Wind power plant multi-target coordination optimization control method considering wake effect

Abstract

The invention relates to a wind power plant multi-objective coordinated optimization control method considering wake effect, comprising the following steps: firstly, improving a wake effect model, taking a Jensen model as a basis, and providing a Jensen model after three times of correction according to the influence effect of each wake influence factor; then, in consideration of fan layout characteristics of the large-scale wind power plant, a model of wake flow superposition of a plurality of wind turbines is provided; on the basis, a double-layer optimization model with the maximum output power and the minimum load loss of the wind turbine generator set as objective functions is established, a wake effect model is combined, an SOM-FCM clustering algorithm is adopted for solving, on the basis of a traditional wake effect, a membership degree is combined with a weight value to form multi-condition constraints, the situation that the wind turbine generator set is caught in global optimum is avoided, and the optimal performance of the wind turbine generator set is improved. After clustering, fuzzy sample data are removed to improve the modeling precision; the method has the advantages that the Jensen wake effect model improved for three times is constructed, the multiple wind generator superposition model is introduced, the wake effect influence is effectively reduced, and the overall output efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of wind farm coordinated control, and in particular relates to a wind farm multi-objective coordinated optimization control method considering wake effects. Background Art

[0002] With the rapid development of the economy, the energy crisis has become increasingly serious and the environment has suffered unprecedented pollution. Wind power generation, as a renewable and clean form of energy, plays an important role in the global energy transition. Wind power generation has attracted much attention due to its wide distribution, abundant resources, and zero emissions. Its development has benefited from technological progress and cost reduction. The capacity and efficiency of wind turbines have continued to increase, and production costs have gradually decreased. Over time, wind power generation has become an indispensable and important part of the global energy structure and has made positive contributions to achieving clean energy goals. However, traditional wind power generation systems still have some challenges in wind turbine layout, wake effect control, optimization algorithms, etc., which affect their efficiency and stability. These challenges include the problem of efficiency reduction caused by mutual interference between wind turbines in traditional systems and the problem that the traditional single wind turbine layout limits the system energy utilization. Therefore, it is very necessary to provide a wind farm multi-objective coordinated optimization control method that considers the wake effect, constructs a three-time improved Jensen wake effect model, introduces a superposition model of multiple wind turbines, effectively reduces the impact of the wake effect, and improves the overall output efficiency. Summary of the Invention

[0003] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a wind farm multi-objective coordinated optimization control method that takes the wake effect into consideration by constructing a three-times improved Jensen wake effect model, introducing a superposition model of multiple wind turbines, effectively reducing the impact of the wake effect, and improving the overall output efficiency.

[0004] The object of the present invention is achieved by: a multi-objective coordinated optimization control method for a wind farm considering the wake effect, the method comprising the following steps:

[0005] Step 1: Improve the wake effect model. Based on the Jensen model, consider the effects of various wake influencing factors and propose a three-fold modified Jensen model. This reduces the impact of the wake effect on the output power and load loss of wind farm units, achieving optimal control under complex wake models.

[0006] Step 2: Considering the layout characteristics of wind turbines in large-scale wind farms, a model for the superposition of the wakes of multiple wind turbines is proposed;

[0007] Step 3: On this basis, a two-layer optimization model is established with the maximum output power of the wind turbine and the minimum load loss as the objective function;

[0008] Step 4: Combine the wake effect model and use the SOM-FCM clustering algorithm to solve the problem;

[0009] Step 5: Based on the traditional wake effect, membership and weights are combined to form multi-condition constraints to avoid falling into the global optimum. After clustering, fuzzy sample data are eliminated to improve modeling accuracy.

[0010] The Jensen wake model proposed in step 1 is modified three times, and each parameter is modified on this basis to construct an optimized calculation model based on the Jensen model, which specifically includes the following steps:

[0011] Step 1.1: The Jensen model has the following assumptions: ① The initial diameter of the wake is equal to the rotor diameter; ② The wake radius changes linearly; ③ The wind speed on the same cross section in the control body is equal. Under the above three assumptions, the wake model of formula (2) is obtained based on the mass conservation law equation of the wake control body of formula (1). Where v is the wake wind speed at x behind the rotor; v0 is the initial wind speed from upstream; r0 is the rotor radius; r w is the wake radius at x downstream of the wind rotor;

[0012] Step 1.2: Based on this, an improved wake model is derived: Where v0 is the initial wind speed of the wake; C T is the inference coefficient of the wind turbine; a is the axial coefficient;

