A wind farm location optimization method based on grid-coordinate genetic algorithm

Abstract

The present invention discloses a method for optimizing the position of a wind farm based on a grid-coordinate genetic algorithm, specifically: using the Jensen wake model to predict the wake velocity of a wind turbine, using the square superposition model of the wake effect of the wind turbine to obtain the total kinetic energy loss in the wake of the wind turbine, and then minimizing the cost of power generation according to the optimization objective function; combining the grid genetic algorithm with the coordinate genetic algorithm, firstly optimizing the number and layout of wind turbines in the wind farm by the grid genetic algorithm, and then further optimizing the initial value of the layout of the wind turbines by the coordinate genetic algorithm, so that the total power generation of the wind farm is further increased. The present invention not only solves the problem of optimizing the number of wind turbines in the optimization of the layout of the wind farm and its high computational cost, but also avoids the limitation of the grid genetic algorithm on the flexibility of the location of the wind turbines; it provides a feasible method for optimizing the layout of the wind farm, and greatly increases its efficiency.

Description

Technical Field

[0001] The invention belongs to the technical field of wind farm planning, and in particular relates to a wind farm machine position optimization method based on a grid-coordinate genetic algorithm. Background Art

[0002] Wind energy is mainly obtained by wind turbines, but due to the wake effect, the airflow passing through the wind turbine will decrease in speed behind the wind turbine, reducing the power generation efficiency of the wind turbine behind. Although the wake effect can be reduced by increasing the distance between turbines, the site of the wind farm is limited in actual applications. Increasing the distance between wind turbines will reduce the number of wind turbines that can be installed, reduce the total power generation of the wind farm, and thus affect its economic benefits. Therefore, optimizing the layout of wind turbines, reducing the impact of the wake effect on the overall power generation of the wind farm, and thus increasing the power generation of the entire wind farm has become an important research topic.

[0003] In 1994, Mosetti et al. [1] first proposed the use of genetic algorithms to determine the number and location of wind turbines to obtain the maximum energy extraction at the lowest installation cost. Grady et al. [2] further improved the optimization by using a larger population size and number of iterations. Mittal [3] found through further research that grid density has a key impact on the optimization problem of wind farms. High-density grids produce better optimization results, but require high computational costs. Therefore, Hassoine et al. [4] used real number coding for optimization. Based on these studies, Chen [5] considered optimizing the power generation efficiency and unit power generation cost of wind farms at the same time and proposed the use of a multi-objective genetic algorithm. In order to improve the efficiency of the algorithm, Liu et al. [6] introduced an adaptive genetic algorithm with position swapping. Yang et al. [7] found that by changing the fixed probability of the initial population generated by the genetic algorithm and the mutation probability into a dynamic probability, the population diversity can be increased, the results can be further optimized, and an improved algorithm was proposed. In addition to genetic algorithms, researchers have also studied other different optimization algorithms. In recent years, Eroglu et al. [8] studied the WELOP algorithm based on particle filtering and optimized three different situations using the particle filtering method. The results were comparable to those of the genetic algorithm. Bilbao et al. [9] used the simulated annealing method to achieve the maximum power generation of the wind farm. Beatriz et al.

[10] first set the initial random layout through a heuristic algorithm, and then used nonlinear mathematical programming technology to perform local optimization to maximize the power generation of the wind farm, which was significantly improved. There are also improved firefly algorithms, sequential convex optimization algorithms (sequential convex programming algorithms), improved particle swarm algorithms, etc. [11-13]. However, considering the accuracy and efficiency of optimization, the genetic algorithm is still the most commonly used optimization algorithm in the wind farm layout optimization problem.

[0004] The genetic algorithms currently studied are divided into two types: grid-based and coordinate-based. When the grid-based genetic algorithm optimizes the layout of a wind farm, the wind turbines can only be placed in the center of the grid. The flexibility and accuracy of the wind turbine layout are greatly limited, and as the grid density increases, the computational cost is high. The coordinate-based genetic algorithm can only be used when the number of wind turbines is determined. A very important research topic is to determine the number of wind turbines.

