Offshore wind farm multi-type wind turbine arrangement optimization method, device and readable medium

Abstract

This invention discloses a method, equipment, and readable medium for optimizing the layout of multiple types of wind turbines in offshore wind farms. It belongs to the technical field of offshore wind farm planning, design, and intelligent optimization algorithms. The method employs four-dimensional chromosome encoding to achieve integrated encoding of turbine location, type, and activation state. It retains gradient information through a soft-constraint smoothing penalty mechanism, combines guided multi-strategy mutation operators incorporating wind power engineering knowledge to correct spatial violations, and supplements this with population diversity restoration and elite-random environment selection strategies to balance the algorithm's global exploration and local development capabilities. This invention, using the aforementioned method, equipment, and readable medium for optimizing the layout of multiple types of wind turbines in offshore wind farms, can minimize the penalized levelized cost of electricity (LCOE) while meeting the target capacity. It is suitable for irregular, non-convex offshore wind farms and those with complex engineering constraints, improving the economic benefits and layout optimization efficiency of offshore wind farms throughout their entire lifecycle.

Description

Technical Field

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

[0002] The global offshore wind power development has entered a phase of rapid growth, and wind farm layout design is a core element determining its economic benefits throughout its entire life cycle. Heterogeneous arrangements of various wind turbine types can adapt to complex offshore wind resource conditions, further improving power generation efficiency and reducing the cost per kilowatt-hour, and have become an important development direction for offshore wind farm layout optimization. Current wind farm layout research mostly focuses on optimizing single-type wind turbines or simply combining multiple types of turbines, lacking efficient and intelligent optimization methods that can adapt to multiple turbine types, making it difficult to meet the engineering design needs of complex offshore wind farms.

[0003] However, existing technologies have the following shortcomings: Traditional algorithms use hard constraint processing mechanisms, directly discarding solutions that violate boundary, spacing, and capacity rules, resulting in a discontinuous search space and loss of optimal layout and gradient information near feasible boundaries; the mutation operator is a general random perturbation that does not incorporate knowledge from the wind farm engineering domain, making it unable to specifically correct spatial violations, resulting in low efficiency in exploring heterogeneous optimization spaces; the coding design only targets the turbine location, without incorporating type and activation variables, and cannot simultaneously optimize turbine location, type selection, and activation state, nor can it automatically determine the optimal number and combination of turbines; at the same time, it lacks a population diversity management mechanism, making the algorithm prone to premature convergence to local optima and difficult to find the globally optimal layout in complex spaces.

[0004] Therefore, there is a need in this field to develop methods, equipment, and readable media for optimizing the layout of various types of wind turbines in offshore wind farms, so as to effectively solve the above problems, adapt to irregular, non-convex offshore wind farm scenarios with complex engineering constraints, improve layout optimization efficiency and global optimization capabilities, reduce the levelized cost of electricity in wind farms, and improve the economic benefits of wind farms throughout their entire life cycle. Summary of the Invention

[0005] The purpose of this invention is to provide a method, equipment, and readable medium for optimizing the layout of various types of wind turbines in offshore wind farms. This method effectively solves the problems in existing technologies, such as discontinuous search space caused by hard constraints, inability of random mutation to specifically correct space violations, inability of single encoding to simultaneously optimize multiple variables, and easy premature convergence and getting trapped in local optima. It can retain gradient information to improve the integrity of layout optimization, improve the exploration and convergence efficiency of heterogeneous optimization space, automatically determine the optimal number and type combination of wind turbines to adapt to heterogeneous layout scenarios of various types of wind turbines, stably output the global optimal layout, and improve the reliability of optimization results.

