A method and apparatus for optimizing wind farm layout considering load constraints
By combining load simulation and load proxy models with engineering wind farm models and improved genetic algorithms, the problem of load influence in wind farm layout optimization was solved, achieving more efficient and realistic wind farm layout optimization, reducing wind turbine loads and improving optimization efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHEJIANG UNIV
- Filing Date
- 2023-12-13
- Publication Date
- 2026-05-26
Smart Images
Figure CN117725824B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of numerical simulation, and specifically relates to a method and apparatus for optimizing wind farm layout considering load constraints. Background Technology
[0002] Wind energy, as a clean and renewable energy source, is developing rapidly. However, previous studies have shown that downwind wind turbines are affected by the wake of upwind wind turbines during operation, leading to reduced output power and wasting resources and resources. Therefore, the early planning of wind farms is crucial. In recent years, wind farm layout optimization methods based on optimization algorithms have gradually become common tools for early planning of wind farms due to their good optimization effect and consideration of wake losses. For example, application number CN201810060152.9 uses minimizing the life-cycle cost of the wind farm and maximizing the annual power generation as optimization objectives, and uses the minimum installation spacing of wind turbines and site limitations as constraints to perform multi-objective optimization of wind farm layout, while finding a set of Pareto optimal solutions with advantages and disadvantages. Application number CN202110322312.4 uses a genetic algorithm to obtain the optimal wind farm layout scheme by combining the sum of power generation model and the sum of cost model as the objective function. Application No. CN202110889731.6 considers the wind farm scheduling strategy in wind turbine layout optimization, and applies greedy algorithms and particle swarm optimization algorithms to optimize the wind turbine layout. Application No. CN202111402387.X uses the Gaussian model to describe the wake effect in the wind farm and uses a local search algorithm to solve the layout optimization problem. Application No. CN202111402365.3 discloses a wind farm layout optimization method based on mathematical programming, innovatively applying mathematical programming algorithms to optimize the wind farm layout. Application No. CN202210975919.7 combines gridded genetic algorithms and coordinate-based genetic algorithms, solving the problem of optimizing the number of wind turbine units and their high computational cost in wind farm layout optimization, while avoiding the limitation of gridded genetic algorithms on the flexibility of wind turbine unit location. Application No.: CN202310138620.0 This paper adopts the improved NSGA-Ⅲ wind turbine layout optimization method to optimize the layout of wind turbines for multiple objectives such as annual power generation, total investment cost and noise of wind farms.
[0003] However, existing patents have not considered the load impact during wind farm layout optimization. Although some wind turbines may generate more power after layout optimization, they may also be affected by higher wind speeds and turbulence, leading to a reduction in their lifespan. Therefore, considering the load on wind turbines within the wind farm during the optimization process is extremely important. Application No.: CN202310926769.5 proposes a fatigue life assessment method for wind turbines using external meteorological parameters and the actual operating status of in-service wind turbines. Application No.: CN202210767889.0 proposes collecting SCADA data and using a DNN for training to achieve real-time calculation of damage equivalent loads. Application No.: CN202210291363.X establishes an explicit ultimate load prediction model to predict the ultimate load at wind turbine locations. Application No.: CN201510873111.8 proposes a long-term load assessment method for wind farms, training an artificial neural network model based on measurement data, suitable for long-term load monitoring of operating wind farms. However, most of the above load assessment methods require measured data from wind turbines and wind farms or require a large amount of simulation calculations, which makes them impractical and have not yet been applied to the layout optimization of wind farms.
[0004] Furthermore, optimizing wind farm layouts considering load constraints using optimization algorithms requires numerous iterations, which significantly increases the computational burden on wind farm loads. Therefore, there is a need to further develop and improve efficient layout optimization methods and devices that consider load constraints. Summary of the Invention
[0005] The purpose of this invention is to solve the problems existing in the current technology and to provide a method and apparatus for optimizing wind farm layout considering load constraints.