[0005] References:

[0006] [1]MOSETTI G, POLONI C, DIVIACCO B. Optimization of wind turbinepositioning in large wind farms by means of a genetic algorithm[J]. Journal of wind engineering&industrial aerodynamics, 1994, 51(1): 105-116.

[0007] [2]GRADY SA, HUSSAINI MY, ABDULLAH M. Placement of wind turbines using genetic algorithms[J]. Renew energy, 2005, 30(1): 259-270.

[0008] [3]MITTAL A.Optimization of the layout of large wind farms using agenetic algorithm[C] / / International Mechanical Engineering Congress&Exposition,Vol.7.Texas,USA,20 12.

[0009] [4]HASSOINE M A,LAHLOU F,ADDAIM A,et al.Wind farm layout optimizationusing real coded multi-population genetic algorithm[C] / / InternationalConference on Wireless Technologies,Embedded and Intelligent Systems,USMBAUniv,ESNA Fez,ERSI&IPI Lab,Fez,MOROCCO,2019.

[0010] [5]CHEN Y,LI H,HE B,et al.Multi-objective genetic algorithm basedinnovative wind farm layout optimization method[J].Energy conversion&management,2015,105(NOV):1318-1327.

[0011] [6]LIU F,WANG Z F.Offshore wind farm layout optimization usingadapted genetic algorithm:a different perspective[J].Electrical and ComputerEngineering,2014,103(3):917-922.

[0012] [7]YANG Q S,HU J X,LAW S.Optimization of wind farm layout withmodified genetic algorithm based on boolean code[J].Journal of windengineering and industrial aerodynamics,2018,181:61-68.

[0013] [8]EROGLU Y,SECKINER S U.Design of wind farm layout using ant colonyalgorithm[J].Renewable energy,2012,44:53-62.

[0014] [9] BILBAO M, ALBA E.Simulated annealing for optimization of wind farmannual profit[C] / / International Symposium on Logistics&IndustrialInformatics, IEEE, Linz, Austria, 2009.

[0015]

[10] BEATRIZ P, ROBERTO M, RAUL G. Offshore wind farm layout optimization using mathematical programming techniques [J]. Renewable energy, 2013, 53(MAY): 389-399.

[0016]

[11] Liu Yongqian, Shao Zhenzhou, Yan Lingwei, et al. Research on micro-site selection optimization of wind farms based on improved binary firefly algorithm [J]. Renewable Energy, 2019, 37(1): 112-119.

[0017]

[12] Zhang Wei. Optimal layout of wind farms in complex terrain based on improved particle swarm optimization algorithm and wind speed distribution regression function method [J]. Hydropower Energy Science, 2016, 34(1): 190-194.

[0018]

[13] PARK J, LAW K H.Layout optimization for maximizing wind farm powerproduction using sequential convex programming[J].Applied energy, 2015, 151(aug1): 320-334. Summary of the invention

[0019] In order to overcome the above problems and further improve the power generation of wind farms, the present invention provides a method for optimizing wind farm machine locations based on a grid-coordinate genetic algorithm.

[0020] The wind farm machine location optimization method based on a grid-coordinate genetic algorithm of the present invention is specifically as follows:

[0021] Step 1: Use the Jensen wake model to predict the wake velocity of the wind turbine, and use the square superposition model of the wind turbine wake effect to obtain the total kinetic energy loss in the wind turbine wake; the superposition model is divided into four operating conditions (including full overlap, no overlap, and partial overlap) according to the degree of influence of the wind turbine wake on the downstream wind turbines, and then the cost of power generation is minimized according to the optimization objective function.

[0022] Step 2: Combine the grid genetic algorithm and the coordinate genetic algorithm. The method used is divided into two steps:

[0023] (1) The number of wind turbines and the optimal layout that minimize the objective function are optimized through a grid-based 0-1 coded genetic algorithm.

[0024] Under given wind farm and wind turbine parameters, consider that wind turbines are not placed within the specified range between wind turbines, divide the area into grids of specified size, and set the number of grids obtained in the given area as m. If the wind turbine is placed in the grid, it is recorded as 1, and if it is not placed, it is recorded as 0. The arrangement of wind turbines is converted into a binary code represented by a length of m.

