Wind power plant multi-target yaw control method considering power load comprehensive benefits

Abstract

The invention discloses a wind power plant multi-target yaw control method considering power load comprehensive benefits, and belongs to the technical field of wind power generation. The whole wind power plant wake flow modeling considering the yaw state is carried out by analyzing the wake flow model, the inflow data of each fan in the wind power plant is determined, and the total power and the average life of the wind power plant are rapidly and accurately predicted by combining the fan power, the thrust coefficient curve and the S-N curve of the transition section pipe node. Single fan power optimization and fan fatigue life optimization with power constraint are integrated, a two-stage multi-target control optimization system is established, and positioning of an optimal yaw combination of each stage is accelerated by establishing a Bayesian machine learning network and utilizing efficient and accurate power and life prediction data. According to the method, the optimal yaw strategy considering the power load comprehensive benefit can be accurately and efficiently provided according to the real-time wind speed data, and the load is reduced to the maximum extent and the service life of the fan is prolonged while the power generation requirement of the wind power plant is met.

Description

Technical Field

[0001] The invention belongs to the technical field of wind power generation, and in particular relates to a multi-objective yaw control method for a wind farm taking into account the comprehensive benefits of power load. Background Art

[0002] With the increasing global demand for renewable energy, wind power generation, as an important green energy, has become an important part of the energy structure transformation of various countries. As the core unit of wind power generation, the operating efficiency and economy of wind farms directly affect the overall contribution of wind power. However, in the actual operation of wind farms, the wake effect caused by the interaction between wind turbines, as well as the instability of wind speed and wind direction, will have a significant impact on the power output and fatigue load of wind turbines. Reasonable yaw control can significantly weaken the wake effect, which plays a vital role in optimizing the power generation efficiency of wind farms and reducing mechanical wear and fatigue damage.

[0003] Traditional wind farm yaw control methods mostly focus on the optimization of a single goal, usually giving priority to maximizing power output. For example, patent application CN 110728066 A discloses a wind farm sector optimization method and system, which combines the power curve determined by historical data and the effective wind speed calculated by the model to calculate the total power of the wind farm, and uses an optimization algorithm to determine the optimal yaw angle when the total power is maximum, which is applied to the rapid and effective optimization of the wind farm sector. Although this method can improve the power generation efficiency of the wind turbine to a certain extent, the fatigue load of the wind turbine components gradually emerges during long-term operation, especially when the wind turbine is in the wake area for a long time or the wind speed changes drastically. Over-reliance on power optimization may aggravate the fatigue damage of the wind turbine components, thereby shortening the service life of the wind turbine and increasing the cost of maintenance and replacement. Therefore, how to balance the power output and fatigue load of the wind turbine in the wind farm has become a key issue in the current research on wind farm control technology. In addition, considering the wind speed, wind direction changes and wake effect in the wind farm, a more efficient fatigue life prediction model is needed to consider the interaction between the wind turbine and the wind farm environment to achieve a more accurate fatigue life assessment. Summary of the invention

[0004] To solve the problems of the prior art, the present invention provides a multi-objective yaw control method for a wind farm that takes into account the comprehensive benefits of power load. Through a two-stage multi-objective control strategy, the method can optimize the yaw angle in real time and improve the power generation efficiency of the wind farm. At the same time, it can effectively extend the service life of the wind turbine and reduce the maintenance cost of the wind farm, thereby improving the overall economy and sustainability of the wind farm.

[0005] The present invention adopts the following technical solutions:

[0006] A multi-objective yaw control method for a wind farm considering the comprehensive benefits of power load comprises the following steps:

[0007] Step 1: Obtain the layout information of the wind farm, basic parameters of wind turbines and measured data of wind speed and direction;

[0008] The basic parameters of the wind turbine include: rotor diameter D, hub height H, power-wind speed curve, thrust coefficient curve and SN curve of the wind turbine transition section pipe node;

[0009] Step 2: Establish a local three-dimensional coordinate system with the current wind flow direction as the positive direction of the x-axis, and determine the local three-dimensional coordinate data of each wind turbine in the wind farm according to the layout information of the wind farm obtained in step 1;

[0010] Step 3: Based on the measured wind speed and direction data obtained in step 1 and the local three-dimensional coordinate data of the wind turbine obtained in step 2, the yaw analytical wake model is used to model the entire wind farm wake considering the yaw state, and a whole-field wake prediction model is established;

[0011] Step 4: Establish the relationship P(γ) between the wind farm yaw control strategy and the total power generation. Based on the wake prediction model for the entire field established in step 3, determine the inflow wind speed of each wind turbine in the wind farm. Combined with the power-wind speed curve, calculate the power of each wind turbine and add them up to get the total power of the wind farm.

