Layout method, device and readable medium for offshore wind farm of single-type wind turbine

Abstract

This invention discloses a method, equipment, and readable medium for offshore wind farm layout targeting a single type of wind turbine. Belonging to the field of offshore wind farm planning, design, and intelligent optimization algorithm technology, this invention focuses on minimizing the levelized cost of electricity (LCoE) over its entire lifecycle. It employs a Gaussian wake model for accurate and efficient evaluation of wake effects. A soft-constraint penalty mechanism integrates wind farm engineering constraints into the objective function, preserving gradient information to explore high-quality solutions at the feasible region boundary. A multi-strategy guided mutation operator incorporating engineering knowledge is designed, coupled with an active population diversity restoration mechanism, to achieve full-process optimization of wind turbine layout. This invention, using the aforementioned method, equipment, and readable medium for offshore wind farm layout targeting a single type of wind turbine, can improve algorithm convergence speed and search efficiency, adapt to irregular site boundaries in offshore wind farms, and enhance the overall lifecycle economics of offshore wind farms.

Description

Technical Field

[0001] This invention belongs to the field of offshore wind farm planning and design and intelligent optimization algorithm technology, specifically involving offshore wind farm layout methods, equipment and readable media for single-type wind turbines. Background Technology

[0002] Wind farm layout design is a core element determining the economic feasibility of offshore wind power projects throughout their entire lifecycle. During the planning phase before construction, adapting to different wind turbine types and determining their optimal spatial arrangement to minimize wake losses is crucial for maximizing wind farm energy output and achieving a return on investment. The core of wind farm layout optimization revolves around three fundamental dimensions: wake effect assessment, optimization objective setting, and optimization algorithm design.

[0003] However, existing technologies have the following shortcomings: First, existing algorithms generally use the death penalty method to handle engineering constraints such as boundary constraints and minimum turbine spacing constraints in offshore wind farms. Although this method is computationally simple, it artificially compresses the algorithm's search space and is prone to losing high-quality feasible solutions containing important genetic information near the constraint boundary. While constraint handling schemes such as the penalty function method and the Lagrange multiplier method have theoretical advantages, they are extremely difficult to apply in practical engineering for the irregular and non-convex site boundaries commonly found in offshore wind farms, making them difficult to implement. Second, traditional genetic algorithms use general random mutation operators and do not incorporate engineering domain knowledge of wind farm layout optimization. They cannot address core issues such as turbine spacing violations, coordinate out-of-bounds errors, and over-concentration of the layout, and can only complete the search through random trial and error. This results in slow convergence speed, low search efficiency, and difficulty in fully exploring the complex non-convex solution space of layout optimization. Third, existing algorithms lack proactive population diversity management mechanisms. As iterative evolution progresses, genotypic similarity within the population increases rapidly, leading to a rapid loss of genetic diversity. This makes them prone to premature convergence and getting trapped in local optima, failing to guarantee that the final output layout scheme is the globally optimal solution. Furthermore, most existing layout optimization schemes are based on regular grid layout designs, which can only adapt to regular rectangular site boundaries. They cannot cope with irregular non-convex boundaries formed by navigation channel avoidance, ecological protection zones, and underwater geological conditions in offshore wind farms. The optimization results are difficult to meet the compliance requirements for engineering implementation, and they have not been systematically adapted to the layout characteristics of single-type offshore wind turbines, significantly reducing the optimization effect of generalized schemes.

[0004] Therefore, a new method is urgently needed. Summary of the Invention

[0005] The purpose of this invention is to provide a method, equipment, and readable medium for offshore wind farm layout for single-type wind turbines. This method can improve the convergence speed and search efficiency of the algorithm, adapt to the irregular site boundaries of offshore wind farms, and improve the economic efficiency of offshore wind farms throughout their entire life cycle.