[0006] To achieve the above objectives, this invention provides a method, equipment, and readable medium for optimizing the layout of various types of wind turbines in offshore wind farms, comprising the following steps: S1. Input wind farm boundary coordinate data, wind resource wind speed-wind direction distribution data, specifications of multiple types of wind turbine generators, project target installed capacity and algorithm preset parameters, and construct an initial population using four-dimensional chromosome coding. The four-dimensional chromosome coding synchronously represents the wind turbine location, model and activation status. Output the initial population to S2. S2. Receive the initial population output by S1, select individuals from the initial population to construct a mating population, balance the selection pressure of the algorithm with population diversity, and output the mating population to S3. S3 receives the mating population output from S2, periodically monitors the genotypic similarity of the population, and triggers diversity repair when the similarity meets a preset threshold to maintain the genetic diversity of the population. The repaired mating population is then output to S4. S4. Receive the repaired mating population output by S3, perform a uniform crossover operation on the parent individuals in the mating population to generate a crossover offspring population, and output the crossover offspring population to S5. S5. Receive the crossover offspring population output by S4, perform multi-strategy mutation that integrates knowledge from the wind power engineering field on the crossover offspring population, and specifically correct the problems of wind turbine spatial layout violations and unreasonable configurations, and output the mutated offspring population to S6. S6. Receive the mutated offspring population output by S5, decode the mutated offspring population, calculate the punitive electricity cost as the individual fitness, the punitive electricity cost integrates boundary constraints, spacing constraints, and capacity constraints through a soft constraint smoothing penalty mechanism, and output the offspring population with fitness values ​​to S7. S7. Receive the parent population and the offspring population with fitness values ​​output by S6, merge the parent population and the offspring population, select elite individuals and randomly supplement offspring individuals to generate a new generation population, and output the new generation population to S8. S8 receives the new generation population output from S7, updates the global optimal solution, visualizes and monitors the algorithm iteration status and constraint satisfaction, and outputs the global optimal solution and iteration monitoring data to S9. S9. Receive the global optimal solution and iterative monitoring data output by S8, and determine whether the preset convergence condition is met. If not, return the new generation population to S2 to re-execute the iteration process. If it is met, output the global optimal solution to S10. S10: Receive the global optimal solution output by S9, decode the global optimal solution, and output the optimal layout scheme of multiple types of wind turbines and all-dimensional performance indicators.

[0007] Preferably, in S1, the four-dimensional chromosome is encoded as a length The encoding structure of a single individual in a one-dimensional row vector is as follows: ; In the formula, For the first Normalized X-coordinate of the typhoon generator; For the first Normalized Y-coordinate of the typhoon generator; For the first Typhoon generator type variable; For the first The active state variable of the typhoon generator; This refers to the maximum number of wind turbines that can be deployed in a wind farm. This is a serial number subscript.

[0008] Preferably, in step S2, individuals are selected from the initial population using a biased tournament operator to construct a mating population, specifically as follows: A fixed-size group of individuals is randomly selected from the population. The individual with the best fitness in the group is selected with an 80% probability, and the non-optimal individual in the group is randomly selected with a 20% probability. This process is repeated until a mating population with the same size as the original population is generated.

[0009] Preferably, in step S3, the preset threshold is 20%, that is, diversity restoration is triggered when the genotypic similarity of the population is below 20%; the diversity restoration specifically includes: Repair methods are randomly selected from a strategy pool containing Gaussian mutation, uniform mutation, type mutation, and activation mutation to perform gene correction on duplicate or highly similar individuals.

[0010] Preferably, in S5, the multi-strategy mutation is divided into spatial mutation and configuration mutation; the spatial mutation includes Gaussian mutation, uniform mutation, spacing repair mutation, discrete mutation, boundary-guided mutation, and coordinate exchange mutation, which are used to correct spatial layout violations such as wind turbine boundary crossing and insufficient spacing; the configuration mutation includes turbine model switching mutation and active state reversal mutation, which are used to optimize wind turbine model combination and installed quantity.

[0011] Preferably, in step S6, the formula for calculating the penalty cost per kilowatt-hour is: ; In the formula, Punitive cost per kilowatt-hour; The basic cost per kilowatt-hour for wind farms; This is a boundary constraint penalty term; This is a penalty term for spacing constraints; This is a capacity constraint penalty term; The boundary constraint penalty item Spacing constraint penalty term The capacity constraint penalty term is calculated using the tanh smoothing function. Based on actual installed capacity With target installed capacity Relationship segmentation settings: when hour ,when hour .

[0012] Preferably, in step S7, merging the parent and offspring populations, selecting elite individuals, and randomly supplementing offspring individuals specifically involves: The parent and offspring populations are merged to form a total candidate set. The selection ratio of elite individuals is preferably 40% to 60%. In this embodiment, the top 50% of individuals in terms of fitness are selected as elite surviving individuals, and the remaining 50% of individuals are randomly supplemented from the offspring that have not been selected as elites, so as to always maintain a constant population size.

[0013] Preferably, in step S9, the preset convergence condition includes at least one of the following: The preset maximum number of iterations is reached; the global optimal fitness remains unchanged for multiple consecutive generations; all engineering constraints are met and the cost per kilowatt-hour meets the target.