[0013] Step 1.3: The first modified wake model mainly takes into account the air flow loss in the control volume. The loss is serious in the near-wake flow field and gradually becomes zero in the far-wake flow field. The ratio of the inflow flow to the total airflow mass is The formula for variation with distance is: Where β is the first-order correction coefficient, which represents the ratio of airflow flowing into the side boundary of the control body to compensate for the loss; is the ratio of inflow flow to total airflow;

[0014] Step 1.4: The secondary corrected wake model takes into account that the attenuation of the wake with distance is nonlinear, and introduces the attenuation factor k d , represents the attenuation of the wake effect with downstream distance, assuming the attenuation factor k d It has an exponential relationship with the downstream distance x of the wind turbine, that is: Combined with the obtained first-corrected Jensen wake expression, the second-corrected exponential decay wake model is obtained:

[0015]

[0016] Step 1.5: The three-times corrected wake model corrects the initial wake radius and initial wind speed. The factors affecting the initial wake radius due to the tip vortex are converted into a relationship with the axial coefficient. The corrected initial wake radius is calculated using the axial coefficient and the rotor radius. Then, the wake model after three corrections is obtained by combining equation (8): Where: is the corrected initial wake radius; r0 is the original initial wake radius and the rotor radius; a is the axial coefficient.

[0017] The modified wake model in step 1.3 is specifically: According to formula (4), the ratio of the external airflow added to the control body from the downstream distance x to infinity is further obtained. Then the ratio of the cross-sectional flow rate to the inflow rate at the downstream distance x of the wind turbine is Substituting this into the mass conservation equation, we get: The expression of the Jensen wake model after a correction is obtained by using equations (4) and (5):

[0018] The modification of the initial wake radius in step 1.5 will result in a decrease in the downstream wake wind speed. Let the modified initial wake wind speed be u x , the corrected initial inflow wind speed is The initial wind speed adjustment coefficient of the wake is λ, which is substituted into formula (11) to obtain formula (12): Substituting equation (12) into equation (10), we can obtain the wake model expression after triple correction:

[0019]

[0020] The superposition model of the wake effect in step 2 is specifically as follows: r1 is the radius of the wake area; r2 is the radius of the wind wheel; d is the distance from the center of the wake area to the center of the wind wheel; the intersection area A can be calculated based on the intersection area of ​​the wind turbine wake. j,i ,have to: The wake wind speed of upstream wind turbine j at downstream wind turbine i is: Where u j,i is the wake wind speed of upstream wind turbine j at downstream wind turbine i; u j is the incoming wind speed of wind turbine j; C Tj is the thrust coefficient of wind turbine j; when there are i-1 wind turbines in front of wind turbine i and the incoming wind speed at infinity is u0, the calculation formula for the wind speed and area intersection weight at wind turbine i is as follows: Where, α j,i A is the weight of the intersection of the wake area of ​​upstream wind turbine j and the rotor area of ​​downstream wind turbine i; j,iis the intersection area of ​​the wake area of ​​upstream wind turbine j and the rotor area of ​​downstream wind turbine i; D is the diameter of the rotor.

[0021] The two-layer optimization model in step 3 is specifically: Where, P sum is the total output power of the wind farm; N is the number of wind turbines in the wind farm; P i is the output power of the i-th unit in the wind farm; M sum is the load on the entire wind turbine in the wind farm; M Ti is the load on the tower of the i-th wind turbine in the wind farm; M Bi is the load on the blade of the i-th wind turbine in the wind farm.

[0022] The maximum output power and minimum load loss in step 3 are specifically achieved by adjusting the axial induction factor of the wind turbines in the wind farm to coordinate their output power and load size so that the overall power output of the wind farm is maximized and the load loss is minimized. The overall objective function is: maxF = λ1P sum -λ2M sum (20), where λ1 and λ2 are the wind turbine output power and the load weight value; P sum is the total output power of the wind farm; M sum It is the load on the entire wind turbine in the wind farm.

[0023] The steps 4 and 5 are specifically as follows: using the SOM-FCM algorithm combined with the wind farm equivalent model of the wake effect, taking wind speed-power as the clustering index and using membership-weight as the multi-condition constraint, optimizing the clustering of each unit in the wind farm, avoiding the influence of modeling accuracy by using only a single index, and improving the accuracy of the constructed model.