[0006] To achieve the above objectives, the present invention provides a method, apparatus, and readable medium for offshore wind farm layout for a single type of wind turbine, comprising the following steps: S1. Based on wind farm engineering data, complete the initial population construction and fitness assessment; output the initial population, standardized engineering data, and fitness assessment results; S2. Obtain the initial population and fitness assessment results output by S1, or the next-generation population and fitness data of S8 iteratively returned; use the biased tournament selection strategy to screen parent individuals and generate a mating pool; output the mating pool to S3. S3. Obtain the mating pool output from S2; monitor genotype similarity periodically, and perform diversity repair if the conditions are met; output the repaired mating pool to S4. S4. Obtain the repaired mating pool output from S3; use a uniform crossover operator to complete gene recombination and generate offspring population; output the offspring population to S5. S5. Obtain the offspring population output from S4 and the standardized engineering data output from S1; perform multi-strategy mutation according to the mutation probability; output the mutated offspring population to S6. S6. Obtain the mutated offspring population output from S5 and the standardized engineering data output from S1; complete the population uniqueness screening and parallel fitness evaluation, calculate and assign the penalized levelized electricity cost; output the offspring population and fitness array to S7. S7. Obtain the offspring population and fitness array, and the current parent population output from S6; merge the populations and perform elite retention and random sampling selection to generate a new generation population; output the new generation population and fitness data to S8; S8. Obtain the next generation population and fitness data output by S7; update the global best individual, record the optimization data, and determine the convergence condition; if convergence fails, output the next generation population and fitness data and send it back to S2; if convergence occurs, output the global best individual to S9. S9. Obtain the globally optimal individual output from S8 and the standardized engineering data output from S1; perform coordinate decoding to complete full index review and compliance verification; output the final wind turbine layout scheme and performance indicators.

[0007] Preferably, in S2, the execution rules of the bias tournament selection strategy are as follows: a preset number of individuals are randomly selected from the current population to form a competition group, the individual with the best fitness in the group is selected with an 80% probability, and the individual in the group is randomly selected with a 20% probability. This process is repeated until a mating pool of the same size as the initial population is generated.

[0008] Preferably, in step S3, genotype similarity is monitored periodically, and diversity repair is performed if certain conditions are met, specifically as follows: Similarity is determined by calculating the Euclidean distance between genotypes of individuals. When the proportion of duplicate or highly similar individuals in the population is less than 20%, a diversity restoration operation is triggered, and Gaussian or uniform mutation is performed on the corresponding individuals, with directional adjustments guided by the current global optimal solution.

[0009] Preferably, in step S5, multi-strategy mutation is performed according to the mutation probability, specifically as follows: The preset multi-strategy mutation pool includes the following 5 targeted strategies: Gaussian mutation for local fine-tuning, uniform mutation for global search, spacing repair mutation for correcting wind turbine spacing violations, dispersed layout mutation for reducing layout clustering, and boundary guidance mutation for pulling out-of-bounds wind turbines back to the compliant area.

[0010] Preferably, in step S6, the formula for calculating the punitive levelized cost of electricity (LCOE) is: ; In the formula, The punitive levelized cost of electricity, i.e., the individual fitness value; Penalties for crossing the boundary; This is a penalty item for violating the minimum spacing requirement for wind turbines; Based on the levelized cost of electricity.

[0011] Preferably, the formula for calculating the boundary violation penalty term is: = ; In the formula, This is the boundary penalty coefficient; For the first The distance the typhoon turbine crosses the site boundary; this value is 0 when the turbine is located within the compliant area. For serial number subscripts; This represents the total number of single-type fans. The formula for calculating the penalty for violations of the minimum spacing between wind turbines is as follows: = ; In the formula, This is the spacing penalty coefficient; For the first Taiwan and the The value of non-compliance in typhoon turbine spacing is 0 when the turbine spacing meets the minimum safety distance requirement; This is a serial number subscript.

[0012] Preferably, in step S7, the combined selection rule for elite retention and random sampling is as follows: The parent and offspring populations are merged to form a mixed pool. Individuals are ranked from best to worst fitness, and the top 50% of the best individuals are selected as elites and directly retained to the next generation. The remaining 50% of the population positions are filled by random sampling from non-elite offspring individuals.