[0014] Therefore, the present invention employs the above-mentioned method, equipment, and readable medium for optimizing the layout of various types of wind turbines in offshore wind farms. Compared with the prior art, the technical solution of the present invention has the following beneficial effects: (1) This invention solves the problem of discontinuous search space and discarding high-quality solutions of feasible boundaries caused by hard constraints through a soft constraint smoothing penalty mechanism; it retains gradient information and improves the integrity of layout optimization. (2) This invention solves the problem that random mutation cannot specifically correct spatial violations by using guided multi-strategy mutation operators; it integrates knowledge from the wind power field to significantly improve the efficiency of heterogeneous optimization space exploration and convergence. (3) This invention solves the problem that a single code cannot simultaneously optimize multiple variables by using four-dimensional chromosome coding; it automatically determines the optimal number and model combination of wind turbines and adapts to heterogeneous layout scenarios of multiple types of wind turbines. (4) This invention solves the problems of premature convergence and easy getting trapped in local optima by using population diversity restoration and elite-random selection strategies; it stably outputs the global optimal layout and improves the reliability of optimization results.

[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 an embodiment of the method, equipment, and readable medium for optimizing the layout of multiple types of wind turbines in offshore wind farms according to the present invention. Figure 2 This is a visualization of the wind turbine layout and annual power generation (AEP) distribution of an offshore wind farm, representing an embodiment of the method, equipment, and readable medium for optimizing the layout of various types of wind turbines in an offshore wind farm according to the present invention. Figure 3 This is a simulation comparison of the effects of the improved genetic algorithm in an embodiment of the method, equipment and readable medium for optimizing the layout of multiple types of wind turbines in offshore wind farms according to 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 Figures 1-2 As shown, this embodiment provides a method, equipment, and readable medium for optimizing the layout of various types of wind turbines in offshore wind farms. 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, equipment, and readable medium for optimizing the layout of various types of wind turbines in offshore wind farms, comprising the following steps: S1. This step is used to set the basic parameters of the algorithm and construct the initial population. The initial candidate schemes for wind turbine layout are determined through standardized coding rules, so as to build a stable and diverse initial solution space for subsequent iterative optimization. In this step, the following data is input: boundary coordinates of the offshore wind farm, wind speed and direction distribution data, specifications of various types of wind turbine generators, and the project's target installed capacity. The algorithm also includes preset parameters such as population size, maximum number of iterations, crossover probability, mutation probability, and similarity threshold.

[0020] Based on the target installed capacity and the minimum unit capacity of the candidate wind turbines Calculate the maximum number of wind turbines that can be deployed in a wind farm. The calculation formula is: ; In the formula, This refers to the maximum number of wind turbines that can be deployed in a wind farm. Set a target installed capacity for the project; This represents the minimum capacity of a single wind turbine among the candidate turbines.

[0021] Subsequently, a four-dimensional integrated chromosome coding method was adopted to set each individual as... A 3D row vector is used to synchronously encode the wind turbine location, model, and operational status; the encoding method is represented as follows: ; In the formula, For the first Normalized X-coordinate of the typhoon generator; For the first Normalized Y-coordinate of the typhoon generator; For the first Typhoon generator type variable; For the first The active state variable of the typhoon generator; This refers to the maximum number of wind turbines that can be deployed in a wind farm. For serial number subscripts; Finally, an initial population is generated through a composite strategy. 30% of the individuals are generated using a combination of heuristic layout and expert seed generation to ensure the engineering feasibility of the initial solution, while 70% of the individuals are generated through random sampling to cover different installed capacity schemes and ensure that the initial population has sufficient diversity.

[0022] Obtain the initial population ; and the initial population The data is transmitted to the S2 bias tournament selection operation as input data for this step.

[0023] S2. Select high-quality individuals from the initial population to form a mating population, and reasonably balance the selection pressure of the algorithm with the diversity of the population to prevent the algorithm from getting trapped in a local optimum due to excessive selection of high-quality solutions. In this step, the initial population output is S1. This will serve as the input data for this step; A biased tournament selection operator is used to screen individuals. A fixed-size group of individuals is randomly selected from the population. The individual with the best fitness in the group is selected with a probability of 75% to 85%, and the non-optimal individual in the group is randomly selected with a probability of 15% to 25%. In this embodiment, the optimal individual is selected with a probability of 80% and the non-optimal individual is randomly selected with a probability of 20%. By repeatedly performing the above screening operation, a mating population with the same size as the original population is finally generated.