[0025] The initial population is generated randomly, updated and iterated through crossover and mutation, and the fitness of the individuals in the population is calculated with the objective function as the fitness function. The roulette wheel and the best individual retention method are combined to select the best individual. The initial optimal layout of the wind turbine can be obtained by decoding the best individual. The position of gene 1 in the individual code is the placement position of the wind turbine, and the total number of genes 1 is the number of wind turbines.

[0026] (2) The final optimization of the coordinate genetic algorithm is to re-optimize the location of the wind turbines based on the obtained number and layout of the wind turbines, and further improve the power generation.

[0027] Assume that the optimal number of wind turbines in the wind farm obtained in step (1) is n turbines, and the position of the wind turbines is represented by vector coordinates, denoted as (x i ,y i ), i = 1, ..., n, the wind farm is analyzed in the plane coordinate axis, the incoming wind direction is set to be along the negative direction of the y-axis, and the wind turbines are reordered from large to small according to the ordinate y. If the ordinate y is equal, the horizontal coordinate x is sorted from small to large. After the new sorting, the i-th unit is only affected by the wake of the previous i-1 units.

[0028] The wind turbine layout optimized by the grid genetic algorithm is further optimized and converted into the optimization of the wind turbine coordinates. The coordinate genetic algorithm is used to encode directly by the coordinates, and the objective function is used as the individual fitness calculation method to select individuals. Through the update and iteration of the population, the optimal individual is obtained, which is the optimal layout of the wind farm, thereby obtaining the optimal power generation and objective function.

[0029] Furthermore, no wind turbines are placed within a specified range between wind turbines, and the distance is set to 5D, where D is the diameter of the wind turbine blades; this constraint is also applicable in the coordinate genetic algorithm.

[0030] The beneficial technical effects of the present invention are:

[0031] The present invention not only solves the problem of optimizing the number of wind turbines in wind farm layout optimization and its high computational cost, but also avoids the limitation of the grid genetic algorithm on the flexibility of wind turbine locations. More importantly, it provides a feasible method for optimizing the layout of wind farms, which greatly increases its efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a flow chart of wind farm machine position optimization according to the present invention.

[0033] Figure 2 is the Jensen wake model.

[0034] Figure 3 There are four wake overlap area conditions.

[0035] Figure 4 This is a schematic diagram of wind turbine sorting.

[0036] Figure 5 There are three types of wind conditions.

[0037] Figure 6 is the wind speed probability distribution of wind condition three.

[0038] Figure 7 This is the optimized wind field layout for wind condition 1 (a is gridding and b is coordinate).

[0039] Figure 8 This is a schematic diagram of coordinate transformation for wind condition 2.

[0040] Fig. 9 This is the optimized wind field layout for wind condition 2 (a is gridding, b is coordinates).

[0041] Fig.10 This is the optimized wind field layout for wind condition three (a is gridding, b is coordinate).

[0042] Fig.11 This is the power curve used by the wind turbine in wind condition three.

[0043] Fig.12 This is the wind speed probability distribution of the improved wind condition three.

[0044] Fig.13 This is the optimized wind field layout for the improved wind condition three (a is gridding, b is coordinates). DETAILED DESCRIPTION

[0045] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0046] The present invention firstly optimizes the number and layout of wind turbines in a wind farm by using a grid genetic algorithm, and then further optimizes the initial layout of the wind turbines by using a coordinate genetic algorithm, so that the total power generation of the wind farm is further increased.

[0047] A wind farm machine location optimization method based on a grid-coordinate genetic algorithm of the present invention is as follows Figure 1 As shown, specifically:

[0048] Step 1: Use the Jensen wake model (such as Figure 2 The wind turbine wake velocity is predicted by using the square superposition model of the wind turbine wake effect to obtain the total kinetic energy loss in the wind turbine wake. The superposition model is further divided into four working conditions according to the degree of influence of the wind turbine wake on the downstream wind turbine (such as Figure 3 As shown, there are two types: full overlap, no overlap, and partial overlap), and then the cost of power generation is minimized according to the optimization objective function.