[0013] Preferably, S1 specifically includes: Configure the core operating parameters of the algorithm, load and standardize the basic data of wind farm engineering, use real number coding to complete the coding of single-type wind turbine layout schemes and construct the initial population, and complete the fitness evaluation of the initial population based on the Gaussian wake model. The encoding formula for real numbers is: ; In the formula, The wind turbine layout encoding vector for a single individual; For the first Normalized plane coordinates corresponding to the typhoon generator; The decoding formula for normalized coordinates and actual engineering coordinates is: ; In the formula, For the first Actual engineering plane coordinates of the typhoon generator; These represent the minimum and maximum values ​​of the X-axis coordinates of the wind farm site; These represent the minimum and maximum values ​​of the Y-axis coordinate of the wind farm site; The formula for calculating the wake velocity deficit in the Gaussian wake model is: ; ; In the formula, This represents the wind speed loss due to the wake. The initial wind speed of the incoming flow; This is the thrust coefficient of the fan at the corresponding inflow velocity. The diameter of the wind turbine rotor; The wake spread radius; The wake spread ratio; This represents the axial distance between the downstream and upstream wind turbines. This is the lateral distance between the center axes of the wakes of the downstream and upstream wind turbines.

[0014] Therefore, the present invention employs the above-mentioned offshore wind farm layout method, equipment, and readable medium for single-type wind turbines. Compared with the prior art, the technical solution of the present invention has the following beneficial effects: (1) This invention takes minimizing LCoE as its core objective and achieves a significant reduction in levelized cost of electricity (LCOE) through layout optimization. For four different models of single-type offshore wind turbines with a target installed capacity of 300MW, compared with the conventional grid layout baseline scheme, LCoE reductions of 2.5%, 0.9%, 2.0%, and 2.1% were achieved, respectively; compared with the traditional genetic algorithm optimization scheme, the LCoE was further reduced by 0.003 yuan / kWh under the optimal operating condition, which greatly improved the economic benefits and return on investment of offshore wind power projects throughout their entire life cycle. (2) This invention minimizes wake interference between wind turbines through accurate evaluation and layout optimization using the Gaussian wake model. In the four wind turbine optimization cases, the wake loss of the optimal layout is reduced to as low as 1.8%, which is much lower than the 7.1%, 6.1%, 3.9%, and 4.1% of the traditional grid layout scheme. At the same time, the annual power generation AEP is significantly improved. Among them, the AEP of the optimal layout of the MySE6.25-180 wind turbine is about 2.65% higher than that of the grid layout scheme, which effectively improves the energy utilization efficiency and annual production capacity of the wind farm. (3) This invention achieves a comprehensive upgrade in algorithm performance through multi-strategy guided mutation operators, soft constraint handling mechanisms, and proactive diversity repair. Compared with traditional genetic algorithms, this invention significantly improves convergence speed and can reach near-optimal solutions in fewer iterations. At the same time, through full-process algorithm architecture optimization, it achieves a precise balance between global exploration and local development, effectively avoiding premature convergence and getting trapped in local optima. It can fully explore complex non-convex solution spaces and output globally optimal layout schemes. In addition, the fitness evaluation strategy of deduplication and parallel computation significantly reduces the amount of redundant computation and improves computational efficiency in large-scale optimization scenarios. (4) This invention is specifically designed for engineering scenarios of single-type offshore wind turbines. It adopts a continuous coding method for wind turbine coordinates, which breaks through the limitations of traditional regular grid layout. It can perfectly adapt to the irregular non-convex site boundaries of offshore wind farms due to channel avoidance, environmental protection zone restrictions, and geological conditions. The optimization process fully incorporates engineering constraints such as wind farm boundaries and minimum wind turbine spacing. The output optimal layout scheme can meet the engineering compliance requirements and can be directly applied to the early planning and design of offshore wind power projects, replacing the traditional experience-based layout scheme. (5) The optimized layout of this invention achieves a balanced distribution of power generation among wind turbines, effectively avoiding disproportionate performance degradation of a single wind turbine due to wake interference. Taking the MySE12-242 wind turbine as an example, the minimum annual power generation of a single unit after optimization is increased by 0.74% and the maximum annual power generation is increased by 0.38% compared with the traditional genetic algorithm scheme; at the same time, the reduction in wake turbulence intensity reduces fatigue damage to the wind turbine, effectively extends the service life of the wind turbine, fundamentally reduces the long-term operation and maintenance costs of the project, and maximizes the comprehensive benefits of the wind farm throughout its entire life cycle.

