On-site historical evolution behavior analysis method for aerodynamic performance of large-scale wind turbine generator

Abstract

The invention discloses an on-site historical evolution behavior analysis method for aerodynamic performance of a large wind turbine generator, which takes a power coefficient as a basic index for on-site analysis of the aerodynamic performance, and overcomes the influence of two adverse factors, namely deviation and randomness, of SCADA (Supervisory Control And Data Acquisition) wind speed data by adopting data compensation and data filtering. The relation between the wind speed and the wind wheel rotating speed is regarded as an inertial system, and a data driving time constant determination method is provided. A practical calculation working condition region division algorithm is designed, and a data binning method and a conservative data comparison method are adopted to solve the problems of scatter data singularization and interval critical value determination which need to be solved during working condition division respectively. Then, power coefficients based on the instantaneous power and the cumulative power are calculated, respectively. The five-year deterioration degree of the two units is respectively given by observing the effectiveness of the algorithm in a multi-angle manner through four indexes, the historical evolution of the on-site aerodynamic performance of the units is described, and the result shows that the method is effective and feasible.

Description

Technical Field

[0001] The present invention relates to a method for analyzing the on-site historical evolution behavior of aerodynamic performance of a large wind turbine generator set, and belongs to the technical field of aerodynamics in the wind energy industry. Background Art

[0002] The wind power industry has great potential. Wind turbines are devices that convert wind energy into electrical energy. The ability to capture air kinetic energy is closely related to the aerodynamic performance of the unit. Therefore, when designing wind turbines, the analysis of structure and aerodynamic performance is the focus of attention. Some aerodynamic performance analysis methods, such as the classic blade element-momentum theory (BEM) and computational fluid dynamics (CFD), have received widespread attention. For example, Amoretti et al. conducted aerodynamic design of a twin-rotor wind turbine based on the BEM theory, while Vaz et al. considered the influence of wake on the wind rotor plane and established a wind turbine design model based on the BEM theory. Esfahanian et al. used CFD and BEM hybrid analysis to simulate the flow field around the wind rotor of the unit to predict aerodynamic performance. In the previous study, the author's team constructed a wind turbine aerodynamic load calculation model that combined the BEM theory and the dynamic stall model. The combination of aerodynamic design and control design can design a wind turbine power characteristic curve as the basis for the unit operation control strategy. With the increase in service life, the performance degradation caused by the natural aging of wind turbines is inevitable. This problem has become increasingly serious as a large number of wind turbines enter their "middle and old" years. Research on wind turbine aging and its evaluation has attracted the attention of more and more researchers. In wind farms, describing the historical evolution of the aerodynamic performance of on-site units is a difficult problem related to the operation, maintenance and optimization of the units. First, because wind turbines are super-large mechanical structures, it is impossible to conduct aerodynamic flow field tests on site. Second, the on-site data is limited and there are no operating parameters that directly describe the aerodynamic performance.

[0003] For the units in service on site, it is not realistic to add an additional hardware test system. Making full use of the existing SCADA data is a feasible strategy. The SCADA data records many operating parameter data such as the nacelle wind speed, the rotor (generator) speed, and the generator power, which can indirectly reflect the aerodynamic performance of the wind turbine. For example, the aerodynamic performance of the wind turbine will affect its power coefficient, and the power coefficient can be calculated based on the SCADA data, so that a correlation is established between the aerodynamic performance of the unit and the SCADA parameters. However, SCADA data is often not used directly, and some zero values, null values ​​and abnormal values ​​interspersed in it need to be processed. In recent years, scholars from various countries have carried out targeted data preprocessing research. Xiang et al. used the power curve as a benchmark and used the quartile method to distribute SCADA data to eliminate abnormal power data; in order to solve the problem of small-amplitude sensor data being "swamped" by large-amplitude sensor data, Sun et al. normalized the multi-sensor data within a similar range; Yao et al. proposed a combined method to process abnormal values ​​of SACAD data by identifying abnormal data types. However, general data preprocessing cannot solve the "randomness" and "deviation" problems of wind speed. These two problems are the main difficulties that affect the reliable calculation of the on-site power coefficient of wind turbines. The randomness of wind speed means that the wind speed is mixed with high-frequency components. From the perspective of control theory, for a large inertia system such as a wind rotor, high-frequency wind speed will not affect its speed change. When performing power coefficient analysis, its influence should be suppressed as much as possible, otherwise it will lead to deviations in the power coefficient calculation. The deviation of wind speed means that the wind speed measured by the nacelle anemometer is smaller than the wind speed at the front of the wind rotor, and deviation compensation is required. Although some literature has carried out research on wind speed filtering, there is a lack of effective filtering methods that meet the actual characteristics of wind turbines. For example, the problem of how to determine the filter cutoff frequency (inertial system time constant) has not been effectively solved. Summary of the invention

[0004] Based on the above background technology analysis, it can be known that with the increase of service life, the performance degradation caused by the natural aging of wind turbines is inevitable. Describing the historical evolution behavior of the aerodynamic performance of the on-site unit is a difficult problem related to the operation, maintenance and optimization of the unit. To this end, the present invention proposes an analysis method based on on-site SCADA data, taking the power coefficient as the basic indicator of on-site analysis of aerodynamic performance, and using data compensation and data filtering to overcome the influence of the two unfavorable factors of "bias" and "randomness" of SCADA wind speed data. The relationship between wind speed and wind wheel speed is regarded as an inertial system, and a data-driven time constant determination method is proposed. A calculation condition area division algorithm that conforms to reality is designed, and the data binning method and the "conservative" data comparison method are used to solve the problems of "single-valued" scattered data and interval critical value determination that need to be solved when dividing the working conditions. Then, the power coefficient based on instantaneous power and cumulative power is calculated respectively. The effectiveness of the multi-angle observation algorithm with four indicators is used to give the 5-year degradation degree of two units respectively, and the historical evolution of the on-site aerodynamic performance of the unit is described. The results show that the proposed method is effective and feasible.