[0024] Obtain the mating population M; transfer the selected mating population M to the S3 periodic population diversity restoration step as input data for subsequent processing.

[0025] S3 is used to monitor the genotypic similarity of the population in real time, actively correct highly similar individuals, avoid premature convergence of the algorithm caused by the homogenization of the population's genes, and continuously ensure the global search capability of the algorithm. In this step, the mating population M output by S2 is used as the input data for this step; the genotypic similarity of individuals within the population is calculated at a fixed period. The population genotypic similarity threshold is preferably set to 15%~25%. In this embodiment, the threshold is set to 20%. When the similarity is lower than this threshold, the population diversity restoration program is automatically started.

[0026] The repair method is randomly selected from a strategy pool containing Gaussian mutation, uniform mutation, type mutation, and enabled mutation. Genetic correction is performed on duplicate or highly similar individuals with a 70% probability of being guided by random individuals within the population and a 30% probability of being guided by the global optimal solution.

[0027] Obtain the restored mating population M'; transfer the mating population M' that has completed diversity restoration to the S4 uniform crossover operation as input data for this step.

[0028] S4. By performing unbiased gene recombination on parent individuals, the optimization solution space is expanded, offspring individuals that inherit the advantageous characteristics of the parents are generated, and the gene combination forms of the population are enriched. In this step, the repaired mating population M' output by S3 is used as the input data for this step; the parent individuals in the mating population are randomly paired in pairs, and the gene recombination operation is performed using a uniform crossover operator. Each gene locus (coordinate, model, enabled variable) has an equal probability of being inherited from the father and mother, so as to achieve unbiased recombination of genetic information without changing the chromosome coding structure.

[0029] Output crossover offspring population The offspring population generated by the crossover operation The data is transferred to the S5-guided multi-policy mutation operation as input data for this step.

[0030] S5 integrates knowledge from the field of wind power engineering to carry out targeted corrections for issues such as wind turbine boundary crossing, insufficient spacing, and unreasonable turbine configuration, abandoning the blindness of traditional random variation and significantly improving the exploration efficiency of heterogeneous space; In this step, the crossover offspring population is output as S4. The input data for this step is used as input data; a guided multi-strategy mutation operator is used to perform the mutation operation, and the mutation strategies are divided into two categories: spatial and configurational. Spatial mutations include Gaussian mutation, uniform mutation, spacing repair mutation, discretization mutation, boundary-guided mutation, and coordinate exchange mutation, used to correct violations in wind turbine spatial layout. Configuration mutations include turbine model switching mutation and activation state reversal mutation, used to optimize wind turbine model combinations and installed capacity. A mutation strategy is randomly selected for each offspring and executed to complete targeted layout and configuration corrections.

[0031] Obtain a mutant offspring population The offspring population that has undergone mutation treatment The data is transmitted to S6 and used as input for the fitness evaluation step.

[0032] S6 is used to calculate the individual fitness value, integrate the soft constraint smoothing penalty mechanism, construct a continuous and differentiable fitness function, and retain gradient information to guide the algorithm to accurately find the optimal solution in the feasible region. In this step, the mutant offspring population output by S5 is... As input data for this step; decode the individual units, round the model variable to determine the wind turbine model, convert the activation variable to binary activation status with a threshold of 0.5, filter and activate the wind turbines, and calculate the actual installed capacity. .

[0033] Then, the core indicators are calculated in sequence: The formula for calculating annual power generation is: ; In the formula, Annual power generation; The total number of segments representing the incoming wind direction; The total number of wind speed segments within a single wind direction; For the first The wind direction, the first Wind frequency probability for each wind speed range To consider the superposition effect of the flow, the first Effective wind speed at the fan location within the wind speed range; This represents the total annual power generation duration; This represents the total number of activated wind turbines within the wind farm. For the first The power output of the typhoon generator; This is the subscript number.

[0034] The formula for calculating the basic cost per kilowatt-hour is: ; ; In the formula, The basic cost per kilowatt-hour; For the initial capital expenditure of the wind farm; Fixed fee rate; Annual operating and maintenance expenses; This refers to the annual power generation of the wind farm. The discount rate; For the number of years of operation; Annual equal amount of funds recovered ; represents the present value of the initial investment.

