optimization results. If adjustment is required, update the reference point and / or weight vector, and return to step S2 to execute the optimization process. If no adjustments are needed, output the final wind farm layout scheme; In step S7, individuals in the merged population are rated and ranked based on a predetermined target value, and the optimal individuals are selected. The specific process for each individual is as follows: Calculate the relationship between all individuals in the merged population and the reference point. The weighted Euclidean distance; Define the dominance relation r: if Pareto Domination ;or and They are mutually non-Pareto dominants and satisfy... , Then r dominates ;in Let be a function defined based on weighted Euclidean distance; where and They are two individuals in the population. For non-r-dominated thresholds, ; The hierarchy of all individuals in the merged population is determined by the dominance relationship of r; Non-r-dominated individuals at the same level are sorted from largest to smallest by crowding distance; Start by selecting all non-dominated layers from the lowest level until a certain layer is added, exceeding the total number. At that time, within this layer, select a subset of individuals based on their crowding distance, from largest to smallest, to fill the gaps. .

2. The wind farm layout optimization method for balancing noise and power generation according to claim 1, characterized in that, Step S1 specifically includes: S11. Acquire and preprocess wind energy data, wind farm and turbine parameters, and establish wind speed probability distribution model, wake model and turbine power curve model. S12. Based on the wind speed probability distribution model, wake model and wind turbine power curve model, calculate the expected power output function of each wind turbine, and then sum them to obtain the annual power generation of the wind farm. Take the negative value of the annual power generation as the first objective of the multi-objective optimization model. S13. Based on wind farm parameters and acoustic standards, establish a two-dimensional wind farm acoustic model, take a series of preset points at the wind farm boundary and center as noise receivers, calculate the maximum noise value among all noise receivers, and take it as the second objective of the multi-objective optimization model.

3. The wind farm layout optimization method for balancing noise and power generation according to claim 2, characterized in that, Annual power generation in step S12 The calculation formula is: in For wind turbine i The expected power output value, i =1,…, N， in N This refers to the number of wind turbines in the wind farm. Wind direction The probability density function, For the first i Typhoon generators at wind speed Power output at that time; In the wind direction wind speed The Weibull probability distribution model; Wind direction Scaling parameters considering wake effect, Wind direction The shape parameters below.

4. The wind farm layout optimization method for balancing noise and power generation according to claim 1, 2, or 3, characterized in that, In step S4, the difference strategy based on neighborhood search is as follows: in For the mutation vector, The first vector represents the first vector. One dimension, This is the scaling factor; , Each represents a different individual randomly selected from the current subpopulation. It is the best individual in the current subpopulation.

5. The wind farm layout optimization method for balancing noise and power generation according to claim 1, 2, or 3, characterized in that, In step S5, the specific process of performing a local difference operation on the mutation vector to obtain the candidate layout solution is as follows: The mutation vector is re-encoded, and the coordinates of each wind turbine are treated as an independent individual, forming a temporary population of size N. The temporary population is then subjected to local differencing, and the differencing population is re-encoded again to obtain candidate layout solutions.

6. The wind farm layout optimization method for balancing noise and power generation according to claim 1, 2, or 3, characterized in that, In step S6, the specific process of performing adaptive replacement and repair replacement operations sequentially on each target vector in the current parent population is as follows: S61. Adaptive replacement operation: During the algorithm learning cycle, coordinates in the target vector are randomly selected for replacement and the replacement success rate information is recorded; after the learning cycle, based on the accumulated success rate information, a random traversal sampling method is used to determine the index of the replaced coordinates to improve search efficiency. S62. Repair and Replacement Operation: Detect the wind turbine coordinates that violate the distance constraint in the test vector obtained after the adaptive replacement operation, and replace them sequentially with the subsequent coordinates in the candidate layout solution until the constraint condition is met or the maximum number of replacements is reached.

7. The wind farm layout optimization method for balancing noise and power generation according to claim 1, characterized in that, The expression is: in and They are two individuals in the population. and They are weighted Euclidean distances. The maximum and minimum boundary values; Individuals and With reference point The weighted Euclidean distance.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program, when run by a processor, performs the steps of the method as described in any one of claims 1-7.

9. A wind farm layout optimization system for balancing noise and power generation, comprising an interconnected memory and a processor, wherein the memory stores a computer program, characterized in that, The computer program, when run by a processor, performs the steps of the method as described in any one of claims 1-7.

Images