[0005] The technical solution adopted by the present invention is a method for analyzing the on-site historical evolution behavior of aerodynamic performance of large wind turbines, which comprises:

[0006] Step 1, determine the basic evaluation index of the on-site historical evolution behavior of the aerodynamic performance of the wind turbine; according to the blade element theory and on-site aerodynamic performance analysis, determine the power coefficient as the basic evaluation index for the on-site analysis of the aerodynamic performance of the wind rotor blades, and provide an evaluation index for the on-site historical evolution behavior analysis of the aerodynamic performance of the wind turbine in steps 2, 3, and 4.

[0007] Step 2, preprocessing of on-site wind speed data; considering the flow field characteristics of the on-site wind speed data, preprocessing the collected SCADA wind speed data, using data compensation to overcome the deviation of the SCADA wind speed data, determining the system time from the preprocessed SCADA wind speed data itself, proposing a data-driven time constant determination method, determining a reasonable filtering time constant, overcoming the randomness of the SCADA wind speed data, and providing more accurate wind speed data for calculating the power coefficient in step 4.

[0008] Step 3, calculation of working condition area division; design a calculation working condition area division algorithm that conforms to the actual characteristics of blade field data, and use data binning method and "conservative" data comparison method to solve the problems of "single-valued" scattered data and interval critical value determination that need to be solved when dividing working conditions, so as to provide accurate maximum wind energy tracking working condition area MPPT for step 4 power coefficient calculation.

[0009] Step 4: Calculate the power coefficient based on the field data; calculate the power coefficient based on the instantaneous power and cumulative power, and propose a four-index multi-angle observation algorithm, that is, the interval power coefficient mean value obtained based on the instantaneous power calculation and the maximum value C PmaxI , the mean of the interval power coefficient calculated based on the cumulative power and the maximum value C PmaxA Four evaluation indicators are used to observe the on-site historical evolution of aerodynamic performance of large wind turbines.

[0010] Further, the implementation process of step 1 is as follows: the lift and drag acting on the blade element are written as dL and dD respectively, and according to the principle of force composition and decomposition, they are written as normal force and tangential force dF n 、dF t The torque dM acting on a blade ring of the wind wheel is:

[0011]

[0012] Where B is the number of wind wheel blades, ρ is the air density, v0 is the synthetic air velocity, C t is the tangential force coefficient, and r is the radius of the circle where the current blade element is located.

[0013] The axial force dT acting on a blade ring of the wind wheel can be written as:

[0014]

[0015] Where c is the chord length of the blade section, C n is the normal force coefficient.

[0016] Assume that the rotor radius is R and the rotor rotation speed is ω. According to equations (1) and (2), the torque M, axial force T and rotor shaft power P acting on the entire rotor are: rt for:

[0017]

[0018]

[0019] P rt =Mω (5)

[0020] Wind rotor power coefficient C p for

[0021]

[0022] From the above analysis process, it can be seen that the description of the aerodynamic performance of the blade involves multiple aspects such as lift characteristics, drag characteristics, thrust characteristics, torque characteristics, etc. These factors are all matters that need to be fully considered in the blade design stage. However, in wind farms, there are very few quantities that can be directly observed on site, and the description of the aerodynamic performance of the blade can only be mined and analyzed from SCADA data. According to formula (6), the wind rotor power coefficient can be calculated through power and wind speed data, plus air density information. Therefore, in wind farms, the power coefficient is the basic indicator for analyzing the aerodynamic performance of wind rotor blades.

[0023] Furthermore, the implementation process of step 2 is as follows;

[0024] Step 2.1 Wind speed data compensation: In the SCADA system, the wind speed data comes from the wind speed sensor installed on the nacelle. The wind speed before the wind rotor is set to v1, and the wind speed when passing through the wind rotor is v d , the wind speed after passing through the wind wheel is set to v2, obviously v1>v2, under relatively stable wind speed conditions:

[0025]

[0026] Where P is the generator power and S is the swept area of ​​the wind rotor.

[0027] According to formula (7), the relationship between v1 and v2 is:

[0028]

[0029] Step 2.2 Wind speed data filtering: The relationship between wind speed and wind rotor speed is approximately described by a first-order inertial system, expressed as:

[0030]

[0031] Where W(s) and V(s) are the Laplace transform results of the wind rotor speed and wind speed respectively, and τ is the time constant of the system.

[0032] A first-order inertial system is a filter system with a cutoff frequency f c (f c =1 / 2πτ) will be filtered out and will not have any effect on the change in the wind rotor speed. Filtering is performed when conducting relevant performance analysis of wind turbines based on wind speed data. There are two methods to determine the time constant of the wind rotor inertial system, one is based on the model-driven time constant determination method, and the other is the data-driven time constant determination method. The model-driven time constant determination method is based on a simplified dynamic analysis model. The actual rotational inertia of the wind rotor will bring calculation difficulties due to the flexibility of the blades. Therefore, the present invention proposes a data-driven time constant determination method.

[0033] In the SCADA data of the wind farm, data such as wind speed, rotation speed, and power are recorded synchronously. From different perspectives, there are different ways to divide the operating conditions of wind turbines. From the perspective of the "speed-power" relationship, the operating conditions of wind turbines are divided into four stages. The first stage is transition zone 1, which is the initial stage when the wind speed reaches the cut-in wind speed and the wind turbine starts to operate. The second stage is the maximum wind power tracking zone MPPT. This area maintains the tip speed ratio at the optimal value to obtain the maximum power coefficient. The third stage is the second transition zone, which enters the constant speed zone after this area. In the constant speed zone, even if the wind speed increases, the rotor speed remains basically unchanged at the set value, and the utilization rate of wind energy is adjusted by controlling the change in the pitch angle.