[0006] The objective of this invention is achieved through the following technical solution: Firstly, this invention provides a wind farm layout optimization method considering load constraints, the steps of which are as follows:
[0007] (1) Based on the range of hub wind speed, turbulence intensity and yaw angle of the wind turbine, random sampling is performed within the range using low difference sequences to generate multiple sets of inflow condition data;
[0008] (2) Based on the inflow condition data and relevant parameters required for simulation in step (1), the wind turbine is subjected to load simulation and the time-series load of the wind turbine is calculated. The time-series load data is processed into equivalent loads within a certain time period by rainflow counting method to form a load proxy model training dataset.
[0009] (3) The load proxy model training dataset obtained in (2) is used to train the artificial neural network after hyperparameter tuning to obtain the load proxy model for each load channel.
[0010] (4) Use the engineering wind field model to calculate the velocity field and turbulence field of the wind farm; use the load proxy model obtained in (3) to quickly evaluate the equivalent load of the wind turbines in the field within a certain time; calculate the life cycle equivalent load according to the wind frequency distribution; and encapsulate the evaluation process into a wind farm load fast evaluation function.
[0011] (5) Use an improved genetic algorithm to optimize the layout of the wind farm; during the iteration of the genetic algorithm, use the wind farm load fast evaluation function obtained in step (4) to calculate the load corresponding to the layout; add the wind turbine spacing constraint and load constraint in the wind farm as a penalty function to the genetic algorithm so that all wind turbines in the wind farm can keep the load below the limit value while maintaining the spacing.
[0012] Further, in step (1), the low-difference sequence is a Sobol sequence; the multiple sets of inflow condition data are 1024 sets of inflow condition data.
[0013] Further, in step (2), the inflow condition data includes hub wind speed, turbulence intensity, and wind turbine yaw angle; the relevant parameters required for simulation include wind turbine aerodynamic data, elastic dynamics data, hydrodynamic data, servo system data, and structural mechanics data; the load simulation of the wind turbine is implemented based on writing an OpenFAST case setting file; the load proxy model training dataset includes hub wind speed, turbulence intensity, wind turbine yaw angle, and equivalent load data of each load channel within a certain time period.
[0014] Furthermore, in step (3), the hyperparameter tuning method is specifically the Keras-Tuner library; the artificial neural network is specifically a fully connected neural network, with three variables as input to the neural network of each load channel, namely hub wind speed, turbulence intensity, and yaw angle, and a single variable as output, namely the equivalent load of the load channel.
[0015] Furthermore, in step (4), the engineering wind field model includes the Bastankhah Gaussian wake velocity deficit model, the Steen Frandsen turbulence model, and the sum of squares superposition model; the method for calculating the lifetime equivalent load is as follows:
[0016]
[0017] Where f is the sampling frequency of DEL, T Lifetime N represents the total number of seconds in the lifecycle of the wind turbine. eqThe material cycle number is used to calculate the equivalent load. P(wd,ws) represents the frequency of wind direction wd and wind speed ws, DEL(wd,ws) represents the equivalent load of the wind turbine under wind conditions of wind direction wd and wind speed ws, and m represents the material... The exponent; the physical meaning of LDEL is: in order to achieve N eq The equivalent load required to achieve the same total fatigue damage as the entire lifespan in a single cycle.
[0018] Further, in step (5), the optimization object of the improved genetic algorithm is the coordinates (X,Y) of all wind turbines in the wind farm; the mutation operator of the improved genetic algorithm is to randomly move the wind turbine to a new position; the crossover operator of the improved genetic algorithm is to exchange the positions of several wind turbines within two individuals; the spacing constraint penalty function and the load constraint penalty function are respectively:
[0019]
[0020]
[0021] Where, N wt For the number of wind turbines, d ij Let be the distance between the i-th and j-th wind turbines, and be the minimum allowable spacing. Load i Let be the load of the i-th wind turbine, denoted as DEL for single-wind-condition operation and LDEL for all-wind-condition operation. limit This represents the maximum allowable load.