[0012] Step 5: Establish the relationship between the wind farm yaw control strategy and the average fatigue life L(γ). Based on the wake prediction model for the entire field established in step 3, determine the inflow wind speed and turbulence of each wind turbine in the wind farm. Combined with the SN curve of the wind turbine transition section pipe node, calculate the fatigue life of each wind turbine and take the average value to obtain the average life of the wind farm.

[0013] Step 6: Based on the relationship P(γ) between the wind farm yaw control strategy and the total power generation established in step 4, combined with the Bayesian optimization learning network, the yaw control optimization considering the single objective function of power generation in the first stage is performed, and the optimal yaw combination and the corresponding total power are output as the baseline yaw strategy and baseline power. At the same time, the relationship between the wind farm yaw control strategy and the average fatigue life established in step 5 is used to calculate the fatigue life of the wind farm under the baseline yaw strategy as the baseline average life;

[0014] Step 7: Based on the relationship between the wind farm yaw control strategy and the total power generation P(γ), and the relationship between the wind farm yaw control strategy and the average fatigue life L(γ), combined with the Bayesian optimization learning network, the life extension factor relative to the baseline average life is taken as the optimization target, and based on the actual power generation demand of the wind farm, the allowable reduction factor relative to the baseline power is set as a constraint, and the second stage of multi-objective yaw control optimization considering the comprehensive benefits of power load is carried out;

[0015] Step 8: Output the final optimal yaw control strategy and its corresponding total power and average fatigue life, and visualize the wake distribution of the entire field to complete the multi-objective yaw control of the wind farm considering the comprehensive benefits of power load.

[0016] Preferably, the layout information of the wind farm in step 1 is the global three-dimensional coordinate data of each wind turbine in a global three-dimensional coordinate system with the east direction as the positive direction of the x-axis.

[0017] Preferably, in step 2, determining the local three-dimensional coordinate data of each wind turbine in the wind farm according to the layout information of the wind farm obtained in step 1 specifically comprises: converting the global three-dimensional coordinate data of each wind turbine obtained in step 1 into the local three-dimensional coordinate data of each wind turbine in a local three-dimensional coordinate system;

[0018] The transformation between the global 3D coordinate data and the local 3D coordinate data of the fan satisfies the following relationship:

[0019]

[0020] Among them, (x, y) is the global three-dimensional coordinate data of the wind turbine, (x1, y1) is the local three-dimensional coordinate data of the wind turbine, and θ is the angle between the local coordinate system and the global coordinate system in the counterclockwise direction.

[0021] Preferably, the process of establishing the whole wake prediction model in step 3 is:

[0022] Step 3.1: Number all wind turbines in ascending order of the horizontal axis (1,…,N), where N represents the number of wind turbines;

[0023] Step 3.2: According to the wind turbine number, determine the inflow wind speed and turbulence of each wind turbine in the wind farm in turn, where the inflow wind speed and turbulence of the first wind turbine are the measured wind speed data. For the i-th downstream wind turbine, its inflow condition is obtained by superimposing the wakes of all upstream wind turbines, where the inflow wind speed and turbulence superposition adopt the square sum and linear superposition methods respectively;

[0024] Step 3.3: According to the inflow wind speed and turbulence of each wind turbine, the inflow wind speed and turbulence at any position in the downstream wake area are calculated using the yaw analytical wake model that fully considers the influence of atmospheric turbulence, thrust coefficient and yaw angle. The velocity loss and additional turbulence are Gaussian and double Gaussian distributions in the radial direction, respectively, as shown in the following formula:

[0025]

[0026] Among them, U(s,r') is the inflow wind speed, I(s,r') is the turbulence, U0 and I0 are the upstream undisturbed wind speed and turbulence, C'T is the thrust coefficient considering the yaw effect, I a is the fan inlet turbulence, D is the fan rotor diameter, s is the wake diffusion distance, r' is the radial distance considering the yaw wake centerline offset, σ is the standard deviation of the Gaussian function, k1 and k2 are the wake diffusion coefficients.