[0034] Based on the analysis of each stage of the operating conditions, it can be seen that in the maximum wind energy tracking area MPPT, the tip speed ratio λ is basically unchanged and is defined as

[0035]

[0036] Where R is the radius of the wind rotor swept area.

[0037] Let V = {v t1 ,v t2 ,v t3 ,···,v tn ,v t(n+1) ,v t(n+2) ,···} is a section of wind speed data in the maximum wind energy tracking area MPPT, the sampling time interval is 1 second, and the time constant τ=n, n is an integer, then:

[0038]

[0039] The mean filtered array is obtained from Start taking values, the new array becomes The array is represented by V = {v1,v2,v3,···,v m}, where m is the size of the array after sliding mean filtering.

[0040] Substitute the filtered array into equation (10) to calculate the tip speed ratio, and we have

[0041] λ={λ1,λ2,λ3,···,λ m} (12)

[0042] In the formula, And so on.

[0043] The array λ obtained when the filtering time constant is n is expressed as λ n , changing the value of n to obtain multiple different arrays, for a λ n , the standard deviation is:

[0044]

[0045] In the formula, is the mean.

[0046] Taking multiple different time constants, we can get a standard deviation array of the tip speed ratio, that is,

[0047] σ={σ1,σ2,σ3,···,σ q} (14)

[0048] Wherein, the subscript q is the number of times the time constant takes value.

[0049] Based on the control principle of wind turbines, in the maximum wind energy tracking area, the tip speed ratio should be a constant value. Under the most ideal time constant value, the calculated standard deviation should be the minimum, thus finding min(σ1,σ2,σ3,···,σ q ) is used as the time constant of the wind turbine, that is,

[0050]

[0051] Furthermore, the implementation process of step 3 is as follows: in order to more reliably analyze the aerodynamic characteristics of the wind rotor during its service life, the power coefficient of the wind rotor within comparable working conditions should be selected. Among the multiple operating conditions of the wind turbine, the maximum wind energy tracking condition is characterized by keeping the power coefficient unchanged, and is therefore the most suitable calculation condition. According to the speed-power curve of the wind turbine, the key to determining the calculation condition is to clarify the minimum speed and maximum speed of the maximum wind energy tracking area. The speed-power curve is selected as the basis for judgment because these two parameters are stable and slow-changing relative to the wind speed. Taking the maximum wind energy tracking area as the calculation condition area, two problems need to be solved: 1) Single-value processing of the speed-power scatter plot; 2) Clarifying the minimum speed and maximum speed of the maximum wind energy tracking area.

[0052] Problem 1): Use the data binning method to discretize the speed into several equally spaced data intervals, that is, N = {N1(n1-Δn,n1+Δn),···,N m (n m -Δn,n m +Δn)};

[0053] For any small interval N n (n n -Δn,n n +Δn), and use the kernel density estimation method (KDE) to extract the power point with the highest probability, and use this point as the only power value in the interval. Let the power point in this interval be P n1 , P n2, P n3 ,···,P nq The kernel density estimate of the population density function f(P) at any point P is

[0054]

[0055] Where K() is the kernel density function, q is the number of sample points, and h is the window width.

[0056] The probability density set of all points in this interval is defined as Then there is

[0057] Finally, the final value of the interval is:

[0058]

[0059] Problem 2): Use the "conservative" data comparison method. First, select a conservative speed interval [ω1, ω2] within the maximum wind energy tracking area. That is, based on the operating characteristics of the wind turbine, the speed in this interval must be within the maximum wind energy tracking area. Use the power-speed expression (18) in the maximum wind energy tracking area for data fitting.

[0060]

[0061] In the formula, λ opt is the optimal tip speed ratio, C Pmax For the best power factor.

[0062] The above formula is abbreviated as

[0063]

[0064] Where k is the proportionality coefficient.

[0065] In the speed range [ω1,ω2], for a speed ω i , there is a corresponding power P i , denoted as (ω i ,P i ). Assume that the number of discrete points in the interval is n, and use the least squares method to find the value of parameter k, that is:

[0066]

[0067] Then, let (ω c ,P c ) is a point outside the speed range [ω1,ω2], and the fitting curve P=k is obtained by substituting opt ω 3 , calculate the difference ΔP as:

[0068]

[0069] Set a critical value. When ΔP is greater than this critical value, it is considered that the point (ω c ,P c ) is no longer in the maximum wind energy tracking area. The value at the time of analysis gradually moves outward from the interval [ω1, ω2] port to both sides. When calculating, the critical value is selected from the fitting curve P = k opt ω 3 exist The maximum difference ΔP within the interval max , the calculation method is the same as formula (21), and the speed range of the wind turbine maximum wind energy tracking area is obtained as [ω c1 ,ω c2 ]. At this point, the division of the calculation condition area is completed.

[0070] Furthermore, the implementation process of step 4 is as follows: the power coefficient based on instantaneous power and cumulative power is calculated according to formula (6) and formula (22) respectively, the incoming wind speed is calculated according to formula (8), and the wind speed filtering and power accumulation integration are performed according to the time constant τ (40s).

[0071]

[0072] In the formula, τ=t2-t1.