[0015] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the implementation of the offshore wind farm layout method for a single type of wind turbine according to the present invention. Figure 2 This is a comparison curve of the iterative convergence characteristics of the penalized genetic algorithm and the original genetic algorithm for the MySE6.25-180 wind turbine in Embodiment 1 of the offshore wind farm layout method for a single type of wind turbine of the present invention. Figure 3 This is a comparison curve of the iterative convergence characteristics of the penalized genetic algorithm and the original genetic algorithm for the EN-226-8.5 wind turbine in Embodiment 1 of the offshore wind farm layout method for a single type of wind turbine of the present invention. Figure 4 This is a comparison curve of the iterative convergence characteristics of the penalized genetic algorithm and the original genetic algorithm for the MySE12-242 wind turbine in Embodiment 1 of the offshore wind farm layout method for a single type of wind turbine of the present invention. Figure 5 This is a comparison curve of the iterative convergence characteristics of the penalized genetic algorithm and the original genetic algorithm for the EN-252-14 wind turbine in Embodiment 1 of the offshore wind farm layout method for a single type of wind turbine of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Unless otherwise defined, the technical or scientific terms used in the present invention should have the ordinary meaning understood by those skilled in the art.

[0018] Example 1 like Figure 1 As shown, this embodiment provides a method, equipment, and readable medium for offshore wind farm layout for a single type of wind turbine. It should be understood that the specific parameters, models, and protocols mentioned in this embodiment are merely examples to help those skilled in the art understand the present invention, and are not intended to limit the present invention.

[0019] The present invention provides a method, apparatus, and readable medium for offshore wind farm layout for a single type of wind turbine, comprising the following steps: S1. First, complete the configuration of key algorithm parameters, such as population size of 50~200 (preferably 100), maximum number of iterations of 100~500 (preferably 300), crossover probability of 0.6~0.9 (preferably 0.8), mutation probability of 0.05~0.2 (preferably 0.1), genotype similarity threshold of 15%~25% (preferably 20%), boundary penalty coefficient of 10~100 (preferably 50), spacing penalty coefficient of 10~100 (preferably 50), tournament size of 3~5 (preferably 3), and other core operating parameters.

[0020] Simultaneously, the data on wind farm site boundaries, wind energy resources statistics, technical parameters of single-type wind turbines, and project life cycle cost parameters are loaded and standardized. Extreme value normalization processing is performed on the site coordinates to establish a conversion benchmark between normalized coordinates and actual engineering coordinates.

[0021] Each wind turbine layout scheme is encoded using real-number encoding as a normalized coordinate vector with a length equal to twice the number of wind turbines. The encoding format is as follows: ; In the formula, The wind turbine layout encoding vector for a single individual; This represents the total number of single-type fans. For the first Normalized plane coordinates corresponding to the typhoon generator.

[0022] The normalized coordinates are restored to the actual planar coordinates of the wind farm using a preset decoding formula, which is as follows: ; In the formula, For the first Actual engineering plane coordinates of the typhoon generator; These represent the minimum and maximum values ​​of the X-axis coordinates of the wind farm site; These represent the minimum and maximum values ​​of the Y-axis coordinate of the wind farm site; This is a serial number subscript.

[0023] Based on this, a hybrid strategy is adopted to construct the initial population. One part uses a heuristic approach to generate feasible individuals that meet the boundary and spacing constraints, another part is implanted with expert experience to lay out seeds to provide high-quality prior information, and the remaining individuals are generated using randomized normalized coordinates to balance the feasibility and diversity of the initial population.

[0024] After deduplication of the initial population, the initial fitness assessment was completed using parallel computing. The inflow velocity of the fan was calculated based on the Gaussian wake model, and the wake velocity deficit calculation formula is as follows: ; ; In the formula, This represents the wind speed loss due to the wake. The initial wind speed of the incoming flow; This is the thrust coefficient of the fan at the corresponding inflow velocity. The diameter of the wind turbine rotor; The wake spread radius; The wake spread ratio; This represents the axial distance between the downstream and upstream wind turbines. The lateral distance between the wake center axes of the downstream and upstream fans is [value missing]; the actual inflow velocity of the downstream fan is [value missing]. The superposition of wakes from multiple fans is calculated using the energy conservation method.