[0024] Beneficial effects of the invention: The present invention is a multi-objective coordinated optimization control method for wind farms taking into account the wake effect. It is a wind farm optimization method that combines the three-times modified Jensen model with a model of multiple wake superpositions, a multi-objective optimization model function and a SOM-FCM clustering-wake effect algorithm. In use, firstly, in response to the wake effect, the mutual interference between wind turbines in the traditional system leads to a decrease in efficiency. The present invention proposes a three-times improved Jensen wake effect model. By accurately simulating the interaction between wind turbines, the precise control and optimization of the wake effect are achieved, thereby improving the overall performance of the system; secondly, in response to the traditional single wind turbine layout that limits the energy utilization of the system In order to solve the problem of energy utilization rate, the present invention introduces a superposition model of multiple wind turbines, which effectively improves the energy utilization rate and power generation efficiency of the system by rationally arranging and managing the relationship between multiple wind turbines; finally, in order to better control system operation and optimize power generation efficiency, the present invention also introduces an improved fuzzy clustering-wake algorithm and a two-layer optimization mathematical model. These algorithms combine modern computer technology and optimization theory to quickly and accurately model and optimize the system, further improving the stability and performance of the system; the present invention has the advantages of constructing a three-time improved Jensen wake effect model, introducing a superposition model of multiple wind turbines, effectively reducing the impact of the wake effect, and improving the overall output efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 Flowchart of the present invention.

[0026] Figure 2 This is the wake model correction flow chart of the present invention.

[0027] Figure 3 This is a flow chart of wind farm equivalent modeling using the SOM-FCM algorithm of the present invention combined with the wake effect.

[0028] Figure 4 This is the power load optimization control flow chart of the present invention.

[0029] Figure 5 It is a schematic diagram of the intersection of the wake and the wind wheel of the present invention.

[0030] Figure 6 This is a control block diagram of the present invention. DETAILED DESCRIPTION

[0031] The present invention will be further described below with reference to the accompanying drawings.

[0032] Example 1

[0033] like Figure 1 As shown, a multi-objective coordinated optimization control method for a wind farm considering the wake effect is provided, the method comprising the following steps:

[0034] Step 1: Improve the wake effect model. Based on the Jensen model, consider the effects of various wake influencing factors and propose a three-fold modified Jensen model. This reduces the impact of the wake effect on the output power and load loss of wind farm units, achieving optimal control under complex wake models.

[0035] Step 2: Considering the layout characteristics of wind turbines in large-scale wind farms, a model for the superposition of the wakes of multiple wind turbines is proposed;

[0036] Step 3: On this basis, a two-layer optimization model is established with the maximum output power of the wind turbine and the minimum load loss as the objective function;

[0037] Step 4: Combine the wake effect model and use the SOM-FCM clustering algorithm to solve the problem;

[0038] Step 5: Based on the traditional wake effect, membership and weights are combined to form multi-condition constraints to avoid falling into the global optimum. After clustering, fuzzy sample data are eliminated to improve modeling accuracy.

[0039] The present invention is a multi-objective coordinated optimization control method for wind farms considering the wake effect. It is a wind farm optimization method that combines the three-times modified Jensen model with a model of multiple wake superpositions, a multi-objective optimization model function and a SOM-FCM clustering-wake effect algorithm. In use, firstly, in response to the wake effect, the mutual interference between wind turbines in the traditional system leads to a decrease in efficiency. The present invention proposes a three-times improved Jensen wake effect model. By accurately simulating the interaction between wind turbines, the precise control and optimization of the wake effect are achieved, thereby improving the overall performance of the system; secondly, in response to the problem that the traditional single wind turbine layout limits the energy utilization of the system To solve the problem, the present invention introduces a superposition model of multiple wind turbines, which effectively improves the energy utilization rate and power generation efficiency of the system by rationally arranging and managing the relationship between multiple wind turbines; finally, in order to better control the system operation and optimize the power generation efficiency, the present invention also introduces an improved fuzzy clustering-wake algorithm and a two-layer optimization mathematical model. These algorithms, combined with modern computer technology and optimization theory, can quickly and accurately model and optimize the system, further improving the stability and performance of the system; the present invention has the advantages of constructing a three-times improved Jensen wake effect model, introducing a superposition model of multiple wind turbines, effectively reducing the influence of the wake effect, and improving the overall output efficiency.

[0040] Example 2

[0041] like Figure 1-6 As shown, a multi-objective coordinated optimization control method for a wind farm considering the wake effect is provided, the method comprising the following steps:

[0042] Step 1: Improve the wake effect model. Based on the Jensen model, consider the effects of various wake influencing factors and propose a three-fold modified Jensen model. This reduces the impact of the wake effect on the output power and load loss of wind farm units, achieving optimal control under complex wake models.

[0043] Step 2: Considering the layout characteristics of wind turbines in large-scale wind farms, a model for the superposition of the wakes of multiple wind turbines is proposed;

[0044] Step 3: On this basis, a two-layer optimization model is established with the maximum output power of the wind turbine and the minimum load loss as the objective function;

[0045] Step 4: Combine the wake effect model and use the SOM-FCM clustering algorithm to solve the problem;

[0046] Step 5: Based on the traditional wake effect, membership and weights are combined to form multi-condition constraints to avoid falling into the global optimum. After clustering, fuzzy sample data are eliminated to improve modeling accuracy.