[0073] Define the evaluation index C PI That is, the real-time value of the interval power coefficient calculated based on the instantaneous power and C PA That is, based on the real-time value of the interval power coefficient calculated based on the cumulative power, a "four-index" multi-angle observation algorithm is proposed, that is, the mean value of the interval power coefficient calculated based on the instantaneous power and the maximum value C PmaxI , the mean of the interval power coefficient calculated based on the cumulative power and the maximum value C PmaxA Four evaluation indicators are used to observe the on-site historical evolution of aerodynamic performance of large wind turbines.

[0074] Compared with the prior art, the present invention has the following technical advantages:

[0075] 1. The power coefficient determined based on field data proposed in the present invention is used as a basic indicator for on-site analysis of the aerodynamic performance of wind turbine blades, providing a new evaluation index and method for on-site historical evolution behavior analysis of the aerodynamic performance of large wind turbines;

[0076] 2. The data compensation and data filtering adopted by the present invention overcome the influence of two unfavorable factors of "bias" and "randomness" of SCADA wind speed data;

[0077] 3. The present invention regards the relationship between wind speed and wind rotor speed as an inertial system and proposes a data-driven time constant determination method, which solves the problem of difficulty in calculating the actual moment of inertia of the wind rotor due to blade flexibility in the model-driven time constant determination method;

[0078] 4. The present invention designs a calculation working condition area division algorithm that conforms to the actual situation, and adopts the data binning method and the "conservative" data comparison method to respectively solve the problems of "single-valued" scattered data and determination of interval critical value that need to be solved when dividing working conditions;

[0079] 5. The present invention calculates the power coefficient based on instantaneous power and cumulative power, and proposes a four-index multi-angle observation algorithm to analyze the on-site historical evolution behavior of aerodynamic performance, improve the accuracy of the historical evolution behavior of large units, reduce the maintenance cost of wind turbines, and ensure the safety and economy of wind turbine operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 A diagram depicting the blade's on-site aerodynamic performance.

[0081] Figure 2 Wind speed diagram before and after the wind rotor.

[0082] Figure 3 Wind turbine operating conditions classification diagram.

[0083] Figure 4 Wind speed sliding mean filter graph.

[0084] Figure 5 Calculation condition area division diagram.

[0085] Figure 6 Data-driven time constant analysis plot.

[0086] Figure 7 Analysis diagram of the evolution of aerodynamic performance of the unit. DETAILED DESCRIPTION

[0087] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0088] A method for analyzing the on-site historical evolution behavior of aerodynamic performance of a large wind turbine generator set, the method comprising:

[0089] Step 1: Determine the basic evaluation index of the on-site historical evolution behavior of the aerodynamic performance of the wind turbine; based on the blade element theory and on-site aerodynamic performance analysis, determine the power coefficient as the basic index for on-site analysis of the aerodynamic performance of the wind turbine blades.

[0090] Step 2, preprocessing of on-site wind speed data; considering the flow field characteristics of on-site wind speed data, data compensation is used to overcome the "bias" of SCADA wind speed data, the system time is determined from the preprocessed SCADA wind speed data itself, and a data-driven time constant determination method is proposed to determine a reasonable filtering time constant to overcome the "randomness" of SCADA wind speed data.

[0091] Step 3, calculation condition area division; design a calculation condition area division algorithm that conforms to the actual characteristics of the blade field data, and use the data binning method and the "conservative" data comparison method to solve the problems of "single-valued" scattered data and interval critical value determination that need to be solved when dividing the working conditions.

[0092] Step 4: Calculate the power coefficient based on the field data; calculate the power coefficient based on the instantaneous power and cumulative power, and propose a "four-index" multi-angle observation algorithm, that is, the mean of the interval power coefficient calculated based on the instantaneous power and the maximum value C PmaxI , the mean of the interval power coefficient calculated based on the cumulative power and the maximum value C PmaxA Four evaluation indicators are used to observe the on-site historical evolution of aerodynamic performance of large wind turbines.

[0093] Furthermore, the description of the on-site aerodynamic performance of the blades in step 1 is as follows: the blades are the energy capture components of the wind turbine, converting the kinetic energy of the air into the mechanical energy of the wind rotor. At the same time, the blades are also the main load-bearing components. The blades have the ability to capture aerodynamic energy because they adopt an airfoil structure design, and the airflow acting on the blades will generate lift and drag. Based on lift and drag, the characteristics of various forces such as thrust and torque can be derived. These characteristics can all be attributed to the aerodynamic performance of the blades. The aerodynamic performance of the blades determines the energy capture efficiency and load characteristics of the wind turbine, and is a key performance indicator of the unit. In the design and analysis stage, the aerodynamic performance analysis of the blades can adopt two-dimensional aerodynamics or three-dimensional aerodynamics, and the calculation methods used include blade element-momentum theory (BEM), computational fluid dynamics (CFD), etc.

[0094] like Figure 1 As shown in the figure, the lift and drag acting on the blade element (the wind rotor blade is divided into several micro segments along the span direction, each micro segment is called a blade element) are written as dL and dD respectively. According to the principle of force composition and decomposition, they can also be written as normal force and tangential force dF n 、dF t The torque acting on a blade ring of the wind wheel can be written as

[0095]

[0096] Where B is the number of wind wheel blades, ρ is the air density, v0 is the synthetic air velocity, Ct is the tangential force coefficient, and r is the radius of the circle where the current blade element is located.

[0097] The axial force acting on a blade ring of the wind wheel can be written as

[0098]

[0099] In the formula, C n is the normal force coefficient.

[0100] Assuming the rotor radius is R and the rotor rotation speed is ω, according to equations (1) and (2), the torque M, axial force (thrust) T and rotor shaft power P acting on the entire rotor are: rt It can be written as

[0101]

[0102]

[0103] P rt =Mω (5)

[0104] The wind rotor power coefficient is

[0105]

[0106] From the above analysis, we can see that the description of blade aerodynamic performance involves multiple aspects such as lift characteristics, drag characteristics, thrust characteristics, torque characteristics, etc. These factors are all matters that need to be fully considered during the blade design stage. However, in wind farms, there is very little that can be directly observed on site, and the description of blade aerodynamic performance can only be mined and analyzed from SCADA data.