[0027] Preferably, in step 5, the fatigue life calculation of each wind turbine specifically includes:

[0028] Step 5.1: Calculate the peak stress S according to the average wind speed and turbulence of the fan flow and the fan thrust coefficient curve. max and the mean stress S avg :

[0029]

[0030] Among them, U a and U a are the average wind speed and peak wind speed of the fan flow, r p is the pile radius, I z is the moment of inertia of the pile section, h is the height of the wind turbine hub, A is the swept area of ​​the wind rotor, and ρ is the air density;

[0031] Step 5.2: Calculate the alternating stress amplitude S based on the peak stress and average stress r At the same time, in order to consider the influence of average stress on the fatigue life of the structure, the Goodman curve is used to correct it and the equivalent alternating stress amplitude S is obtained. e :

[0032]

[0033] Among them, S b is the ultimate strength, S y is the yield limit, S m is the corrected mean stress;

[0034] Step 5.3: Based on the SN curve of the fan transition section pipe node, according to the equivalent alternating stress amplitude S e , calculate the corresponding fatigue life, that is, the number of failure cycles N:

[0035]

[0036] Where t is the thickness where fatigue cracks may grow, t ref is the reference thickness. For tubular nodes, it is 32 mm, and k is the thickness coefficient.

[0037] Preferably, in step 6, the yaw control optimization considering the single objective function of power generation in the first stage by using Bayesian machine learning specifically includes:

[0038] Step 6.1: randomly generate a set of yaw control strategies, use the relationship P(γ) between the yaw control strategy and the total power generation of the wind farm established in step 4 to obtain the corresponding total power, and generate an initial training set consisting of the yaw control strategy and the corresponding total power;

[0039] Step 6.2: The training set uses Gaussian process to establish the approximate probability distribution between the yaw control strategy and the total power, and obtains the expectation and variance of the power prediction probability distribution;

[0040] Step 6.3: Based on the expected value and variance, establish an acquisition function, search for the possible optimal yaw combination, and use the relationship between the wind farm yaw control strategy and the total power generation established in step 4 to obtain the corresponding total power, forming a new data point and incorporating it into the training set;

[0041] Step 6.4: Use the training set obtained in step 6.3 to repeatedly generate the power prediction probability distribution and search for the possible optimal yaw control strategy until the optimization convergence condition is met.

[0042] Preferably, in step 7, the second stage multi-objective yaw control optimization considering the comprehensive benefits of power load incorporates the power constraint considering the allowable reduction factor into the objective function life extension factor in the form of a penalty function, and establishes a relationship between a new target value and the yaw control strategy based on the relationship between the wind farm yaw control strategy and the total power generation established in step 4 and the relationship between the wind farm yaw control strategy and the average fatigue life established in step 5:

[0043]

[0044] Among them, L baseline and P baseline are the baseline mean life span and baseline power obtained in step 6 respectively, α0 is the allowable reduction coefficient of power, and λ is the penalty factor.

[0045] The present invention has the following beneficial effects:

[0046] The present invention adopts the advanced Qian & Ishihara yaw analytical wake model to fully consider the influence of atmospheric turbulence, thrust coefficient and yaw angle on wake distribution, accurately and quickly describe the wake field under the yaw state of the wind turbine, so as to simulate the relationship between the yaw control strategy of the wind farm and the total power generation under the actual operating state. At the same time, a new method for rapid prediction of fatigue life of wind farms is creatively proposed, which combines the speed and turbulence prediction based on the yaw wake model with the SN curve of the wind turbine transition section pipe node, and predicts the average fatigue life of the wind farm in real time, providing strong technical support for the application of wind farm optimization problems that require a large number of iterative calculations. On this basis, the power optimization of a single wind turbine and the fatigue life optimization of a wind turbine with power constraints are innovatively integrated to establish a two-stage multi-objective control optimization system, which minimizes the load and extends the service life of the wind turbine on the basis of meeting the power generation needs of the wind farm. In addition, the Bayesian optimization network can make full use of the advantage of a small number of samples to quickly converge to the global optimal solution, and accelerate the positioning of the optimal yaw combination in each stage of multi-objective control. The method described in the present invention is applicable to multi-objective real-time yaw control of any wind farm taking into account the comprehensive benefits of power load, and can accurately and efficiently provide the optimal yaw control strategy based on the measured wind speed data, effectively solving the problems of lack of effective fatigue life prediction means, reasonable power and fatigue life coupling method, and high-precision and efficient optimization algorithm in multi-objective yaw control of wind farms, thereby optimizing the power generation efficiency of wind farms while effectively extending the service life of wind turbines and achieving cost reduction and efficiency improvement. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic diagram of the method flow of the present invention.