[0035] The soft constraint penalty calculation includes: The capacity constraint penalty is defined as: ; In the formula, This is a capacity constraint penalty term; This refers to the actual installed capacity. Set a target installed capacity for the project.

[0036] when At that time, capacity constraint penalty term The preferred value is 1% to 5%, and in this embodiment it is set to no less than 1%.

[0037] Boundary constraint penalty Spacing constraint penalty Defined as: ; ; In the formula, This is a boundary constraint penalty term; This is a penalty term for spacing constraints; This is the upper limit of the penalty coefficient; This is the lower limit of the penalty coefficient. For the proportion of boundary crossing, These represent the percentages of spacing violations; Average violation rate based on spacing; The maximum violation is the spacing.

[0038] Punitive cost per kilowatt-hour The calculation formula is: ; Finally, a genotype deduplication combined with parallel looping was used to complete the fitness assessment of all individuals and eliminate redundant calculations.

[0039] Obtain offspring population with fitness values The offspring population that has completed fitness assessment Transmitted to the S7 Elite-Random Environment Selection step as input data for that step.

[0040] S7. By merging the parent and offspring populations, high-quality individuals are selected to form a new generation of population. While monotonically preserving the historical best solution, new genetic information is continuously injected into the population. In this step, the current parent population is used. Offspring population output from step S6 Both are used as input data for this step; the parent and offspring populations are merged to form a total candidate set, and the top 50% of individuals in terms of fitness are selected as elite surviving individuals. The remaining 50% of individuals are randomly added from the offspring that were not selected as elites, so as to always maintain a constant population size.

[0041] Obtain a new generation of population The new generation of population obtained through screening The data is transmitted to the S8 global optimal solution update and iteration monitoring step as input data for that step.

[0042] S8 is used to track the algorithm iteration process in real time, update the global optimal solution, and intuitively monitor the convergence status and constraint satisfaction of the algorithm through visualization. In this step, the new generation population output by step S7... This serves as the input data for this step; comparing the current best individual in the population with the historical global best solution. It retains better solutions and updates the global optimal solution; at the same time, it regularly generates iterative convergence curves and wind turbine layout visualization charts to monitor the algorithm iteration progress and the satisfaction of various engineering constraints in real time.

[0043] Obtain the global optimal solution Iterative monitoring data; updating the global optimal solution The iterative monitoring data is transmitted to the S9 convergence condition judgment step as input data for that step.

[0044] S9 is used to control the optimization iteration process. It determines whether to terminate the optimization based on preset conditions, thereby achieving automatic closed-loop control of the entire algorithm process. In this step, the global optimal solution is output by step S8. The iterative monitoring data is used as the input data for this step; it is determined whether the preset convergence conditions are met, including three cases: reaching the maximum number of iterations, no significant change in the global optimal fitness over multiple consecutive generations, and the constraints being met and the cost per kilowatt-hour being achieved. The subsequent process direction is determined based on the judgment results.

[0045] Obtain the convergence judgment result; if the convergence condition is not met, the new generation population will be... Return to the S2 biased tournament selection operation and re-execute the iteration process; if the convergence condition is met, obtain the global optimal solution. Transmit to S10 optimal solution decoding and output steps.

[0046] S10. Decode and analyze the global optimal solution, output the final wind turbine layout scheme and all-dimensional performance indicators, and complete the optimization process of the layout of multiple types of wind turbines in the entire offshore wind farm. In this step, the globally optimal solution is output by S9. The input data for this step is used to fully decode the optimal chromosome, accurately obtaining the wind turbine deployment coordinates, model configuration, operational status, and actual installed capacity; the output includes the optimal layout scheme and the penalty cost per kilowatt-hour (LCOE). Annual power generation () The system generates a comprehensive performance report covering wake loss, capacity factor, and constraint satisfaction, and produces a visual drawing of the wind turbine layout, thus completing the entire optimization process.

[0047] Output the optimal layout technical solution and performance index report for various types of wind turbines in offshore wind farms.