[0047] The three-times-corrected Jensen wake model proposed in step 1 is not able to accurately simulate the wind farm because the traditional Jensen wake model idealizes each parameter. Therefore, the three-times-corrected Jensen wake model is proposed. On this basis, each parameter is corrected and an optimization calculation model based on the Jensen model is derived. Figure 2 The flowchart for correcting the wake model in step 1 includes the following steps:

[0048] Step 1.1: The Jensen model has the following assumptions: ① The initial diameter of the wake is equal to the rotor diameter; ② The wake radius changes linearly; ③ The wind speed on the same cross section in the control body is equal. Under the above three assumptions, the wake model of formula (2) is obtained based on the mass conservation law equation of the wake control body of formula (1). Where v is the wake wind speed at x behind the rotor; v0 is the initial wind speed from upstream; r0 is the rotor radius; r w is the wake radius at x downstream of the wind rotor;

[0049] Step 1.2: On this basis, due to the actual operation of the wind turbine, the wind energy conversion coefficient C p <0.592. Taking into account the influence of wind turbine accessory turbulence, hub height and wind farm surface roughness on the wake wind speed and inference coefficient, an improved wake model is further derived: Where v0 is the initial wind speed of the wake; C T is the inference coefficient of the wind turbine; a is the axial coefficient;

[0050] Step 1.3: The first modified wake model mainly takes into account the air flow loss in the control volume. The loss is serious in the near-wake flow field and gradually becomes zero in the far-wake flow field. The ratio of the inflow flow to the total airflow mass is The formula for variation with distance is: Where β is the first-order correction coefficient, which represents the ratio of airflow flowing into the side boundary of the control body to compensate for the loss; is the ratio of inflow flow to total airflow;

[0051] Step 1.4: The secondary corrected wake model takes into account that the attenuation of the wake with distance is nonlinear, and introduces the attenuation factor k d , represents the attenuation of the wake effect with downstream distance, assuming the attenuation factor k d It has an exponential relationship with the downstream distance x of the wind turbine, that is: Combined with the obtained first-corrected Jensen wake expression, the second-corrected exponential decay wake model is obtained:

[0052]

[0053] Step 1.5: The three-times corrected wake model corrects the initial wake radius and initial wind speed. The factors affecting the initial wake radius due to the tip vortex are converted into a relationship with the axial coefficient. The corrected initial wake radius is calculated using the axial coefficient and the rotor radius. Then, the wake model after three corrections is obtained by combining equation (8): Where: is the corrected initial wake radius; r0 is the original initial wake radius and the rotor radius; a is the axial coefficient.

[0054] The modified wake model in step 1.3 is specifically: According to formula (4), the ratio of the external airflow added to the control body from the downstream distance x to infinity is further obtained. Then the ratio of the cross-sectional flow rate to the inflow rate at the downstream distance x of the wind turbine is Substituting this into the mass conservation equation, we get: The expression of the Jensen wake model after a correction is obtained by using equations (4) and (5):

[0055] The modification of the initial wake radius in step 1.5 will lead to a decrease in the downstream wake wind speed. Since the previous Jensen wake model compromised various wake influencing factors and thus ignored the initial wake radius, the initial wake wind speed decreased after the initial wake radius was modified. Let the modified initial wake wind speed be u x , the corrected initial inflow wind speed is The initial wind speed adjustment coefficient of the wake is λ, which is substituted into formula (11) to obtain formula (12): Substituting equation (12) into equation (10), we can obtain the wake model expression after triple correction:

[0056]

[0057] The superposition model of the wake effect in step 2 is specifically as follows: taking into account the wake wind speed in the wake interference area of ​​the units in the large-scale wind farm, further exploring the relationship between the wake wind speed, the wind turbine thrust coefficient and the intersection area between the units, the effective wind speed loss of the downstream wind turbine is smaller than that in the full coverage case, and the effective wind speed of the downstream wind turbine can be expressed by multiplying the speed loss of the single wake model by the area of ​​overlap between the wake and the wind rotor disk: Where, v is the area of ​​the overlap between the cross section of the wake of upstream wind turbine i at downstream wind turbine j and the plane swept by the rotor of wind turbine j; j,i is the effective wind speed of the downstream fan;

[0058] like Figure 5 As shown in the figure, o1 is the center of the wake area of ​​the upstream wind turbine j; o2 is the center of the rotor of the downstream wind turbine i; r1 is the radius of the wake area; r2 is the radius of the rotor; d is the distance from the center of the wake area to the center of the rotor; α is the angle between the straight line from the intersection of the wake area and the rotor to the center of the wake area and d; β is the angle between the straight line from the intersection of the wake area and the rotor to the center of the rotor and d;