[0022] Secondly, the present invention also provides a wind farm layout optimization device considering load constraints, including a memory and one or more processors, wherein the memory stores executable code, and when the processor executes the executable code, it implements the wind farm layout optimization method considering load constraints.
[0023] Thirdly, the present invention also provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the aforementioned wind farm layout optimization method considering load constraints.
[0024] The beneficial effects of this invention are as follows: The wind farm layout optimization method and apparatus considering load constraints established in this invention are an innovation and improvement on traditional methods, possessing advantages such as considering load constraints, being closer to engineering reality, finding the global optimal solution, and high optimization efficiency. The wind farm layout optimization method and apparatus considering load constraints can perform parallel optimization for wind farms or wind farm clusters, and the applicable wind speed and turbulence range covers the wind conditions that wind farms may experience. The optimization effect is even better when the background wind speed and turbulence intensity are high. Attached Figure Description
[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0026] Figure 1 A schematic diagram of the wind farm layout optimization method considering load constraints provided by the present invention;
[0027] Figure 2 This is a schematic diagram illustrating the changing trends of wind farm power and maximum DEL during the layout optimization process of an example of an offshore wind farm in Hangzhou Bay.
[0028] Figure 3 A comparison diagram of the DEL distribution of 63 wind turbines in an offshore wind farm in Hangzhou Bay after the layout optimization of an example.
[0029] Figure 4 A graph showing the changing trend of the maximum values of AEP and LDEL of the Horns Rev1 offshore wind farm during the layout optimization process of the Horns Rev1 offshore wind farm example;
[0030] Figure 5 Comparison of LDEL distribution of 80 wind turbines in the Horns Rev1 offshore wind farm after layout optimization;
[0031] Figure 6 This is a schematic diagram of the wind farm layout optimization device considering load constraints provided by the present invention. Detailed Implementation
[0032] The present invention will be further described and illustrated below with reference to the accompanying drawings and specific embodiments.
[0033] This invention mainly constructs a wind farm layout optimization method considering load constraints. The optimization algorithm couples a fast load evaluation method to perform layout optimization of the wind farm considering load constraints. The specific steps are as follows:
[0034] 1) Based on the range of hub wind speed, turbulence intensity, and yaw angle during wind turbine operation, random sampling was performed using the Sobol low-difference sequence within this range to generate a large number of inflow condition samples. Since the Sobol sequence performs well in generating a power of 2 number of data points, considering both computation time and model accuracy requirements, the number of generated samples was set to 1024.
[0035] 2) OpenFAST is an open-source wind turbine load calculation package composed of multiple sub-modules, such as inflow wind module, aerodynamics module, elastic dynamics module, structural mechanics module, hydrodynamics module, and servo module. Researchers only need to write configuration files for each module to calculate time-series loads, flow field contour maps, and other results. Using the inflow condition data and relevant parameters required for simulation from step 1), an OpenFAST case configuration file is written to simulate the wind turbine load and calculate the time-series load. To use a unified standard to evaluate the load on the wind turbine, the time-series load data is processed into 10-minute Damage Equivalent Loads (DEL) using the rainflow counting method, forming the load surrogate model training dataset. The calculation formula for DEL is as follows:
[0036]
[0037] Where, n i S is the number of cycles for the i-th load cycle obtained by the rainflow counting method. i Let m be the load amplitude of the i-th load cycle obtained by the rainflow counting method, and m be the material... The exponent, f, is the sampling frequency of DEL, which is 1Hz in this case, and T. ref For reference, the time is 600 seconds.
[0038] The final training dataset for the load proxy model includes hub wind speed, turbulence intensity, wind turbine yaw angle, and 10-minute equivalent loads for each load channel.