[0107] In SCADA data, information such as blade lift and drag cannot be collected directly. Mature sensor test systems include measurements of cabin wind speed, generator power, and wind rotor speed. According to formula (6), the wind rotor power coefficient can be calculated through power and wind speed data plus air density information. Although air density is generally not measured directly, it can be supplemented by calculation based on wind farm temperature, humidity and other information. Therefore, in wind farms, the power coefficient is a basic indicator for analyzing the aerodynamic performance of wind rotor blades.

[0108] In SCADA data, generator power and wind rotor speed data are relatively accurate and reliable. This is because neither power nor speed can produce sudden changes, and the corresponding measurement methods are relatively mature. However, wind speed data needs to be effectively preprocessed (data compensation, filtering, etc.) before it can be used for relevant analysis and calculation.

[0109] Furthermore, the implementation process of step 2 is as follows;

[0110] Step 2.1 Wind speed data compensation;

[0111] The wind speed of a wind farm is affected by the atmospheric environment and local terrain, and its typical characteristics are randomness and time-varying. In the SCADA system, wind speed data generally comes from wind speed sensors installed on the nacelle. At this time, there is a certain deviation between the wind speed data and the actual incoming wind speed. On the one hand, this is because the kinetic energy of the air is partially converted into mechanical energy of the wind rotor when the airflow passes through the wind rotor, and on the other hand, the airflow is disturbed by the rotation of the blades when passing through the wind rotor.

[0112] exist Figure 2 In the example, the wind speed before the wind wheel is set to v1, and the wind speed when passing through the wind wheel is v d , the wind speed after passing through the wind wheel is set to v2, obviously v1>v2, under relatively stable wind speed conditions, the relationship between the three can be written as

[0113]

[0114] Where P is the generator power and S is the swept area of ​​the wind rotor.

[0115] According to formula (7), the relationship between v1 and v2 can be written as:

[0116]

[0117] Step 2.2 Wind speed data filtering: The relationship between wind speed and wind rotor speed is approximately described by a first-order inertial system, expressed as:

[0118]

[0119] Where W(s) and V(s) are the Laplace transform results of the wind rotor speed and wind speed respectively, and τ is the time constant of the system.

[0120] A first-order inertial system is a filter system with a cutoff frequency f c (f c =1 / 2πτ) will be filtered out and will not have any effect on the change in the wind rotor speed. Filtering is performed when conducting relevant performance analysis of wind turbines based on wind speed data. There are two methods to determine the time constant of the wind rotor inertial system, one is based on the model-driven time constant determination method, and the other is the data-driven time constant determination method. The model-driven time constant determination method is based on a simplified dynamic analysis model. The actual rotational inertia of the wind rotor will bring calculation difficulties due to the flexibility of the blades. Therefore, the present invention proposes a data-driven time constant determination method.

[0121] In the SCADA data of the wind farm, data such as wind speed, rotation speed, and power are recorded synchronously. From different perspectives, there are different ways to divide the operating conditions of wind turbines. From the perspective of the "speed-power" relationship, the operating conditions of wind turbines are divided into four stages. The first stage is transition zone 1, which is the initial stage when the wind speed reaches the cut-in wind speed and the wind turbine starts to operate. The second stage is the maximum wind power tracking zone MPPT. This area maintains the tip speed ratio at the optimal value to obtain the maximum power coefficient. The third stage is the second transition zone, which enters the constant speed zone after this area. In the constant speed zone, even if the wind speed increases, the rotor speed remains basically unchanged at the set value, and the utilization rate of wind energy is adjusted by controlling the change in the pitch angle.

[0122] Based on the above analysis, it can be seen that in the maximum wind energy tracking area MPPT, the tip speed ratio is basically unchanged, which is defined as

[0123]

[0124] Where R is the radius of the wind rotor swept area.

[0125] Let V = {v t1 ,v t2 ,v t3 ,···,v tn ,v t(n+1) ,v t(n+2) ,···} is a section of wind speed data in the maximum wind energy tracking area MPPT, the sampling time interval is 1 second, and the time constant τ=n, n is an integer, then:

[0126]

[0127] The mean filtered array is obtained from Start taking values, the new array becomes The array is represented by V = {v1,v2,v3,···,v m}, where m is the size of the array after sliding mean filtering.

[0128] Substitute the filtered array into equation (10) to calculate the tip speed ratio, and we have

[0129] λ={λ1,λ2,λ3,···,λ m} (12)

[0130] In the formula, And so on.

[0131] The array λ obtained when the filtering time constant is n is expressed as λ n , changing the value of n to obtain multiple different arrays, for a λ n , the standard deviation is:

[0132]

[0133] In the formula, is the mean.

[0134] Taking multiple different time constants, we can get a standard deviation array of the tip speed ratio, that is,

[0135] σ={σ1,σ2,σ3,···,σ q} (14)

[0136] Wherein, the subscript q is the number of times the time constant takes value.

[0137] Based on the control principle of wind turbines, in the maximum wind energy tracking area, the tip speed ratio should be a constant value. Under the most ideal time constant value, the calculated standard deviation should be the minimum, thus finding min(σ1,σ2,σ3,···,σ q ) is used as the time constant of the wind turbine, that is,

[0138]

[0139] Furthermore, the implementation process of step 3 is as follows: according to the speed-power curve of the wind turbine, the key to determining the calculation condition is to clarify the minimum speed and maximum speed of the maximum wind energy tracking area. The speed-power curve is selected as the basis for judgment because these two parameters are stable and slow-changing relative to the wind speed. Taking the maximum wind energy tracking area as the calculation condition area, two problems need to be solved: 1) single-value processing of the speed-power scatter plot; 2) clarifying the minimum speed and maximum speed of the maximum wind energy tracking area.