[0025] The annual power generation of a wind farm can be obtained by combining statistical values ​​of wind direction and wind speed distribution. The calculation formula is as follows: ; In the formula, This refers to the annual power generation of the wind farm. The total number of hours throughout the year; This represents the total number of wind direction and wind speed operating conditions. For the first The probability of occurrence of a specific wind direction and wind speed condition; For the first The total power generation of all wind turbines in a wind farm under a specific operating condition; This is a serial number subscript.

[0026] The basic levelized cost of electricity (LCOE) is further calculated using the following formula: ; In the formula, Based on the levelized cost of electricity; For the initial capital expenditure of the project; For the project number Annual operating and maintenance costs; The project's entire lifecycle of operation; The benchmark discount rate; For the project number Annual power generation; Use year subscripts.

[0027] The violation rates of boundary constraints and spacing constraints are transformed into continuous penalty terms, and a penalty-levelized cost of electricity is constructed as the individual fitness, calculated as follows: ; In the formula, The punitive levelized cost of electricity, i.e., the individual fitness value; Penalties for crossing the boundary; This is a penalty item for violating the minimum spacing requirement for wind turbines; Based on the levelized cost of electricity.

[0028] The formula for calculating the boundary violation penalty is as follows: = ; In the formula, This is the boundary penalty coefficient; For the first The distance the typhoon turbine crosses the site boundary; this value is 0 when the turbine is located within the compliant area.

[0029] The formula for calculating the penalty for violations of minimum spacing between wind turbines is as follows: = ; In the formula, This is the spacing penalty coefficient; For the first Taiwan and the The value for non-compliance with typhoon fan spacing is 0 when the fan spacing meets the minimum safety distance requirement.

[0030] The fitness values ​​for the initial population are assigned. The initial population, individual fitness arrays, and standardized basic database generated in this step will serve as input data for subsequent selection operations.

[0031] S2. Using the initial population and fitness data output by S1, or the new generation population and fitness data returned by S8 during the iteration process as input, a biased tournament selection strategy is used to complete the selection of parent individuals: A number of individuals are randomly selected from the current population to form a competition group. The individual with the best fitness in the group is selected with a probability of 70% to 90% (preferably 80%). Individuals in the group are randomly selected with a probability of 10% to 30% (preferably 20%). This process is repeated until a mating pool of the same size as the initial population is generated, and then the result is output to the next stage.

[0032] S3: Receive the mating pool output from S2 as the processing target, and perform population diversity monitoring and active repair according to the set iteration cycle. Similarity is determined by calculating the Euclidean distance between genotypes of individuals. When the proportion of duplicate or highly similar individuals in the population is less than 15%~25% (preferably 20%), a diversity restoration operation is triggered. Gaussian or uniform mutation operations are performed on the corresponding individuals, and directional adjustments are made based on the current global optimum. Before the crossover operation, the genetic diversity of the population is replenished to avoid premature convergence and optimization stagnation due to diversity loss. The mating pool after restoration is output to the next stage.

[0033] S4. Using the repaired mating pool output from S3 as the processing target, the uniform crossover operator is used to achieve unbiased recombination of parental genes: Individuals in the mating pool are randomly paired according to the set crossover probability. Each dimension of the gene of each paired parent individual is inherited with equal probability, so that the offspring individuals can freely combine the layout information from different parents and fully expand the scope of solution space exploration. The offspring population formed after the crossover is completed is output to the next stage.

[0034] S5 receives the offspring population output by S4 and the standardized basic data provided by S1. The standardized basic data includes engineering constraint information such as site boundaries and minimum spacing between wind turbines. Guided multi-strategy mutation is performed according to the set mutation probability.

[0035] The mutation strategy pool contains five targeted optimization strategies. Each mutation randomly selects one strategy from the pool for execution, ensuring that the mutation operation closely adheres to the constraints of the wind farm project. This enables targeted correction of violations and efficient local search. The five strategies are: 1. Gaussian variation for local fine-tuning; 2. Uniform mutation for global search; 3. Spacing repair variants used to correct spacing violations; 4. Variations in the decentralized layout used to reduce the concentration of wind turbines; 5. Boundary guidance variation used to pull wind turbines that have overstepped their boundaries back to compliant areas.

[0036] The mutated offspring population is then sent to the next stage.