[0039] 3) Divide the training dataset of the load surrogate model obtained in step 2) into training set, validation set and test set in a ratio of 6:2:2. Build a fully connected neural network, using the Adam optimizer, the loss function is the mean absolute percentage error, the activation function is ReLU, the input layer has three input parameters: hub wind speed, turbulence intensity and wind turbine yaw angle, and the output layer is the DEL output of the target load channel. Perform hyperparameter tuning, and use the tuned artificial neural network for training to obtain the load surrogate model for each load channel.
[0040] 4) To rapidly assess the load on the entire wind farm, an engineering wind farm model was used to calculate the velocity and turbulence fields across the entire farm. This model included the Gaussian wake deficit model, the Steen Frandsen turbulence engineering model, and the sum-of-squares superposition model. Using the calculated wind speeds and turbulence intensities, combined with the load proxy model obtained in 3), a rapid assessment of the 10-minute equivalent load on all wind turbines was performed. The lifetime equivalent load was then calculated based on the wind frequency distribution, using the following formula:
[0041]
[0042] Where f is the sampling frequency of DEL, i.e., 1Hz, and T Lifetime N represents the total number of seconds in the lifecycle of the wind turbine. eq The material cycle number is used to calculate the equivalent load. P(wd,ws) represents the frequency of wind direction wd and wind speed ws, DEL(wd,ws) represents the DEL under wind conditions with wind direction wd and wind speed ws, and m represents the material's... The exponent. The physical meaning of LDEL is: in order to achieve N eq The equivalent load required to achieve the same total fatigue damage as the entire lifespan in a single cycle.
[0043] This evaluation process is encapsulated into a fast wind farm load evaluation function, enabling rapid full-field load simulation for a specific wind farm layout.
[0044] 5) To apply the genetic algorithm to the wind farm layout optimization problem, the genetic algorithm was improved by redesigning the matrix-encoded encoding method and the corresponding mutation and crossover operations. The encoding method is as follows: the entire population consists of multiple solution individuals, each solution individual being a 2-row N-series genetic algorithm. wt A matrix of columns, where N wt Let represent the number of wind turbines, and each column of an individual represents the (X,Y) coordinates of each turbine. The mutation operation randomly selects turbines from the individuals based on the mutation probability and moves them randomly to new positions within the layout boundary. The crossover operation randomly selects two individuals based on the crossover probability and swaps several turbine coordinates between them. Because the turbine coordinates (X,Y) are used for encoding instead of treating X and Y as independent optimization variables, new locations can be easily placed within the wind farm boundary during the initialization and mutation phases. Using this optimization method, layout optimization targeting wind farm power generation can be achieved. However, to consider load constraints, the wind farm load fast evaluation function obtained in step 4) needs to be coupled in the form of a penalty function during the iteration of the genetic algorithm. This ensures that while optimizing wind farm power generation, the load on the wind turbines is controlled below the limit value, ultimately achieving wind farm layout optimization considering load constraints.
[0045] The specific implementation effect of the above method is demonstrated below with reference to the embodiments.
[0046] Example
[0047] In this embodiment, a wind farm in Hangzhou Bay under single wind conditions and the Horns Rev1 offshore wind farm in Denmark under all wind conditions are used as research objects to explore the specific implementation effect and advantages of this wind farm layout optimization algorithm considering load constraints. The relevant parameter settings are as follows:
[0048] The wind farm uses real-world wind condition data, and the wind turbines are NREL-5MW wind turbines with publicly available design parameters. The genetic algorithm population size is set to 64, the number of iterations is set to 5000 generations, and an adaptive mutation rate is used. The specific settings are shown in Table 1, and the crossover rate is set to 0.95.