[0140] Problem 1): Use the data binning method to discretize the speed into several equally spaced data intervals, that is, N = {N1(n1-Δn,n1+Δn),···,N m (n m -Δn,n m +Δn)};

[0141] For any small interval N n (n n -Δn,n n +Δn), and use the kernel density estimation method (KDE) to extract the power point with the highest probability, and use this point as the only power value in the interval. Let the power point in this interval be P n1 , P n2 , P n3 ,···,P nq The kernel density estimate of the population density function f(P) at any point P is

[0142]

[0143] Where K() is the kernel density function, q is the number of sample points, and h is the window width.

[0144] The probability density set of all points in this interval is defined as Then there is

[0145] Finally, the final value of the interval is:

[0146]

[0147] Problem 2): Use the "conservative" data comparison method. First, select a conservative speed interval [ω1, ω2] within the maximum wind energy tracking area. That is, based on the operating characteristics of the wind turbine, the speed in this interval must be within the maximum wind energy tracking area. Use the power-speed expression (18) in the maximum wind energy tracking area for data fitting.

[0148]

[0149] In the formula, λ opt is the optimal tip speed ratio, C Pmax For the best power factor.

[0150] The above formula is abbreviated as

[0151] P=kω 3 (19)

[0152] Where k is the proportionality coefficient.

[0153] In the speed range [ω1,ω2], for a speed ω i , there is a corresponding power P i , denoted as (ω i ,P i ). Assume that the number of discrete points in the interval is n, and use the least squares method to find the value of parameter k, that is:

[0154]

[0155] Then, let (ω c ,P c ) is a point outside the speed range [ω1,ω2], and the obtained fitting curve P = k opt ω 3 , calculate the difference ΔP as:

[0156]

[0157] Set a critical value. When ΔP is greater than this critical value, it is considered that the point (ω c ,P c) is no longer in the maximum wind energy tracking area. The value at the time of analysis gradually moves outward from the interval [ω1, ω2] port to both sides. When calculating, the critical value is selected from the fitting curve P = k opt ω 3 exist The maximum difference ΔP within the interval max , the calculation method is the same as formula (21), and the speed range of the wind turbine maximum wind energy tracking area is obtained as [ω c1 ,ω c2 ]. At this point, the division of the calculation condition area is completed.

[0158] Furthermore, the implementation process of step 4 is as follows: the power coefficient based on instantaneous power and cumulative power is calculated according to formula (6) and formula (22) respectively, the incoming wind speed is calculated according to formula (8), and the wind speed filtering and power accumulation integration are performed according to the time constant τ (40s).

[0159]

[0160] In the formula, τ=t2-t1.

[0161] Define the evaluation index C PI (Real-time value of interval power coefficient calculated based on instantaneous power) and C PA (Real-time value of interval power coefficient calculated based on cumulative power) and propose a "four-index" multi-angle observation algorithm, that is, the mean value of interval power coefficient calculated based on instantaneous power and the maximum value C PmaxI , the mean of the interval power coefficient calculated based on the cumulative power and the maximum value C PmaxA Four evaluation indicators are used to observe the on-site historical evolution of the aerodynamic performance of large wind turbines, including The power coefficient C of each sampling point in the interval is calculated according to formula (1) PI The average value, C PmaxI is the power coefficient C of each sampling point in the interval PI The maximum value in The interval power coefficient C is calculated based on the time constant τ (40s) according to formula (22) PA The average value, C PmaxA is the calculated power coefficient C of all intervals PA The maximum value in .

[0162] Examples and data analysis

[0163] A 2.1MW doubly-fed wind turbine in a mountain wind farm in the south is selected as an example for analysis. The rotor diameter of this type of wind turbine is 110m, the cut-in wind speed is 3m / s, the rated wind speed is 9.5m / s, and the cut-out wind speed is 20m / s; the power regulation method is variable speed and variable pitch; the design life is 20 years. The analysis mainly includes two aspects: data-driven time constant analysis and historical evolution analysis of the unit's aerodynamic performance.

[0164] First, data-driven time constant analysis;

[0165] Extract the parameters such as instantaneous wind speed, active power, hub speed, ambient temperature, etc. from the SCADA data of a unit. According to the working area division method proposed by equations (16) to (21), it can be known that the speed range of the maximum wind energy tracking area is [9.4r / min, 12.8r / min]. It should be noted here that the temperature data in the wind farm SCADA system is collected separately (sampling period 60s) and is minute-level data. In order to make the temperature data sequence consistent with other parameter sequences, the second-level temperature data in the same period is replaced by minute-level data. Since the ambient temperature is a slow variable, this treatment is acceptable. Then, a set of continuous data (data length is 3000 points) in the maximum wind energy tracking area (MPPT) is selected at a sampling interval of 1s, and the time constant τ is taken as 8s, 10s, ..., 18s, 20s, 30s, 40s, ..., 100s respectively.