[0048] like Figure 3 As shown, this embodiment verifies the convergence of the improved genetic algorithm (IGA) and the traditional genetic algorithm (GA) under different operating conditions through simulation comparison. The simulation results fully demonstrate that under the same test conditions (Case2 / Case3 / Case4), the penalty levelized electricity cost (pLCoE, unit: RMB / kWh) curve corresponding to the IGA of this invention is significantly lower than the curve corresponding to the traditional GA throughout the entire process. Its curve declines more steeply and enters the convergence and stabilization stage earlier. This proves that the improved mechanisms such as soft constraint smoothing penalty and guided multi-strategy mutation adopted in this invention can significantly improve the algorithm search efficiency, quickly locate the global optimal solution, greatly reduce invalid iterations, and the final converged pLCoE is much lower than that of the traditional GA, effectively improving the layout optimization accuracy and obtaining a better levelized electricity cost. In this embodiment, the IGA optimization scheme for heterogeneous arrangement of multiple types of wind turbines achieves a final convergence pLCoE of only 0.351~0.352 RMB / kWh, which is significantly better than the convergence value of over 0.352 RMB / kWh for single-type wind turbine operating conditions (Case2 / Case3 / Case4) under the same IGA framework. This fully verifies that this embodiment can further reduce the levelized cost of electricity (LCOE) of wind farms throughout their entire life cycle and achieve optimal economic benefits. At the same time, all IGA curves show no significant fluctuations and the convergence process is smooth and continuous, indicating that the population diversity restoration and elite-random environment selection mechanism in this embodiment effectively avoids the problem of premature convergence of the algorithm, ensuring the stability of the optimization results and their engineering feasibility.

[0049] Example 2 The difference between this modified embodiment and Embodiment 1 is that the biased tournament selection is replaced with roulette selection, while the remaining steps and parameter settings are exactly the same as in Embodiment 1.

[0050] The roulette wheel selection method specifically calculates the selection probability based on the individual's fitness value. Individuals with higher fitness have a higher probability of being selected to enter the mating population. By generating a mating population with the same size as the original population through random sampling, the algorithm's selection pressure can be further reduced while ensuring population diversity. This method is suitable for offshore wind farm layout optimization scenarios where wind resources are more evenly distributed.

[0051] Example 3 The difference between this modified embodiment and Embodiment 1 is that the tanh smoothing penalty function for boundary constraints and spacing constraints is replaced with the sigmoid smoothing penalty function, while the remaining steps and parameter settings are exactly the same as in Embodiment 1.

[0052] The sigmoid smoothing penalty function provides a gentler change in the penalty gradient. When the wind turbine layout is close to the constraint boundary, it can more gently guide the algorithm to converge to the feasible region. It is suitable for offshore wind farm layout optimization with extremely irregular boundary shapes and complex constraints, and improves the stability of the optimization process.

[0053] Therefore, this invention adopts the above-mentioned method, equipment, and readable medium for optimizing the layout of multiple types of wind turbines in offshore wind farms. This method effectively solves the problems in the prior art, such as discontinuous search space caused by hard constraints, inability of random mutation to specifically correct space violations, inability of single encoding to simultaneously optimize multiple variables, and easy premature convergence and getting trapped in local optima. It can retain gradient information to improve the integrity of layout optimization, improve the exploration and convergence efficiency of heterogeneous optimization space, automatically determine the optimal number and type combination of wind turbines to adapt to heterogeneous layout scenarios of multiple types of wind turbines, stably output the global optimal layout, and improve the reliability of optimization results.

[0054] 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.

[0055] 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 optimizing the layout of various types of wind turbines in offshore wind farms, characterized in that, Includes the following steps: S1. Input wind farm boundary coordinate data, wind resource wind speed-wind direction distribution data, specifications of multiple types of wind turbine generators, project target installed capacity and algorithm preset parameters, and construct an initial population using four-dimensional chromosome coding. The four-dimensional chromosome coding synchronously represents the wind turbine location, model and activation status. Output the initial population to S2. S2. Receive the initial population output by S1, select individuals from the initial population to construct a mating population, balance the selection pressure of the algorithm with population diversity, and output the mating population to S3. S3 receives the mating population output from S2, periodically monitors the genotypic similarity of the population, and triggers diversity repair when the similarity meets a preset threshold to maintain the genetic diversity of the population. The repaired mating population is then output to S4. S4. Receive the repaired mating population output by S3, perform a uniform crossover operation on the parent individuals in the mating population to generate a crossover offspring population, and output the crossover offspring population to S5. S5. Receive the crossover offspring population output by S4, perform multi-strategy mutation that integrates knowledge from the wind power engineering field on the crossover offspring population, and specifically correct the problems of wind turbine spatial layout violations and unreasonable configurations, and output the mutated offspring population to S6. S6. Receive the mutated offspring population output by S5, decode the mutated offspring population, calculate the punitive electricity cost as the individual fitness, the punitive electricity cost integrates boundary constraints, spacing constraints, and capacity constraints through a soft constraint smoothing penalty mechanism, and output the offspring population with fitness values ​​to S7. S7. Receive the parent population and the offspring population with fitness values ​​output by S6, merge the parent population and the offspring population, select elite individuals and randomly supplement offspring individuals to generate a new generation population, and output the new generation population to S8. S8 receives the new generation population output from S7, updates the global optimal solution, visualizes and monitors the algorithm iteration status and constraint satisfaction, and outputs the global optimal solution and iteration monitoring data to S9. S9. Receive the global optimal solution and iterative monitoring data output by S8, and determine whether the preset convergence condition is met. If not, return the new generation population to S2 to re-execute the iteration process. If it is met, output the global optimal solution to S10. S10: Receive the global optimal solution output by S9, decode the global optimal solution, and output the optimal layout scheme of multiple types of wind turbines and all-dimensional performance indicators.