[0049] Table 1. Mutation rate settings for the genetic algorithm
[0050] Number of iteration rounds 0-1500 1500-2000 2000-2500 2500-3000 3000-3500 3500-4000 4000-5000 mutation rate 0.01 0.005 0.002 0.001 0.0005 0.0002 0.0001
[0051] By comparing ordinary optimization without considering load constraints and optimization with considering load constraints, Figure 2 This paper illustrates the trends in wind farm power and maximum damper efficiency (DEL) during the layout optimization process of an offshore wind farm in Hangzhou Bay. In the initial optimization phase, the power increase was slower with the load-considered layout optimization compared to ordinary optimization. However, after approximately 2800 iterations, the power output of both optimization methods stabilized and reached similar levels, and the maximum DEL was effectively limited, decreasing by 5.64%. For a detailed view, see the DEL distribution diagram of the 63 wind turbines in the optimized wind farm, as shown below. Figure 3 As shown, the overall DEL of the ordinary optimization is larger than that of the layout optimization considering the load, and two wind turbines are subjected to a large DEL, which will lead to a greater risk of failure during operation.
[0052] Figure 4 This paper illustrates the trends in AEP and maximum LDEL of the Horns Rev1 offshore wind farm during layout optimization. In the initial optimization phase, the load-considered layout optimization showed a slower increase in AEP compared to the standard optimization. However, after approximately 3200 iterations, the load-considered layout optimization method and the standard optimization method after approximately 3800 iterations achieved stable and very similar AEP values. The former was only 0.103% lower than the latter, and the maximum LDEL was effectively limited, decreasing by 1.70%. A detailed view of the LDEL distribution of the 80 wind turbines within the optimized layout is provided. Figure 5 As shown, compared with the optimized layout considering the load, the overall LDEL of the ordinary optimization is larger, and many wind turbines are subjected to LDELs much higher than the average level, which will lead to a greater risk of failure during operation.
[0053] In addition, compared with the traditional optimization method, the load optimization method in this embodiment consumes less computational resources to calculate the load, which can be ignored. Considering that the load is also an important factor to consider in the optimization of wind farm layout, this improved method has more practical engineering significance than the original method.
[0054] Corresponding to the aforementioned embodiment of a wind farm layout optimization method considering load constraints, the present invention also provides an embodiment of a wind farm layout optimization device considering load constraints.
[0055] See Figure 6 The present invention provides a wind farm layout optimization device considering load constraints, comprising a memory and one or more processors. The memory stores executable code, and when the processor executes the executable code, it is used to implement a wind farm layout optimization method considering load constraints as described in the above embodiment.
[0056] An embodiment of a wind farm layout optimization device considering load constraints provided by this invention can be applied to any device with data processing capabilities, such as a computer. The device embodiment can be implemented through software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor of any data processing device loading the corresponding computer program instructions from non-volatile memory into memory for execution. From a hardware perspective, such as... Figure 6 The diagram shown is a hardware structure diagram of any data processing-capable device used in a wind farm layout optimization device considering load constraints, provided by the present invention. (Except for...) Figure 6 In addition to the processor, memory, network interface, and non-volatile memory shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.
[0057] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.
[0058] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the present invention according to actual needs. Those skilled in the art can understand and implement this without creative effort.
[0059] This invention also provides a computer-readable storage medium storing a program thereon, which, when executed by a processor, implements a wind farm layout optimization method considering load constraints as described in the above embodiments.
[0060] The computer-readable storage medium can be an internal storage unit of any data processing device as described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device of any data processing device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units and external storage devices of any data processing device. The computer-readable storage medium is used to store the computer program and other programs and data required by the data processing device, and can also be used to temporarily store data that has been output or will be output.
[0061] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the invention. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the invention. Therefore, all technical solutions obtained through equivalent substitution or transformation fall within the protection scope of the present invention.