[0166] Figure 6 Correlation curves were obtained for data-driven time constant analysis. Figure 6 (a) shows the instantaneous wind speed after data compensation according to formula (8), and the wind speed curve after sliding mean filtering with time constants of 20s, 40s and 60s respectively. From the curve characteristics in the figure, the larger the time constant is, the smoother the wind speed after filtering is. Obviously, the larger the time constant value is, the better. According to formula (10), the value of the time constant will affect the calculation result of the tip speed ratio. Figure 6 (b) The tip speed ratio fluctuation curves under the above three time constant values ​​are given. It can be clearly seen from the figure that the fluctuations of the tip speed ratio after filtering are generally small. From the basic idea of ​​maximum wind energy tracking control, the tip speed ratio λ in this area is ideally a straight line. However, from the actual curve, no matter whether the time constant is large or small, the tip speed ratio in a continuous time area is in a fluctuating state. This shows that due to the fluctuating behavior of wind speed, even in the MPPT area, wind turbines cannot always be in the maximum wind energy tracking state, but can only be infinitely approximated. From the curve in the figure, when the time constant is 40s and 60s, the tip speed ratio curve obtained is obviously less volatile than when it is 20s. However, the tip speed ratio fluctuation curves under the time constant values ​​of 40s and 60s are difficult to compare. For this reason, it is necessary to Figure 6(c) is used for analysis. The standard deviation is the arithmetic square root of the variance and can better reflect the degree of discreteness of a data set. According to the idea of ​​infinite approximation, the smaller the standard deviation within the time interval, the better the MPPT control effect. Conversely, it can be used to evaluate the correctness of the time constant value. The calculation method is shown in equations (13) and (15). It is clearly found in the figure that with the change of the time constant value, the standard deviation curve first decreases and then increases, which further proves that the time constant value is not the larger the better, nor the smaller the better. Here, the time constant corresponding to the minimum standard deviation is 40s.

[0167] Table 1 gives the interval average wind speed under multiple time constant values ​​in detail. (after filtering), average tip speed ratio and standard deviation σ. From the data characteristics in the table, as the time constant increases, the average wind speed gradually decreases; while the average tip speed ratio The standard deviation first decreases and then increases, reaching a minimum value of 0.235 at 40s. Therefore, it can be considered that 40s is a more appropriate time constant for the wind turbine rotor.

[0168] Table 1 Average wind speed (after filtering), average tip speed ratio and standard deviation

[0169]

[0170] Second, the historical evolution analysis of the aerodynamic performance of the unit;

[0171] In order to obtain the historical evolution characteristics of the on-site aerodynamic performance of wind turbines, the SCADA data of two wind turbines in the wind farm in March of each year from 2017 to 2021 (for 5 consecutive years) were selected to extract parameters such as instantaneous wind speed, active power, hub speed, and ambient temperature. During the specific extraction, 2000 continuous data points were selected at a sampling interval of 1s in the maximum wind power tracking area (MPPT). Unit 1 extracted data based on the speed, and unit 2 extracted data based on the wind speed and speed verification. According to formula (6), the power coefficient (C P ) is related to active power, air density, incoming wind speed, and rotor swept area. Since air density does not appear directly in SCADA data, this paper uses the method in the literature to calculate the air density at the sampling time, calculates the incoming wind speed according to formula (8), and performs wind speed filtering and power accumulation integration according to the time constant τ (40s). Then, the power coefficient based on instantaneous power and cumulative power is calculated respectively, among which the power coefficient based on instantaneous power is calculated according to formula (6), and the power coefficient based on cumulative power is calculated according to formula (22).

[0172] Figure 7These are the relevant curves obtained from the aerodynamic performance evolution analysis of the two units. Figure 7 (a) and Figure 7 (b) The MPPT regional power coefficient C calculated based on instantaneous power for units 1 and 2 in different years is given respectively. P Curve. The starting point of the horizontal axis in the figure is marked as 0s, which only indicates the relative value. Judging from the change characteristics of the curve in the figure, the power coefficient fluctuates mainly in the range of 0.25 to 0.45, and the fluctuation amplitude varies greatly in different years. Among them, the maximum power coefficient of unit 1 is 0.476, which occurred in 2019, and the maximum power coefficient of unit 2 is 0.513, which occurred in 2017. It can be seen from the curve that the maximum power coefficient is not a steady-state value, but a transient value, which is mainly caused by the violent fluctuation of wind speed during this period. Figure 7 (c) and Figure 7 (d) The MPPT regional power coefficient C calculated based on the cumulative power of the two units in different years is given respectively. P Curve. Compared to Figure 7 (a) and Figure 7 (b), calculated power coefficient C P The fluctuation amplitude is generally smaller. The curve change characteristics in the figure also show that although the wind turbine is operating in the MPPT area, the power coefficient is not a constant value, but fluctuates within a certain range. This once again proves that even in the MPPT area, the wind turbine cannot always be in the maximum wind energy tracking state, but can only be infinitely close.

[0173] In order to better analyze the historical evolution characteristics of the unit's aerodynamic performance, Figure 7 (e) and Figure 7 (f) shows the power coefficient interval mean values ​​calculated based on the data samples of the two units in different years. and power coefficient interval maximum value C Pmax Table 2 and Table 3 give the specific data of the two units respectively. The expression of annual degradation degree of aerodynamic performance Γ is defined as

[0174]

[0175] In the formula, C PB is the power coefficient of the base year, for example, 2017 is used in this example; C PC is the power factor for the current year, which can be any year following the base year.

[0176] Definition: C PI Indicates the real-time value of the interval power coefficient calculated based on the instantaneous power. represents the mean value of the interval power coefficient calculated based on the instantaneous power, C PmaxI Indicates the maximum value of the interval power coefficient calculated based on the instantaneous power; CPA Indicates the real-time value of the interval power coefficient calculated based on the cumulative power. represents the mean value of the interval power coefficient calculated based on the cumulative power, C PmaxA Indicates the maximum value of the interval power factor calculated based on the cumulative power.