[0049] Step 2: Combine the grid genetic algorithm and the coordinate genetic algorithm. The method used is divided into two steps:

[0050] (1) The number of wind turbines and the optimal layout that minimize the objective function are optimized through a grid-based 0-1 coded genetic algorithm.

[0051] Under given wind farm and wind turbine parameters, considering that no wind turbine is placed within a specified range between wind turbines (set to 5D in the present invention), the area is divided into grids of specified size, and the number of grids obtained in the given area is set to m. If a wind turbine is placed in the grid, it is recorded as 1, and if it is not placed, it is recorded as 0. The arrangement of the wind turbines is converted into a binary code represented by a length of m.

[0052] The initial population is generated randomly, and updated and iterated through crossover and mutation. Unlike the traditional genetic algorithm, the improved genetic algorithm changes the generation probability of 0 and 1 in the initial population and mutation process from fixed probability to dynamic probability to increase population diversity. The fitness of the individuals in the population is calculated with the objective function as the fitness function, and the roulette wheel and the best individual retention method are combined to select the best individual. The initial optimal layout of the wind turbine can be obtained by decoding the best individual; the position of gene 1 in the individual code is the placement position of the wind turbine, and the sum of the number of gene 1 is the number of wind turbines.

[0053] (2) The final optimization of the coordinate genetic algorithm is to re-optimize the location of the wind turbines based on the obtained number and layout of the wind turbines, and further improve the power generation.

[0054] Assume that the optimal number of wind turbines in the wind farm obtained in step (1) is n turbines, and the position of the wind turbines is represented by vector coordinates, denoted as (x i ,y i ), i = 1, ..., n, the wind farm is analyzed in the plane coordinate axis, such as Figure 4 As shown in the figure, the incoming wind direction is set to be along the negative direction of the y-axis, and the wind turbines are reordered from large to small according to the ordinate y. If the ordinate y is equal, they are sorted from small to large according to the abscissa x. After the new sorting, the i-th unit is only affected by the wake of the previous i-1 units.

[0055] The wind turbine layout optimized by the grid genetic algorithm is further optimized and converted into the optimization of wind turbine coordinates. The coordinate genetic algorithm is used to encode directly by coordinates, and the objective function is used as the individual fitness calculation method to select individuals. The optimization of the objective function needs to be combined with the constraints that wind turbines are no longer placed within the 5D range of the wind farm boundary and wind turbines. Through the update and iteration of the population, the optimal individual is obtained, which is the optimal layout of the wind farm, thereby obtaining the optimal power generation and objective function.

[0056] Example:

[0057] For the 2km×2km wind farm in the classic case, the layout optimization is carried out by considering the optimal power generation of the wind farm in one hour under unit cost, and the power generation is determined by formula (1). The wind turbine and wind farm information used are shown in Table 1.

[0058] The power generation calculation formula is:

[0059]

[0060] Where: C P ——Wind turbine efficiency, C P =4a(1-a) 2; ρ——air density, determined by the wind turbine power curve.

[0061] Table 1 Wind turbine and wind farm information

[0062]

[0063]

[0064] Consider 3 wind conditions, see Figure 5 The specific wind speed and direction are as follows:

[0065] Wind condition 1: constant wind direction and speed (12m / s);

[0066] Wind condition 2: 36 wind directions with the same frequency and constant wind speed (12m / s);

[0067] Wind condition 3: 36 wind directions, variable wind speed (8, 12, 17 m / s), the distribution probability of each wind direction is as follows Figure 6 .

[0068] Wind Condition Example 1:

[0069] The optimization results are shown in Table 2 and the layout diagram is shown in Figure 7 .

[0070] Table 2 Optimization results of wind condition case 1

[0071]

[0072] With the same number of wind turbines, the position of wind turbine coordinates is further optimized. The newly optimized wind turbine positions are more concentrated at the upper and lower ends, and appropriately dispersed in the middle, which can reduce the wake effect to a greater extent and increase the total power generation of wind turbines. Compared with the best existing results, the power generation is increased by 7.70%, and the objective function is significantly improved.