[0059] Connect the center of the wake cross-section of wind turbine i at wind turbine j with the center of the circle swept by the rotor of wind turbine j, and use d to represent the distance from the center of the wake to the center of the rotor. Secondly, compare d with the radius of the two circles to determine the influence mode between wind turbines. Finally, calculate the area of ​​the overlapping part according to the triangle cosine theorem. The influence mode between wind turbines can be calculated by the following formula: Among them, A j,i is the area of ​​the rotor of the downstream wind turbine covered by the wake;

[0060] A j,i It can be calculated by the following formula: α and β are the arcs corresponding to the overlapping parts of the wake and the wind rotor, which can be calculated by the cosine theorem: Combining the above three expressions, the intersection area A can be calculated based on the wind turbine wake intersection area in the figure above. j,i ,have to: The wake wind speed of upstream wind turbine j at downstream wind turbine i is: Where u j,i is the wake wind speed of upstream wind turbine j at downstream wind turbine i; u jis the incoming wind speed of wind turbine j; C Tj is the thrust coefficient of wind turbine j; from formula (15), it can be seen that the wind speed at the downstream wind turbine i is only related to the thrust coefficient C of the upstream wind turbine j. Tj Changing the thrust coefficient of the upstream wind turbine can change the incoming wind speed of the downstream wind turbine. When there are i-1 wind turbines in front of wind turbine i and the incoming wind speed at infinity is u0, the calculation formula for the wind speed and area intersection weight at wind turbine i is as follows: Where, α j,i A is the weight of the intersection of the wake area of ​​upstream wind turbine j and the rotor area of ​​downstream wind turbine i; j,i is the intersection area of ​​the wake area of ​​upstream wind turbine j and the rotor area of ​​downstream wind turbine i; D is the diameter of the rotor.

[0061] In this embodiment, step 3 specifically includes: establishing a load loss model for wind turbines in a wind farm, performing load analysis on the tower and blades to obtain tower bending moments and blade bending moments, proposing a two-layer objective function for maximizing the overall output power of the wind farm and minimizing load loss, and further deriving an overall objective function, including the following steps:

[0062] Step 3.1: Wind farm single-unit maximum power tracking control method: The optimal tip speed ratio method is to maintain the tip speed ratio of the wind turbine at the optimal tip speed ratio when the wind speed of the wind farm changes, so that the wind turbine always maintains the maximum power output state under different wind speeds. When the wind speed changes, the wind speed is monitored in real time and the speed or blade angle of the wind turbine is adjusted according to the pre-set optimal tip speed ratio curve or algorithm to ensure that the wind turbine always maintains the maximum power output state. This can maximize the utilization of wind energy and improve the power generation efficiency of the wind farm.

[0063] Figure 6 For the control block diagram, the blade tip speed ratio of the input parameter is calculated by λ=ωr / v and the optimal blade tip speed ratio λ opt Make a comparison, and then use the control system to gradually make the tip speed ratio approach the optimal tip speed ratio;

[0064] Step 3.2: Mathematical model of wind farm wind turbine load loss: The load on the tower can be calculated by the tower bending moment M T The tower bending moment is mainly composed of the rotor torque, tower torque and nacelle eccentricity. For wind turbines above the megawatt level, the rotor torque and nacelle eccentricity can be ignored. Therefore, the tower bending moment M T It is mainly caused by the thrust T of the wind acting on the tower, and the formula is: M T =hT(25), Where h is the height of the wind turbine tower;

[0065] The load on the blade can be expressed as the blade bending moment M B To express it, the blade bending moment is caused by the axial force of the blade rotor and the tangential force of gravity. The blade bending moment M B The expression is as follows: Where m f is the mass of the wind turbine blade; g is the acceleration of gravity; substituting the tip speed ratio λ = ωr / v into formula (27), the blade bending moment expression is obtained:

[0066] Step 3.3: Optimize the control objective function and constraints: The maximum objective function of the wind farm's overall power output is: Where, P sum is the overall output power of the wind farm; N is the number of wind turbines in the wind farm; P i is the output power of the i-th wind turbine in the wind farm; a i is the axial induction factor of the i-th wind turbine in the wind farm;

[0067] The objective function of minimizing load loss is: Where M sum is the load on the entire wind turbine in the wind farm; M Ti is the load on the tower of the i-th wind turbine in the wind farm; M Bi is the load on the blade of the i-th wind turbine in the wind farm;