[0037] S6. Using the mutated offspring population output from S5 and the standardized basic data from S1 as input, the offspring population is first screened for uniqueness, and duplicate genotypes are removed to avoid redundant calculations.

[0038] Then, a parallel computing architecture is used to perform fitness evaluation on the unique individual. The core process is as follows: The wake effect is calculated based on the Gaussian wake model, and the inflow wind speed of each wind turbine and the annual power generation AEP of the wind farm are obtained. The basic levelized cost of electricity (LCoE) is calculated by combining the project investment cost and operation and maintenance cost. The penalty terms of boundary and spacing soft constraints are superimposed to obtain the penalty levelized cost of electricity for the offspring individuals, and the fitness is assigned.

[0039] The calculation formula is: ; After the evaluation is completed, the offspring population and its fitness array are output to the next stage.

[0040] S7. Simultaneously receive the previous generation parent population and the offspring population output from S6, merge the parent and offspring populations to form a mixed pool, sort them from best to worst fitness, select 40%~60% (preferably 50%) of the best individuals as elites and directly retain them to the next generation, and randomly sample the remaining population positions from the non-elite offspring individuals.

[0041] This rule ensures the stable inheritance of historical optimal solutions while continuously introducing new genetic material into the population; the new generation of population formed through selection is then output to the next stage.

[0042] S8 takes the new generation population output by S7 and the current global optimal solution as input, traverses the new generation population to update the global optimal individual and its corresponding layout coordinates, fitness, power generation, cost per kilowatt-hour and other key indicators, and records the optimization process data at fixed intervals to achieve process visualization monitoring.

[0043] The algorithm is then judged to meet the convergence condition. If the maximum number of iterations has not been reached and there is still room for optimization of the optimal fitness, it is judged as non-converged, and the new generation of population is sent back to the S2 bias tournament selection stage to continue the iterative loop. If the maximum number of iterations has been reached, or the optimal value has not changed significantly for several generations, the algorithm is judged as converged, and the final global optimal individual is output to the final stage.

[0044] S9. Based on the global optimal individual output by S8 and the basic engineering data of S1, perform coordinate decoding operation on the optimal individual to convert the normalized coordinates into the actual engineering coordinates of the wind farm, forming a wind turbine spatial layout scheme that can be directly used for engineering design.

[0045] Based on this, a full-indicator review calculation is performed on the optimal layout to obtain core performance data such as levelized cost of electricity, annual power generation, total project cost, wake loss rate, capacity factor, and constraint satisfaction, thus completing the layout compliance verification.

[0046] The final output includes a wind turbine coordinate table, a layout plan, an optimized convergence curve, and a complete optimization report, forming the final layout results that meet the planning requirements of offshore wind farm projects.

[0047] like Figures 2-5 As shown, in order to verify the engineering adaptability, optimization effect and algorithm performance of the method of the present invention, a systematic verification experiment was carried out for an offshore wind farm scenario with a target installed capacity of a preset value and irregular non-convex boundaries formed by constraints such as waterway avoidance, ecological protection zones and underwater geological conditions.

[0048] The experiment sets up multiple optimization cases for single-type wind turbines, each corresponding to a different specification of commercial single-type offshore wind turbine. The number of turbines is determined based on a preset target installed capacity. Each case set up three parallel optimization schemes: the enhanced genetic algorithm (penalized genetic algorithm) proposed in this invention; the baseline scheme of the mainstream traditional genetic algorithm in existing technology; and the standard grid layout benchmark scheme in the offshore wind power industry.

[0049] The comparison dimensions are divided into two categories: the first is the technical indicators of the optimization effect of the scheme, including the levelized cost of electricity (LCoE), annual power generation (AEP), and the satisfaction of engineering constraints; the second is the algorithm performance indicators, including the convergence speed, the quality of the final solution, and the ability to maintain population diversity throughout the entire evolutionary process.

[0050] This experiment is processor-based and utilizes a parallel computing architecture to accelerate the fitness evaluation process. It adopts a deduplication-before-evaluation execution strategy: first, duplicate genotypes within the population are identified, fitness evaluation is performed only once for unique genotypes, and then the results are mapped to all duplicate individuals. This fully utilizes the multi-core architecture to improve computational efficiency while minimizing redundant computation.