[0048] Figure 2 This is a schematic diagram of the wind farm layout of Example 1.

[0049] Figure 3 (a) is a power-wind speed diagram of a wind turbine selected in a wind farm of Example 1; (b) is a thrust coefficient curve of Example 1.

[0050] Figure 4 (a) is a schematic diagram of the distribution of inflow velocity in the Qian&Ishihara yaw wake model; (b) is a schematic diagram of the turbulence distribution.

[0051] Figure 5 Schematic diagram of the Bayesian machine learning optimization process in step 6.

[0052] Figure 6 This is the single power optimization result of the first stage of the wind farm in Example 1.

[0053] Figure 7The multi-objective yaw control results under different allowable power reduction factors of the wind farm in Example 1 are: (a) α0 = 96%; (b) α0 = 97%; (c) α0 = 98%; (d) α0 = 99%; (e) α0 = 100%.

[0054] Figure 8 The actual power reduction and life extension of the wind farm in Example 1 vary with the allowable power reduction coefficient. DETAILED DESCRIPTION

[0055] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings and specific embodiments:

[0056] Example 1

[0057] like Figure 1 As shown, this embodiment takes a small wind farm consisting of 5 wind turbines arranged in a straight line as an example. Figure 2 As shown, the wind turbine model is Vestas V80 2MW. In practical applications, the time-series flow information can be equivalent to the addition of multiple steady-state flow fields considering the average inflow within the control interval. Therefore, this embodiment is aimed at a certain steady-state fixed inflow: the wind speed at the height of the wind turbine hub is 12m / s, and the turbulence is 4%. The specific steps are:

[0058] Combination Figures 1 to 8 A multi-objective yaw control method for a wind farm considering the comprehensive benefits of power load is characterized by comprising the following steps:

[0059] Step 1: Obtain the layout information of the wind farm, basic parameters of wind turbines and measured data of wind speed and direction.

[0060] The layout information of the wind farm is the global three-dimensional coordinate data of each wind turbine in a global three-dimensional coordinate system with the east direction as the positive direction of the x-axis.

[0061] The basic parameters of the wind turbine in this embodiment are that the interval between adjacent wind turbines is 5D, the wind turbine rotor diameter D=80m, the hub height H=70m, and the power-wind speed and thrust coefficient curves are as follows: Figure 3 As shown, the SN curve of the fan transition section pipe node satisfies the following relationship:

[0062]

[0063] The measured data of wind speed and direction are as follows: the inflow wind speed of the wind farm is 12m / s, the turbulence is 4%, and the wind direction is along the axis of the wind turbine arrangement.

[0064] Step 2: Establish a local three-dimensional coordinate system with the current wind flow direction as the positive direction of the x-axis, and determine the local three-dimensional coordinate data of each wind turbine in the wind farm according to the layout information of the wind farm obtained in step 1.

[0065] The global three-dimensional coordinate data of each wind turbine obtained in step 1 is converted into the local three-dimensional coordinate data of each wind turbine in the local three-dimensional coordinate system.

[0066] The transformation between the global 3D coordinate data and the local 3D coordinate data of the fan satisfies the following relationship:

[0067]

[0068] Among them, (x, y) is the global three-dimensional coordinate data of the wind turbine, (x1, y1) is the local three-dimensional coordinate data of the wind turbine, and θ is the angle between the local coordinate system and the global coordinate system in the counterclockwise direction.

[0069] The local three-dimensional coordinate data of the five wind turbines obtained in this embodiment are (0, 0), (5D, 0), (10D, 0), (15D, 0), and (20D, 0).

[0070] Step 3: Based on the measured wind speed and direction data obtained in step 1 and the local three-dimensional coordinate data of the wind turbine obtained in step 2, the yaw analytical wake model is used to model the entire wind farm wake considering the yaw state, and a whole-field wake prediction model is established.

[0071] Step 3.1: Number all wind turbines in ascending order of the horizontal axis (1,…,5).