[0068] By adjusting the axial induction factor of the wind turbines in the wind farm to coordinate their output power and load size, the overall power output of the wind farm is maximized and the load loss is minimized. The overall objective function is: max F = λ1P sum -λ2M sum (20), where λ1 and λ2 are the wind turbine output power and the load weight value; P sum is the total output power of the wind farm; M sum is the load on the entire wind turbine in the wind farm; is the load on the unified calculation sum , M sum The dimension and magnitude of the wind farm are normalized by dividing the wind farm's overall output power and load value during single-machine maximum power tracking, and the final objective function is: maxF = λ1P sum / P MPPT -λ2M sum / M MPPT (33), the range of the axial induction factor in the objective function above is: 0≤a i ≤1 / 3(34), the power output constraint is: 0≤P i ≤P rate (35), where P rateis the rated output power of the wind turbine.

[0069] The steps 4 and 5 are specifically as follows: using the SOM-FCM algorithm in combination with the equivalent model of the wind farm with the wake effect, taking wind speed-power as the clustering index and membership-weight as the multi-condition constraint, optimizing and clustering each unit in the wind farm, avoiding the influence of modeling accuracy by using only a single index, and improving the accuracy of the built model; the SOM algorithm can automatically cluster according to the characteristics of the input data and give a more accurate number of clusters; the FCM algorithm determines the degree of membership of the data point to each class by calculating the Euclidean distance between each data and the cluster center, and then divides the data points according to the degree of membership to determine the final clustering result; specifically including the following steps:

[0070] Step 4.1: Wind speed probability density weight: Using the three-parameter Weibull distribution as the wind random model, the probability density function describing the statistical law of wind speed is expressed as: Where: x1 is the location parameter, indicating the starting position of the distribution curve; v j is the value of the j-th wind speed sample; α is the shape parameter, which represents the shape of the curve; β is the scale parameter; Integrating the above formula, we can get the wind speed distribution function expression F(v j ), represents the weight distribution at each wind speed, and its expression is:

[0071] Step 4.2: Implementation of SOM-FCM clustering algorithm: ① Initialization: First, for each weight node W in the output layer j (j=1,2,...,m) are randomly assigned small initial values ​​and normalized together with the input data set X. Normalization is to process vectors of different directions and lengths according to a fixed direction and unit length. The vector normalization formula is: The normalized vector obtained after initialization and As shown in the following formula: Where: W j is a weight node (j=1,2,...,m), which represents the influence of each data node on the network. j ∈[0,1];

[0072] ② Output the winning neuron and adjust the weights: Use the Euclidean algorithm to obtain each input vector of the input layer The geometric distance d between the nodes in the output layer, the Euclidean algorithm can express the similarity of the input data. The calculation formula is shown below, and the weight vector of the minimum Euclidean geometric distance d is obtained. Named the winning neuron, Where n is the total number of data points; the winning neuron outputs are represented by "1" and the other neurons outputs are represented by "0" respectively: In order to make other neurons in the attachment closer to the winning neuron, their weights are adjusted according to formula (42): Where: α is the learning rate, α∈[0,1], α determines the convergence speed of the objective function. The objective function formula is shown in formula (43). As the calculation α continues to decrease, the adjustment of the weight vector also continues to decrease, gradually approaching the cluster center. Where u ij is the membership degree; m is the fuzzy index;

[0073] ③ Re-normalize: After the above two steps, the unit vector changes and needs to be re-normalized, and then continue with step ② to make α continue to decay; if α>α min , then continue to enter the loop, if α≤α min , then the optimal number of clusters is obtained and the process goes to step ④;

[0074] ④ Calculate the membership matrix U according to the following formula: Where: x is the input data set, 1≤i≤C, 1≤j≤n; v is the cluster center;

[0075] ⑤ Update the cluster center V according to the following formula (t+1) , and repeat steps ④ and ⑤, Where: u ij is the membership degree; x j is the input dataset;

[0076] ⑥ After clustering, select the typical wind turbines in each center as samples and remove wind turbines that are far away from the cluster center. The removal principle is: Where p j is the value of the jth power sample; p i is the power value of the cluster center; ε and η are the thresholds of the wind speed and power values ​​in the sample from the cluster center (ε = 1.2, η = 120);

[0077] ⑦ The SOM-FCM algorithm model can obtain p clusters with similar characteristics and the corresponding cluster center values ​​v i , so it can be concluded that the mth equivalent wind speed v of the SOM-FCM algorithm combined with the wake effect is eq : Where: C T is the wind turbine thrust coefficient; R, R w are the impeller radius and the wake radius respectively; n2 is the number of fans after removing the fans; γ is defined as v eq The coefficients in ;

[0078] ⑧ According to the principle of the shortest distance, several groups of equivalent data closest to the cluster center can be identified. The obtained data sets can be used through formula (48) to obtain the equivalent output power P of the wind turbine.eq : The total power output of wind turbines P Σ for: The power of p clusters is calculated by weighted average to obtain the sum of the output power. The optimized clustering algorithm combined with the wake effect to obtain the equivalent output power has the characteristics of high accuracy and fast calculation speed.