Claims
1. A wind farm layout optimization method considering load constraints, characterized in that, The steps of this method are as follows: (1) Based on the range of hub wind speed, turbulence intensity and yaw angle of the wind turbine, random sampling is performed within the range using low difference sequences to generate multiple sets of inflow condition data; (2) Based on the inflow condition data and relevant parameters required for simulation in step (1), the wind turbine is subjected to load simulation and the time-series load of the wind turbine is calculated; the time-series load data is processed into equivalent loads within a certain time period by rainflow counting method, and a load proxy model training dataset is formed. (3) The load proxy model training dataset obtained in (2) is used to train the artificial neural network after hyperparameter tuning to obtain the load proxy model for each load channel. (4) Use the engineering wind field model to calculate the velocity field and turbulence field of the wind farm; use the load proxy model obtained in (3) to quickly evaluate the equivalent load of the wind turbines in the field within a certain time; calculate the life cycle equivalent load according to the wind frequency distribution; and encapsulate the evaluation process into a wind farm load fast evaluation function. (5) Use an improved genetic algorithm to optimize the wind farm layout; during the iteration process of the genetic algorithm, use the wind farm load fast evaluation function obtained in step (4) to calculate the load corresponding to the layout; add wind turbine spacing constraints and load constraints in the wind farm as penalty functions to the genetic algorithm so that all wind turbines in the wind farm maintain the spacing while controlling the load below the limit value; the optimization object of the improved genetic algorithm is the coordinates (X, Y) of all wind turbines in the wind farm, the mutation operator of the improved genetic algorithm is to randomly move the wind turbine to a new position, and the crossover operator of the improved genetic algorithm is to exchange the positions of several random wind turbines in two individuals; the spacing constraint penalty function and the load constraint penalty function are respectively: Where, N wt For the number of wind turbines, d ij Let be the distance between the i-th and j-th wind turbines, and be the minimum allowable spacing. Load i Let be the load of the i-th wind turbine, denoted as DEL for single-wind-condition operation and LDEL for all-wind-condition operation. limit This represents the maximum allowable load.
2. The wind farm layout optimization method considering load constraints according to claim 1, characterized in that, In step (1), the low-difference sequence is a Sobol sequence; the multiple sets of inflow condition data are 1024 sets of inflow condition data.
3. The wind farm layout optimization method considering load constraints according to claim 1, characterized in that, In step (2), the inflow condition data includes hub wind speed, turbulence intensity, and wind turbine yaw angle; the relevant parameters required for simulation include wind turbine aerodynamic data, elastic dynamics data, hydrodynamic data, servo system data, and structural mechanics data; the load simulation of the wind turbine is implemented based on writing an OpenFAST case setting file; the load proxy model training dataset includes hub wind speed, turbulence intensity, wind turbine yaw angle, and equivalent load data of each load channel within a certain time.
4. The wind farm layout optimization method considering load constraints according to claim 1, characterized in that, In step (3), the hyperparameter tuning method is specifically the Keras-Tuner library; the artificial neural network is specifically a fully connected neural network. The neural network input for each load channel is three variables, namely hub wind speed, turbulence intensity, and yaw angle, and the output is a single variable, namely the equivalent load of the load channel.
5. The wind farm layout optimization method considering load constraints according to claim 1, characterized in that, In step (4), the engineering wind field model includes the Bastankhah Gaussian wake velocity deficit model, the Steen Frandsen turbulence model, and the sum of squares superposition model; the lifecycle equivalent load calculation method is as follows: Where f is the sampling frequency of DEL, T Lifetime N represents the total number of seconds in the lifecycle of the wind turbine. eq The material cycle number is used to calculate the equivalent load. P(wd,ws) represents the frequency of wind direction wd and wind speed ws, DEL(wd,ws) represents the equivalent load of the wind turbine under wind conditions of wind direction wd and wind speed ws, and m is the Wöhler exponent of the material. The physical meaning of LDEL is: to calculate the equivalent load of the material under N conditions... eq The equivalent load required to achieve the same total fatigue damage as the entire lifespan in a second cycle.
6. A wind farm layout optimization device considering load constraints, comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that, When the processor executes the executable code, it implements a wind farm layout optimization method considering load constraints as described in any one of claims 1-5.
7. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements a wind farm layout optimization method considering load constraints as described in any one of claims 1-5.