[0072] Step 3.2: According to the wind turbine number, determine the inflow wind speed and turbulence of each wind turbine in the wind farm in turn. The inflow wind speed and turbulence of the first wind turbine are measured wind speed data, representing the undisturbed free flow situation. For the downstream i-th wind turbine, its inflow condition (u i ,I i ) is obtained by superimposing the wakes of all upstream wind turbines, where the inlet wind speed and turbulence are superimposed using the square sum and linear superposition methods, respectively.

[0073] The speed loss superposition adopts the square sum model: The turbulent kinetic energy superposition adopts a linear model: The subscript j indicates the upstream fan of fan i.

[0074] u i represents the inflow wind speed of the i-th fan, u j represents the inflow wind speed of the jth fan, u ij represents the wake velocity of the jth wind turbine at the ith wind turbine; k i represents the inflow turbulent kinetic energy of the i-th wind turbine, k0 represents the turbulent kinetic energy of the upstream undisturbed flow, Δk ij is the additional turbulent kinetic energy of the wake of the jth wind turbine at the ith wind turbine.

[0075] Step 3.3: According to the inflow wind speed and turbulence of each wind turbine, the Qian & Ishihara yaw analytical wake model that fully considers the influence of atmospheric turbulence, thrust coefficient and yaw angle is used to calculate the inflow wind speed and turbulence at any position in the downstream wake area, where the velocity loss and additional turbulence are Gaussian and double Gaussian distributions along the radial direction, respectively, as follows: Figure 4 And the following formula:

[0076]

[0077] Among them, U(x,r') is the inflow wind speed, I(x,r') is the turbulence, U0 and I0 are the upstream undisturbed wind speed and turbulence, U0 = 12m / s, I0 = 4%, C' T is the thrust coefficient considering the yaw effect, I a is the fan inlet turbulence, D is the fan rotor diameter, s is the wake diffusion distance, r' is the radial distance considering the yaw wake centerline offset, σ is the standard deviation of the Gaussian function, k1 and k2 are the wake diffusion coefficients.

[0078] Step 4: Establish the relationship between the wind farm yaw control strategy and the total power generation P(γ). Based on the whole field wake prediction model established in step 3, determine the effective inflow wind speed of each wind turbine in the wind farm, and combine the power-wind speed curve ( Figure 3 a), calculate the power of each wind turbine and add them up to get the total power of the wind farm.

[0079] Step 5: Establish the relationship L(γ) between the wind farm yaw control strategy and the average fatigue life. Based on the wake prediction model for the entire field established in step 3, determine the effective inflow wind speed and turbulence of each wind turbine in the wind farm. Combined with the SN curve of the wind turbine transition section pipe node, calculate the fatigue life of each wind turbine and take the average value to obtain the average life of the wind farm.

[0080] The fatigue life calculation of each fan specifically includes:

[0081] Step 5.1: Based on the average wind speed and turbulence of the fan flow, combined with the fan thrust coefficient curve ( Figure 3 b), calculate the peak stress S max and the mean stress S avg :

[0082]

[0083] Among them, U a and U a are the average wind speed and peak wind speed of the fan flow, r p is the pile radius, I z is the moment of inertia of the pile section, h is the height of the wind turbine hub, A is the swept area of ​​the wind rotor, and ρ is the air density.

[0084] Step 5.2: Calculate the alternating stress amplitude S based on the peak stress and average stress r At the same time, in order to consider the influence of average stress on the fatigue life of the structure, the Goodman curve (Goodman fatigue limit line) is used to correct it and obtain the equivalent alternating stress amplitude S e :

[0085]

[0086] Among them, S b is the ultimate strength, S y is the yield limit, S m is the corrected mean stress.

[0087] Step 5.3: Based on the SN curve of the fan transition section pipe node, according to the equivalent alternating stress amplitude S e , calculate the corresponding fatigue life, that is, the number of failure cycles N:

[0088]

[0089] Where t is the thickness where fatigue cracks may grow, t ref is the reference thickness, for tubular nodes, t ref =32mm, k is the thickness coefficient.

[0090] The mean of fatigue life of all wind turbines in the wind farm is taken as the average life of the wind farm L, and finally the relationship between the wind farm yaw control strategy γ and the average fatigue life L is established:

[0091]

[0092] N i (γ) represents the fatigue life of the i-th fan.

[0093] Step 6: Based on the relationship P(γ) between the wind farm yaw control strategy and the total power generation established in step 4, combined with the Bayesian optimization learning network, the first stage yaw control optimization considering the single objective function of power generation is carried out, and the optimal yaw combination and the corresponding total power are output as the baseline yaw strategy and baseline power.