[0051] As shown in Table 1, the experimental results show that, in all test cases, the punitive genetic algorithm proposed in this invention is significantly better than the two industry-standard benchmark layout schemes of grid layout and staggered layout in terms of optimization effect, algorithm performance and engineering adaptability, and its overall performance also surpasses that of traditional genetic algorithms.

[0052] The penalized genetic algorithm converges significantly faster than the traditional genetic algorithm, reaching a near-optimal solution in fewer iterations and maintaining superior solution quality throughout the entire evolutionary process. Its soft constraint handling mechanism effectively explores high-quality layout solutions in the vicinity of the site boundary. In all test cases, the layout schemes output by the penalized genetic algorithm strictly satisfy engineering constraints such as site boundaries and minimum turbine spacing, and can adapt to irregular, non-convex site boundary shapes. Compared to grid layout schemes, staggered layout schemes, and traditional genetic algorithms, the layout obtained by this algorithm significantly reduces wake loss, and simultaneously improves annual power generation and computational efficiency, demonstrating excellent optimization performance and engineering practicality.

[0053] Table 1. Comprehensive Comparison of Optimization Results

[0054] Example 2 The difference between this embodiment and Embodiment 1 lies in the simplification of the multi-strategy mutation pool. Only three core directional mutation strategies are retained: Gaussian mutation, spacing-fixing mutation, and boundary-guided mutation. Uniform mutation and dispersed layout mutation are removed. The remaining steps and parameter settings are the same as in Embodiment 1. This modified example can further reduce the computational complexity of the algorithm and improve the running efficiency while ensuring the core optimization effect.

[0055] Example 3 The difference between this embodiment and Embodiment 1 is that the biased tournament selection strategy is replaced with a selection strategy combining roulette wheel selection and elite retention. All other steps, models, and parameter settings are the same as in Embodiment 1. This variation can adapt to the selection needs of different population evolution stages, expanding the algorithm's applicable scenarios.

[0056] Therefore, the present invention adopts the above-mentioned offshore wind farm layout method, equipment and readable medium for single-type wind turbines. This method can improve the algorithm convergence speed and search efficiency, adapt to the irregular site boundaries of offshore wind farms, and improve the economic efficiency of offshore wind farms throughout their entire life cycle.

[0057] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for offshore wind farm layout oriented towards a single type of wind turbine, characterized in that, Includes the following steps: S1. Based on wind farm engineering data, complete the initial population construction and fitness assessment; output the initial population, standardized engineering data, and fitness assessment results; S2. Obtain the initial population and fitness assessment results output by S1, or the next-generation population and fitness data of S8 iteratively returned; use the biased tournament selection strategy to screen parent individuals and generate a mating pool; output the mating pool to S3. S3, Obtain the mating pool output by S2; Genotype similarity is monitored periodically, and diversity restoration is performed if conditions are met; the restored mating pool is then output to S4. S4. Obtain the repaired mating pool output by S3; Gene recombination is performed using a uniform crossover operator to generate a progeny population; the progeny population is then output to S5. S5. Obtain the offspring population output from S4 and the standardized engineering data output from S1; perform multi-strategy mutation according to the mutation probability; output the mutated offspring population to S6. S6. Obtain the mutated offspring population output from S5 and the standardized engineering data output from S1; complete the population uniqueness screening and parallel fitness evaluation, calculate and assign the penalized levelized electricity cost; output the offspring population and fitness array to S7. S7. Obtain the offspring population and fitness array, and the current parent population output from S6; merge the populations and perform elite retention and random sampling selection to generate a new generation population; output the new generation population and fitness data to S8; S8. Obtain the next generation population and fitness data output by S7; update the global best individual, record the optimization data, and determine the convergence condition; If the convergence fails, output the next generation population and fitness data and send it back to S2; if the convergence occurs, output the globally optimal individual to S9. S9. Obtain the globally optimal individual output from S8 and the standardized engineering data output from S1; perform coordinate decoding to complete full index review and compliance verification; output the final wind turbine layout scheme and performance indicators.

2. The offshore wind farm layout method for a single type of wind turbine according to claim 1, characterized in that, In S2, the execution rules of the biased tournament selection strategy are as follows: a preset number of individuals are randomly selected from the current population to form a competition group. The individual with the best fitness in the group is selected with an 80% probability, and an individual in the group is randomly selected with a 20% probability. This process is repeated until a mating pool of the same size as the initial population is generated.