[0079] Step 4.3: The equivalent modeling process of wind farm using SOM-FCM algorithm combined with wake effect is as follows: Figure 3 As shown;

[0080] Step 4.4: Optimal control of wind turbine power output for maximum and load loss: Based on wind turbine aerodynamic theory, assuming that the wind turbine can control the axial induction factor from 0 to 1 / 3, the wind turbine active power and load distribution are coordinated by controlling the axial induction factor to achieve optimal control for maximum wind farm power output and minimum load loss.

[0081] When the wind speed v is greater than the cut-in wind speed, the wind turbine starts to operate; when the wind speed v is less than v rate When the wind speed v rate When v<v1, the wind turbines that are not fully powered are operated using the power load optimization control method, and the wind turbines that are partially fully powered are operated using the control method that minimizes the load when the full power is maintained; when the wind speed v1<v<v cut When the wind speed v cut When the load of wind turbine is too high, it will be shut down for protection; the power load optimization control flow chart is as follows: Figure 4 shown.

[0082] The present invention is a multi-objective coordinated optimization control method for wind farms considering the wake effect. It is a wind farm optimization method that combines the three-times modified Jensen model with a model of multiple wake superpositions, a multi-objective optimization model function and a SOM-FCM clustering-wake effect algorithm. In use, firstly, in response to the wake effect, the mutual interference between wind turbines in the traditional system leads to a decrease in efficiency. The present invention proposes a three-times improved Jensen wake effect model. By accurately simulating the interaction between wind turbines, the precise control and optimization of the wake effect are achieved, thereby improving the overall performance of the system; secondly, in response to the problem that the traditional single wind turbine layout limits the energy utilization of the system To solve the problem, the present invention introduces a superposition model of multiple wind turbines, which effectively improves the energy utilization rate and power generation efficiency of the system by rationally arranging and managing the relationship between multiple wind turbines; finally, in order to better control the system operation and optimize the power generation efficiency, the present invention also introduces an improved fuzzy clustering-wake algorithm and a two-layer optimization mathematical model. These algorithms, combined with modern computer technology and optimization theory, can quickly and accurately model and optimize the system, further improving the stability and performance of the system; the present invention has the advantages of constructing a three-times improved Jensen wake effect model, introducing a superposition model of multiple wind turbines, effectively reducing the influence of the wake effect, and improving the overall output efficiency.

Claims

1. A multi-objective coordinated optimization control method for a wind farm considering wake effects, characterized by: The method comprises the following steps: Step 1: Improve the wake effect model. Based on the Jensen model, consider the effects of various wake influencing factors and propose a three-fold modified Jensen model. This reduces the impact of the wake effect on the output power and load loss of wind farm units, achieving optimal control under complex wake models. Step 2: Considering the layout characteristics of wind turbines in large-scale wind farms, a model for the superposition of the wakes of multiple wind turbines is proposed; Step 3: On this basis, a two-layer optimization model is established with the maximum output power of the wind turbine and the minimum load loss as the objective function; Step 4: Combine the wake effect model and use the SOM-FCM clustering algorithm to solve the problem; Step 5: Based on the traditional wake effect, membership and weights are combined to form multi-condition constraints to avoid falling into the global optimum. After clustering, fuzzy sample data are eliminated to improve modeling accuracy.