[0094] At the same time, the relationship L(γ) between the wind farm yaw control strategy and the average fatigue life established in step 5 is used to calculate the fatigue life of the wind farm under the baseline yaw strategy as the baseline average life.

[0095] The first stage of yaw control optimization using Bayesian machine learning with a single objective function of power generation specifically includes:

[0096] Step 6.1: Randomly generate a set of yaw control strategies, use the relationship between the wind farm yaw control strategy and the total power generation P(γ) established in step 4 to obtain the corresponding total power, and generate an initial training set D consisting of yaw control strategies and corresponding total power. 1 , the superscript 1 indicates the initial training set used for the first optimization iteration.

[0097] Step 6.2: For the training set D, use the Gaussian process to establish the approximate probability distribution between the yaw control strategy and the total power, and obtain the power prediction probability distribution P(γ|D n )’s expectation and variance N(μ,σ 2 ), the superscript n represents the training set used for the nth optimization iteration.

[0098] Step 6.3: Based on the obtained expectation and variance, establish an acquisition function. This embodiment adopts the expected improvement (EI) function form to search for the possible optimal yaw combination, and use the relationship P(γ) between the wind farm yaw control strategy and the total power generation power established in step 4 to obtain the corresponding total power, forming a new data point and incorporating it into the training set D.

[0099] Step 6.4: Use the training set D obtained in step 6.3 to repeatedly generate the power prediction probability distribution and search for the possible optimal yaw control strategy until the optimization convergence condition is met.

[0100] This embodiment finally obtains Figure 6 The optimal yaw combination shown in the figure outputs the corresponding total power and fatigue life as the baseline power P baseline and the baseline mean life span L baseline .

[0101] Step 7: Based on the single power optimization results of the first stage, according to the relationship between the wind farm yaw control strategy and the total power generation, and the relationship between the wind farm yaw control strategy and the average fatigue life, combined with the Bayesian optimization learning network, the relative baseline average life L baseline The life extension factor β(γ) is taken as the optimization target. Based on the actual power generation demand of the wind farm, different allowable reduction factors α0 relative to the baseline power are set as constraints, and the multi-objective yaw control optimization considering the comprehensive benefits of power load is carried out in the second stage.

[0102] maximize

[0103] Satisfy the condition -π / 6<γ i <π / 6

[0104] Power reduction constraint: P(γ)≥α0P baseline

[0105] In the second stage, the multi-objective yaw control optimization considering the comprehensive benefits of power load is carried out. The power constraint is incorporated into the life extension factor in the form of a penalty function to establish a new objective function Target(γ), which is used to replace the single power objective function in the Bayesian learning framework established in step 6. According to the relationship between the wind farm yaw control strategy and the total power generation established in step 4 and the relationship between the wind farm yaw control strategy and the average fatigue life established in step 5, the relationship between the new target value and the yaw control strategy is established:

[0106]

[0107] Among them, L baseline and P baseline are the baseline mean life span and baseline power obtained in step 6 respectively, α0 is the allowable reduction coefficient of power, and λ is the penalty factor.

[0108] Step 8: Output the final optimal yaw control strategy and its corresponding total power and average fatigue life, and visualize the wake distribution of the entire field to complete the multi-objective yaw control of the wind farm considering the comprehensive benefits of power load.

[0109] Figure 7 The results of multi-objective yaw control of the selected wind farm under different allowable power reduction factors are shown, represented by a velocity distribution cloud diagram, where α0 = 100% represents the situation without baseline power reduction, which corresponds to the single power target optimization in the first stage. As can be seen from the figure, under different allowable power reduction factors, the yaw angle distribution shows a trend of gradually decreasing along the flow direction, among which the first wind turbine has a more obvious wake deflection relative to other downstream wind turbines. However, with the increase of allowable power reduction, the positive effect of the yaw of the rear wind turbines on extending the fatigue life of the wind farm gradually emerges, and larger yaw angles begin to appear. Figure 8 The actual power reduction under different allowable power reduction coefficients and the obtained fatigue life growth are further quantitatively compared. As can be seen from the figure, the change in the actual power reduction is always consistent with the allowable power reduction coefficient, and the obtained life extension factor shows a trend of first rapid growth and then a gradual slowdown in growth as the allowable power reduction increases. For this embodiment, α0 = 98% is the dividing point, that is, a 2% power reduction can obtain an average life extension of nearly 8%, which represents the maximum life growth benefit that can be achieved with a unit power reduction. This embodiment fully verifies that the multi-objective yaw control method for wind farms that considers the comprehensive benefits of power load can achieve a significant increase in the average life of wind farms with a smaller power reduction, thereby achieving the best comprehensive benefits of power load for yaw control of wind farms.