3. The offshore wind farm layout method for a single type of wind turbine according to claim 2, characterized in that, In S3, genotype similarity is monitored periodically, and diversity restoration is performed when conditions are met, specifically as follows: Similarity is determined by calculating the Euclidean distance between genotypes of individuals. When the proportion of duplicate or highly similar individuals in the population is less than 20%, a diversity restoration operation is triggered, and Gaussian or uniform mutation is performed on the corresponding individuals, with directional adjustments guided by the current global optimal solution.

4. The offshore wind farm layout method for a single type of wind turbine according to claim 3, characterized in that, In S5, multi-strategy mutation is performed according to the mutation probability, specifically as follows: The preset multi-strategy mutation pool includes the following 5 targeted strategies: Gaussian mutation for local fine-tuning, uniform mutation for global search, spacing repair mutation for correcting wind turbine spacing violations, dispersed layout mutation for reducing layout clustering, and boundary guidance mutation for pulling out-of-bounds wind turbines back to the compliant area.

5. The offshore wind farm layout method for a single type of wind turbine according to claim 4, characterized in that, In S6, the formula for calculating the punitive levelized cost of electricity is as follows: ； In the formula, The punitive levelized cost of electricity, i.e., the individual fitness value; Penalties for crossing the boundary; This is a penalty item for violating the minimum spacing requirement for wind turbines; Based on the levelized cost of electricity.

6. The offshore wind farm layout method for a single type of wind turbine according to claim 5, characterized in that, The formula for calculating the boundary violation penalty is as follows: = ； In the formula, This is the boundary penalty coefficient; For the first The distance the typhoon turbine crosses the site boundary; this value is 0 when the turbine is located within the compliant area. For serial number subscripts; This represents the total number of single-type fans. The formula for calculating the penalty for violations of the minimum spacing between wind turbines is as follows: = ； In the formula, This is the spacing penalty coefficient; For the first Taiwan and the The value of non-compliance in typhoon turbine spacing is 0 when the turbine spacing meets the minimum safety distance requirement; This is a serial number subscript.

7. The offshore wind farm layout method for a single type of wind turbine according to claim 6, characterized in that, In S7, the combined selection rule for elite retention and random sampling is as follows: The parent and offspring populations are merged to form a mixed pool. Individuals are ranked from best to worst fitness, and the top 50% of the best individuals are selected as elites and directly retained to the next generation. The remaining 50% of the population positions are filled by random sampling from non-elite offspring individuals.

8. The offshore wind farm layout method for a single type of wind turbine according to claim 7, characterized in that, Specifically, S1 is: Configure the core operating parameters of the algorithm, load and standardize the basic data of wind farm engineering, use real number coding to complete the coding of single-type wind turbine layout schemes and construct the initial population, and complete the fitness evaluation of the initial population based on the Gaussian wake model. The encoding formula for real numbers is: ； In the formula, The wind turbine layout encoding vector for a single individual; For the first Normalized plane coordinates corresponding to the typhoon generator; The decoding formula for normalized coordinates and actual engineering coordinates is: ； In the formula, For the first Actual engineering plane coordinates of the typhoon generator; These represent the minimum and maximum values ​​of the X-axis coordinates of the wind farm site; These represent the minimum and maximum values ​​of the Y-axis coordinate of the wind farm site; The formula for calculating the wake velocity deficit in the Gaussian wake model is: ； ； In the formula, This represents the wind speed loss due to the wake. The initial wind speed of the incoming flow; This is the thrust coefficient of the fan at the corresponding inflow velocity. The diameter of the wind turbine rotor; The wake spread radius; The wake spread ratio; This represents the axial distance between the downstream and upstream wind turbines. This is the lateral distance between the center axes of the wakes of the downstream and upstream wind turbines.

9. A computer device, characterized in that, include: A processor configured to be coupled to memory, read and execute instructions and / or program code in the memory to perform the method as described in any one of claims 1-8.

10. A computer-readable medium, characterized in that, The computer-readable medium stores computer program code that, when executed on a computer, causes the computer to perform the method as described in any one of claims 1-8.

Images