2. The multi-objective coordinated optimization control method for a wind farm considering wake effects according to claim 1, characterized in that: The Jensen wake model proposed in step 1 is modified three times, and each parameter is modified on this basis to construct an optimized calculation model based on the Jensen model, which specifically includes the following steps: Step 1.1: The Jensen model has the following assumptions: ① The initial diameter of the wake is equal to the rotor diameter; ② The wake radius changes linearly; ③ The wind speed on the same cross section in the control body is equal. Under the above three assumptions, the wake model of formula (2) is obtained based on the mass conservation law equation of the wake control body of formula (1). Where v is the wake wind speed at x behind the rotor; v0 is the initial wind speed from upstream; r0 is the rotor radius; r w is the wake radius at x downstream of the wind rotor; Step 1.2: Based on this, an improved wake model is derived: Where v0 is the initial wind speed of the wake; C T is the inference coefficient of the wind turbine; a is the axial coefficient; Step 1.3: The first modified wake model mainly takes into account the air flow loss in the control volume. The loss is serious in the near-wake flow field and gradually becomes zero in the far-wake flow field. The ratio of the inflow flow to the total airflow mass is The formula for variation with distance is: Where β is the first-order correction coefficient, which represents the ratio of airflow flowing into the side boundary of the control body to compensate for the loss; is the ratio of inflow flow to total airflow; Step 1.4: The secondary corrected wake model takes into account that the attenuation of the wake with distance is nonlinear, and introduces the attenuation factor k d , represents the attenuation of the wake effect with downstream distance, assuming the attenuation factor k d It has an exponential relationship with the downstream distance x of the wind turbine, that is: Combined with the obtained first-corrected Jensen wake expression, the second-corrected exponential decay wake model is obtained: Step 1.5: The three-times corrected wake model corrects the initial wake radius and initial wind speed. The factors affecting the initial wake radius due to the tip vortex are converted into a relationship with the axial coefficient. The corrected initial wake radius is calculated using the axial coefficient and the rotor radius. Then, the wake model after three corrections is obtained by combining equation (8): Where: is the corrected initial wake radius; r0 is the original initial wake radius and the rotor radius; a is the axial coefficient.

3. The multi-objective coordinated optimization control method for a wind farm considering wake effects according to claim 2, characterized in that: The modified wake model in step 1.3 is specifically: According to formula (4), the ratio of the external airflow added to the control body from the downstream distance x to infinity is further obtained. Then the ratio of the cross-sectional flow rate to the inflow rate at the downstream distance x of the wind turbine is Substituting this into the mass conservation equation, we get: The expression of the Jensen wake model after a correction is obtained by using equations (4) and (5):

4. The multi-objective coordinated optimization control method for a wind farm considering wake effects according to claim 2, characterized in that: The modification of the initial wake radius in step 1.5 will result in a decrease in the downstream wake wind speed. Let the modified initial wake wind speed be u x , the corrected initial inflow wind speed is The initial wind speed adjustment coefficient of the wake is λ, which is substituted into formula (11) to obtain formula (12): Substituting equation (12) into equation (10), we can obtain the wake model expression after triple correction:

5. The multi-objective coordinated optimization control method for a wind farm considering wake effects according to claim 1, characterized in that: The superposition model of the wake effect in step 2 is specifically as follows: r1 is the radius of the wake area; r2 is the radius of the wind wheel; d is the distance from the center of the wake area to the center of the wind wheel; the intersection area A can be calculated based on the intersection area of ​​the wind turbine wake. j,i ,have to: The wake wind speed of upstream wind turbine j at downstream wind turbine i is: Where u j,i is the wake wind speed of upstream wind turbine j at downstream wind turbine i; u j is the incoming wind speed of wind turbine j; C Tj is the thrust coefficient of wind turbine j; when there are i-1 wind turbines in front of wind turbine i and the incoming wind speed at infinity is u0, the calculation formula for the wind speed and area intersection weight at wind turbine i is as follows: Where, α j,i A is the weight of the intersection of the wake area of ​​upstream wind turbine j and the rotor area of ​​downstream wind turbine i; j,i is the intersection area of ​​the wake area of ​​upstream wind turbine j and the rotor area of ​​downstream wind turbine i; D is the diameter of the rotor.

6. The wind farm multi-objective coordinated optimization control method considering wake effect according to claim 1, characterized in that: The two-layer optimization model in step 3 is specifically: Where, P sum is the total output power of the wind farm; N is the number of wind turbines in the wind farm; P i is the output power of the i-th unit in the wind farm; M sum is the load on the entire wind turbine in the wind farm; M Ti is the load on the tower of the i-th wind turbine in the wind farm; M Bi is the load on the blade of the i-th wind turbine in the wind farm.

7. The wind farm multi-objective coordinated optimization control method considering wake effect according to claim 6, characterized in that: The maximum output power and minimum load loss in step 3 are specifically achieved by adjusting the axial induction factor of the wind turbines in the wind farm to coordinate their output power and load size so that the overall power output of the wind farm is maximized and the load loss is minimized. The overall objective function is: maxF = λ1P sum -λ2M sum (20), where λ1 and λ2 are the wind turbine output power and the load weight value; P sum is the total output power of the wind farm; M sum It is the load on the entire wind turbine in the wind farm.

8. The multi-objective coordinated optimization control method for a wind farm considering wake effects according to claim 1, characterized in that: The steps 4 and 5 are specifically as follows: using the SOM-FCM algorithm combined with the wind farm equivalent model of the wake effect, taking wind speed-power as the clustering index and using membership-weight as the multi-condition constraint, optimizing the clustering of each unit in the wind farm, avoiding the influence of modeling accuracy by using only a single index, and improving the accuracy of the constructed model.

Images