[0073] Wind Condition 2 Case Study:

[0074] For the optimized layout of multiple wind directions, further processing is required. When the incoming wind direction is not directly facing the wind turbine, the angle between the incoming wind direction and the wind turbine is θ. Through Cartesian coordinate transformation, the wind turbine coordinate (x, y) is transformed into (x′, y′), as follows: Figure 8 , so that the wind direction after transformation is facing the wind turbine, and the transformed coordinates are substituted into the objective function for optimization, where the Cartesian coordinate transformation formula and process are:

[0075]

[0076] Substitute into the optimization, the results are shown in Table 3, and the layout diagram is shown in Fig. 9 .

[0077] Table 3 Optimization results of wind condition case 2

[0078]

[0079]

[0080] according to Fig. 9 After analysis, it is found that since the probability of 36 wind directions at constant speed is equal, the distribution of wind turbines tends to be symmetrical and more uniform. However, because there is no grid restriction, there are more layout options for staggered placement of wind turbines, which increases the total power generation by 1.68% for the same number of wind turbines and further reduces the objective function.

[0081] Wind Condition 3 Case Study:

[0082] For the third wind condition, since the wind direction also comes from 36 wind directions, the same Cartesian coordinate transformation as wind condition 2 is first performed. However, unlike wind condition 2, the probability of wind directions at different wind speeds is different, so it is necessary to classify and discuss them, and then sum the classifications. The calculation formula is:

[0083]

[0084] Where: n is the number of wind turbines; N m is the number of wind directions; N s is the wind speed quantity; v j is the incoming wind speed; θ k is the deflection angle between the incoming wind direction and the wind turbine; q(v j ,θ k ) is the wind direction angle θ k And the wind speed is v j The probability of wind speed under different wind directions is 1; P i,j,k The corresponding incoming wind speed of the i-th wind turbine is v j And the wind direction angle is θ k The amount of power generated when.

[0085] Substituting wind condition 3 into the results is shown in Table 4, and the wind farm layout is shown in Fig.10 .

[0086] Table 4 Optimization results of three wind condition cases

[0087]

[0088] For wind conditions with variable wind speed and direction, and different probabilities of wind speed and direction, due to the high frequency of wind direction between 270° and 350°, the number of wind turbines placed around the wind farm is large and evenly distributed, while the number of wind turbines in the center is relatively small and scattered. After further optimization, the total power generation of the wind farm increased by 2.26% on the existing results.

[0089] The case-based analysis shows that the method proposed in the present invention has significant improvements. However, the three wind condition cases (multiple wind speeds and multiple wind directions) result in an average wind speed of 14 m / s, which is relatively high and quite different from the actual wind conditions in the wind farm (the average wind speed is 5-8 m / s). Moreover, when calculating the power, formula (1) does not take into account the cut-in and cut-out wind speeds of the wind turbines, and the small wind turbines used therein are difficult to represent the currently used large wind turbines. Therefore, based on actual engineering applications, a wind turbine with a rated power of 3 MW is used to optimize the optimal annual power generation of the wind farm at unit cost. The information of the wind turbines is shown in Table 5, and the power curve used is shown in Fig.11 .

[0090] Table 5 Wind turbine and wind farm information

[0091]

[0092] The wind farm area is determined by the longitude and latitude coordinates of the given six points A (117.27°, 23.42°), B (117.3°, 23.43°), C (117.33°, 23.39°), D (117.33°, 23.38°), E (117.3°, 23.36°), and F (117.24°, 23.4°). First, the longitude and latitude are converted into XY plane coordinates through arcgis to obtain the wind farm area projected on the plane coordinates. Considering that the placement of wind turbines cannot exceed the boundary, a distance of a wind turbine diameter is reserved for the boundary, and the area where wind turbines can be placed is obtained as follows Fig.13 The constraint condition that no wind turbines are placed within the specified interval between wind turbines is required to be met. This paper sets it to 5D, divides the wind farm area into grids with a specified side length, and removes some incomplete grids at the boundaries of the wind farm, thereby obtaining a complete grid where all wind turbines can be placed.