[0110] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the essential scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A multi-objective yaw control method for a wind farm considering the comprehensive benefits of power load, characterized in that: The following steps are involved: Step 1: Obtain the layout information of the wind farm, basic parameters of wind turbines and measured data of wind speed and direction; The basic parameters of the wind turbine include: rotor diameter D, hub height H, power-wind speed curve, thrust coefficient curve and SN curve of the wind turbine transition section pipe node; Step 2: Establish a local three-dimensional coordinate system with the current wind flow direction as the positive direction of the x-axis, and determine the local three-dimensional coordinate data of each wind turbine in the wind farm according to the layout information of the wind farm obtained in step 1; Step 3: Based on the measured wind speed and direction data obtained in step 1 and the local three-dimensional coordinate data of the wind turbine obtained in step 2, the yaw analytical wake model is used to model the entire wind farm wake considering the yaw state, and a whole-field wake prediction model is established; Step 4: Establish the relationship P(γ) between the wind farm yaw control strategy and the total power generation. Based on the wake prediction model for the entire field established in step 3, determine the inflow wind speed of each wind turbine in the wind farm. Combined with the power-wind speed curve, calculate the power of each wind turbine and add them up to get the total power of the wind farm. Step 5: Establish the relationship between the wind farm yaw control strategy and the average fatigue life L(γ). Based on the wake prediction model for the entire field established in step 3, determine the inflow wind speed and turbulence of each wind turbine in the wind farm. Combined with the SN curve of the wind turbine transition section pipe node, calculate the fatigue life of each wind turbine and take the average value to obtain the average life of the wind farm. Step 6: Based on the relationship P(γ) between the wind farm yaw control strategy and the total power generation established in step 4, combined with the Bayesian optimization learning network, the yaw control optimization considering the single objective function of power generation in the first stage is performed, and the optimal yaw combination and the corresponding total power are output as the baseline yaw strategy and baseline power. At the same time, the relationship between the wind farm yaw control strategy and the average fatigue life established in step 5 is used to calculate the fatigue life of the wind farm under the baseline yaw strategy as the baseline average life; Step 7: Based on the relationship between the wind farm yaw control strategy and the total power generation P(γ), and the relationship between the wind farm yaw control strategy and the average fatigue life L(γ), combined with the Bayesian optimization learning network, the life extension factor relative to the baseline average life is taken as the optimization target, and based on the actual power generation demand of the wind farm, the allowable reduction factor relative to the baseline power is set as a constraint, and the second stage of multi-objective yaw control optimization considering the comprehensive benefits of power load is carried out; Step 8: Output the final optimal yaw control strategy and its corresponding total power and average fatigue life, and visualize the wake distribution of the entire field to complete the multi-objective yaw control of the wind farm considering the comprehensive benefits of power load.

2. A wind farm multi-objective yaw control method considering the comprehensive benefits of power load according to claim 1, characterized in that: The layout information of the wind farm in step 1 is the global three-dimensional coordinate data of each wind turbine in a global three-dimensional coordinate system with the east direction as the positive direction of the x-axis.

3. A wind farm multi-objective yaw control method considering the comprehensive benefits of power load according to claim 2, characterized in that: In step 2, the local three-dimensional coordinate data of each wind turbine in the wind farm is determined according to the layout information of the wind farm obtained in step 1, specifically: the global three-dimensional coordinate data of each wind turbine obtained in step 1 is converted into the local three-dimensional coordinate data of each wind turbine in the local three-dimensional coordinate system; The transformation between the global 3D coordinate data and the local 3D coordinate data of the fan satisfies the following relationship: Among them, (x, y) is the global three-dimensional coordinate data of the fan, (x1, y1) is the local three-dimensional coordinate data of the fan, and θ is the angle between the local coordinate system and the global coordinate system in the counterclockwise direction.