2. The method for optimizing the layout of multiple types of wind turbines in offshore wind farms according to claim 1, characterized in that, In S1, the four-dimensional chromosome is encoded as length. The encoding structure of a single individual in a one-dimensional row vector is as follows: ； In the formula, For the first Normalized X-coordinate of the typhoon generator; For the first Normalized Y-coordinate of the typhoon generator; For the first Typhoon generator type variable; For the first The active state variable of the typhoon generator; This refers to the maximum number of wind turbines that can be deployed in a wind farm. This is a serial number subscript.

3. The method for optimizing the layout of multiple types of wind turbines in offshore wind farms according to claim 2, characterized in that, In S2, a biased tournament operator is used to select individuals from the initial population to construct a mating population, specifically as follows: A fixed-size group of individuals is randomly selected from the population. The individual with the best fitness in the group is selected with an 80% probability, and the non-optimal individual in the group is randomly selected with a 20% probability. This process is repeated until a mating population with the same size as the original population is generated.

4. The method for optimizing the layout of multiple types of wind turbines in offshore wind farms according to claim 3, characterized in that, In S3, the preset threshold is 20%, meaning that diversity restoration is triggered when the genotypic similarity of the population is below 20%; the diversity restoration specifically involves: Repair methods are randomly selected from a strategy pool containing Gaussian mutation, uniform mutation, type mutation, and activation mutation to perform gene correction on duplicate or highly similar individuals.

5. The method for optimizing the layout of multiple types of wind turbines in offshore wind farms according to claim 4, characterized in that, In S5, the multi-strategy mutation is divided into spatial mutation and configuration mutation. The spatial mutation includes Gaussian mutation, uniform mutation, spacing repair mutation, discrete mutation, boundary-guided mutation, and coordinate exchange mutation, which are used to correct spatial layout violations such as wind turbine boundary overruns and insufficient spacing. The configuration mutation includes turbine model switching mutation and active state reversal mutation, which are used to optimize wind turbine model combinations and installed capacity.

6. The method for optimizing the layout of multiple types of wind turbines in an offshore wind farm according to claim 5, characterized in that, In S6, the formula for calculating the penalty cost per kilowatt-hour is: ； In the formula, Punitive cost per kilowatt-hour; The basic cost per kilowatt-hour for wind farms; This is a boundary constraint penalty term; This is a penalty term for spacing constraints; This is a capacity constraint penalty term; The boundary constraint penalty item Spacing constraint penalty term The capacity constraint penalty term is calculated using the tanh smoothing function. Based on actual installed capacity With target installed capacity Relationship segmentation settings: when hour ,when hour .

7. The method for optimizing the layout of multiple types of wind turbines in offshore wind farms according to claim 6, characterized in that, In S7, the parent and offspring populations are merged, elite individuals are selected, and offspring individuals are randomly added, specifically as follows: The parent and offspring populations are merged to form a total candidate set. The top 50% of individuals in terms of fitness in the total candidate set are selected as elite individuals, and the remaining 50% are randomly added from the offspring individuals that were not selected as elites, so as to maintain a constant population size.

8. The method for optimizing the layout of multiple types of wind turbines in an offshore wind farm according to claim 7, characterized in that, In step S9, the preset convergence condition includes at least one of the following: Reach the preset maximum number of iterations; The global optimal fitness remained unchanged for several consecutive generations; all engineering constraints were satisfied and the cost per kilowatt-hour met the target.

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