[0093] Similar to the layout optimization of classic wind farms, the grid genetic algorithm is first used to optimize the number of wind turbines and the initial layout. Further, the coordinate genetic algorithm is used for re-optimization. The constraints of the boundary and the spacing between wind turbines need to be considered during optimization. The specific constraints are:

[0094]

[0095] Based on the constraints shown in formula (4), the coordinate genetic algorithm is used to further optimize the location of the wind turbines to obtain the optimal layout diagram and objective function.

[0096] The wind condition case 3 is improved to make it closer to the actual wind farm. Similarly, this wind farm still uses 36 wind directions (adjacent wind directions are different), but the wind speed and wind speed frequency are adjusted. See Fig.12 The average wind speed of the improved wind farm is 7.48m / s.

[0097] The results of the three cases of improved wind conditions are shown in Table 6, and the wind farm layout is shown in Fig.13 .

[0098] Table 6 Optimization results of three cases of improved wind conditions

[0099]

[0100] Under the condition of multiple wind speeds and directions, the annual power generation increased by 1.77% compared with the grid genetic algorithm, and the annual power generation increased by 7963MWh. The further optimized layout makes the wind turbines more dispersed near the boundary line, maximizes the use of the wind farm area to reduce the wake effect, and increases the total power generation of the wind farm to further optimize the objective function.

Claims

1. A wind farm machine location optimization method based on a grid-coordinate genetic algorithm, characterized in that: Specifically: Step 1: Use the Jensen wake model to predict the wake velocity of the wind turbine, and superimpose the square of the wake effect of the wind turbine to obtain the total kinetic energy loss in the wake of the wind turbine; the superposition model is divided into four working conditions according to the degree of influence of the wind turbine wake on the downstream wind turbine, and then the cost of power generation is minimized according to the optimization objective function; Step 2: Combine the grid genetic algorithm and the coordinate genetic algorithm. The method used is divided into two steps: (1) Optimize the number and optimal layout of wind turbines to minimize the objective function through a gridded 0-1 coded genetic algorithm; Under the given wind farm and wind turbine parameters, consider that there is no wind turbine placed within the specified range between wind turbines, divide the area into grids of specified size, and set the number of grids obtained in the given area as m. If the wind turbine is placed in the grid, it is recorded as 1, and if it is not placed, it is recorded as 0. Then the arrangement of wind turbines is converted into a binary code with a length of m. The initial population is generated randomly, updated and iterated through crossover and mutation, and the fitness of the individuals in the population is calculated with the objective function as the fitness function. The best individuals are selected by combining roulette and the best individual retention method. The initial optimal layout of the wind turbines can be obtained by decoding the best individuals. The position of gene 1 in the individual code is the placement position of the wind turbine, and the total number of gene 1 is the number of wind turbines. (2) The final optimization of the coordinate genetic algorithm is to re-optimize the location of the wind turbines based on the number and layout of the wind turbines obtained, and further improve the power generation; Assume that the optimal number of wind turbines obtained by the wind farm in step (1) is n turbines, and the position of the wind turbines is represented by vector coordinates, denoted as (x i ,y i ), i = 1, ..., n, the wind farm is analyzed in the plane coordinate axis, the incoming wind direction is set to be along the negative direction of the y-axis, and the wind turbines are reordered from large to small according to the ordinate y. If the ordinate y is equal, the horizontal coordinate x is sorted from small to large. After the new sorting, the i-th unit is only affected by the wake of the previous i-1 units; The wind turbine layout optimized by the grid genetic algorithm is further optimized and converted into the optimization of the wind turbine coordinates. The coordinate genetic algorithm is used to encode directly by the coordinates, and the objective function is used as the individual fitness calculation method to select individuals. Through the update and iteration of the population, the optimal individual is obtained, which is the optimal layout of the wind farm, thereby obtaining the optimal power generation and objective function.

2. The wind farm machine location optimization method based on grid-coordinate genetic algorithm according to claim 1 is characterized in that: No wind turbine is placed within the specified range between the wind turbines, and the distance is set to 5D, where D is the diameter of the wind turbine blade; this constraint is also applicable when using the coordinate genetic algorithm.

3. The wind farm machine location optimization method based on grid-coordinate genetic algorithm according to claim 1 is characterized in that: The four working conditions specifically include full overlap, no overlap and two partial overlaps.

Images