4. A multi-objective yaw control method for a wind farm considering the comprehensive benefits of power load according to claim 1, characterized in that: The process of establishing the whole wake prediction model in step 3 is as follows: Step 3.1: Number all wind turbines in ascending order of the horizontal axis (1,…,N), where N represents the number of wind turbines; Step 3.2: According to the wind turbine number, determine the inflow wind speed and turbulence of each wind turbine in the wind farm in turn, where the inflow wind speed and turbulence of the first wind turbine are the measured wind speed data. For the i-th downstream wind turbine, its inflow condition is obtained by superimposing the wakes of all upstream wind turbines, where the inflow wind speed and turbulence superposition adopt the square sum and linear superposition methods respectively; Step 3.3: According to the inflow wind speed and turbulence of each wind turbine, the inflow wind speed and turbulence at any position in the downstream wake area are calculated using the yaw analytical wake model that fully considers the influence of atmospheric turbulence, thrust coefficient and yaw angle. The velocity loss and additional turbulence are Gaussian and double Gaussian distributions in the radial direction, respectively, as shown in the following formula: Among them, U(s,r') is the inflow wind speed, I(s,r') is the turbulence, U0 and I0 are the upstream undisturbed wind speed and turbulence, C' T is the thrust coefficient considering the yaw effect, I a is the fan inlet turbulence, D is the fan rotor diameter, s is the wake diffusion distance, r' is the radial distance considering the yaw wake centerline offset, σ is the standard deviation of the Gaussian function, k1 and k2 are the wake diffusion coefficients.

5. The multi-objective yaw control method for a wind farm considering the comprehensive benefits of power load according to claim 1, characterized in that: The fatigue life calculation of each fan in step 5 specifically includes: Step 5.1: Calculate the peak stress S according to the average wind speed and turbulence of the fan flow and the fan thrust coefficient curve. max and the mean stress S avg : Among them, U a and U a are the average wind speed and peak wind speed of the fan flow, r p is the pile radius, I z is the moment of inertia of the pile section, h is the height of the wind turbine hub, A is the swept area of ​​the wind rotor, and ρ is the air density; Step 5.2: Calculate the alternating stress amplitude S based on the peak stress and average stress r At the same time, in order to consider the influence of average stress on the fatigue life of the structure, the Goodman curve is used to correct it and the equivalent alternating stress amplitude S is obtained. e : Among them, S b is the ultimate strength, S y is the yield limit, S m is the corrected mean stress; Step 5.3: Based on the SN curve of the fan transition section pipe node, according to the equivalent alternating stress amplitude S e , calculate the corresponding fatigue life, that is, the number of failure cycles N: Where t is the thickness where fatigue cracks may grow, t ref is the reference thickness. For tubular nodes, it is 32 mm, and k is the thickness coefficient.

6. A multi-objective yaw control method for a wind farm considering the comprehensive benefits of power load according to claim 1, characterized in that: In step 6, the yaw control optimization using Bayesian machine learning in the first stage considering the single objective function of power generation specifically includes: Step 6.1: randomly generate a set of yaw control strategies, use the relationship P(γ) between the yaw control strategy and the total power generation of the wind farm established in step 4 to obtain the corresponding total power, and generate an initial training set consisting of the yaw control strategy and the corresponding total power; Step 6.2: The training set uses Gaussian process to establish the approximate probability distribution between the yaw control strategy and the total power, and obtains the expectation and variance of the power prediction probability distribution; Step 6.3: Based on the expected value and variance, establish an acquisition function, search for the possible optimal yaw combination, and use the relationship between the wind farm yaw control strategy and the total power generation established in step 4 to obtain the corresponding total power, forming a new data point and incorporating it into the training set; Step 6.4: Use the training set obtained in step 6.3 to repeatedly generate the power prediction probability distribution and search for possible optimal yaw control strategies until the optimization convergence conditions are met.

7. A wind farm multi-objective yaw control method considering the comprehensive benefits of power load according to claim 1, characterized in that: In step 7, the second stage multi-objective yaw control optimization considering the comprehensive benefits of power load is incorporated into the objective function life extension factor in the form of a penalty function. According to the relationship between the wind farm yaw control strategy and the total power generation established in step 4 and the relationship between the wind farm yaw control strategy and the average fatigue life established in step 5, a new relationship between the target value and the yaw control strategy is established: Among them, L baseline and P baseline are the baseline mean life span and baseline power obtained in step 6 respectively, α0 is the allowable reduction coefficient of power, and λ is the penalty factor.

Images