[0177] from Figure 7 (e) and Table 2 show that except for the C PmaxI In addition to showing different characteristics, the four indicators C PmaxI , and C PmaxA All show an overall downward trend with the increase of service years. Compared with 2017, the degradation degrees (Γ) calculated by the four indicators in 2021 are 20.1%, 24.9%, 19.6%, and 21.9%, respectively. By observing these four indicators and combining them with wind turbine design theory and experience, the indicator C PmaxA It can better reflect the actual performance of wind turbines. PmaxA is 0.439, and in 2021, C PmaxA The value of dropped to 0.343. From the above analysis, it can be seen that with the increase of service life, the aerodynamic performance of wind turbines tends to deteriorate significantly.

[0178] from Figure 7 (f) and Table 3 show that, similar to Unit 1, Unit 2 exhibits different data characteristics in 2019. C PmaxI and All three indicators show an overall downward trend with increasing service years. PmaxAThis indicator showed different data characteristics in both 2019 and 2021. Even so, the overall trend of change is still "downward". Compared with 2017, the degradation degree (Γ) of unit 2 calculated from four indicators in 2021 was 13.8%, 22.8%, 13.9%, and 16.8%, respectively. From the above analysis, it can be seen that for both unit 1 and unit 2, with the increase of service life, the aerodynamic performance of wind turbines has a significant degradation trend. However, the calculated decline of the two units is different. Due to space limitations, only limited calculation results are given in this article. During the analysis process, the author's team carried out more data analysis and more groups of calculation results. From the perspective of wind turbine operation theory and data analysis, there are three main factors that affect the above calculation results: first, the wind turbines themselves are different. Although they are of the same model, the personalized influencing factors cannot be ignored; second, the influence of SCADA data samples. The randomness of on-site wind speed is the main factor affecting the analysis of the above indicators. Although a new data processing method is proposed in this paper to eliminate the influence of random factors as much as possible, it is impossible to completely eliminate it; third, the influence of data extraction methods. As mentioned earlier, data extraction is performed from different angles to better analyze the effectiveness of this method. Different processing methods will inevitably have certain influences.

[0179] In summary, the proposed method for analyzing the on-site historical evolution behavior of aerodynamic performance of large wind turbines is effective, and the aerodynamic performance evolution characteristics of wind turbines are discovered. The proposed "four-index" observation method can more reliably evaluate the evolution characteristics. The data analysis is consistent with the theoretical analysis and provides new ideas and basis for quantitative evaluation.

[0180] Table 2 Average power factor and maximum power factor (unit 1)

[0181]

[0182]

[0183] Table 3 Average power factor and maximum power factor (unit 2)

[0184]

[0185] Whether it is an onshore wind turbine or an offshore wind turbine, as the service life increases, the performance degradation caused by the natural aging of the unit is inevitable. This paper proposes a method for analyzing the on-site historical evolution behavior of the aerodynamic performance of large wind turbines based on on-site SCADA data. This method uses the power coefficient as the basic indicator for on-site analysis of the aerodynamic performance of wind turbine blades, fully considers the flow field characteristics of on-site wind speed data, and uses data compensation and data filtering to overcome the influence of the two unfavorable factors of "randomness" and "bias" of SCADA wind speed data.

[0186] The relationship between wind speed and rotor speed is regarded as an inertial system, and a data-driven time constant determination method is proposed. The system time constant is determined from the data itself, overcoming the inconvenience of the model-driven time constant determination method, which requires precise unit parameters and does not need to consider the influence of blade flexibility.

[0187] An algorithm for dividing the calculation working condition area that conforms to the actual characteristics of the field data is designed. The speed-power curve is selected as the basis for judgment, making full use of the stability and slow-changing advantages of these two parameters relative to wind speed. The data binning method and the "conservative" data comparison method are used to solve the "single-value" problem of scattered data and the problem of determining the interval critical value that need to be solved when dividing the calculation working condition area.

[0188] The effectiveness of the "four-index" multi-angle observation algorithm is proposed. The calculation case analysis shows that the proposed method is reliable and effective, and the evolution characteristics of the aerodynamic performance of large wind turbines on site are discovered. The analysis results are consistent with the theoretical analysis and provide new ideas and basis for quantitative evaluation. This provides important knowledge support for the operation, maintenance and performance improvement of large wind turbines after they have served to their "middle age".

Claims

1. A method for analyzing the on-site historical evolution behavior of aerodynamic performance of large wind turbines, characterized in that: The method includes: Step 1, determine the basic evaluation index of the on-site historical evolution behavior of the aerodynamic performance of the wind turbine; according to the blade element theory and on-site aerodynamic performance analysis, determine the power coefficient as the basic evaluation index of the on-site analysis of the aerodynamic performance of the wind turbine blade, and provide an evaluation index for the on-site historical evolution behavior analysis of the aerodynamic performance of the wind turbine in steps 2, 3, and 4; Step 2, preprocessing of on-site wind speed data; considering the flow field characteristics of the on-site wind speed data, preprocessing the collected SCADA wind speed data, using data compensation to overcome the deviation of the SCADA wind speed data, determining the system time from the preprocessed SCADA wind speed data itself, proposing a data-driven time constant determination method, determining a reasonable filtering time constant, overcoming the randomness of the SCADA wind speed data, and providing more accurate wind speed data for calculating the power coefficient in step 4; Step 3, calculation of working condition area division; design a calculation working condition area division algorithm that conforms to the actual characteristics of blade field data, and use data binning method and "conservative" data comparison method to solve the problems of "single value" of scattered data and interval critical value determination that need to be solved when dividing working conditions, so as to provide accurate maximum wind energy tracking working condition area MPPT for step 4 power coefficient calculation; Step 4: Calculate the power coefficient based on the field data; calculate the power coefficient based on the instantaneous power and cumulative power, and propose a four-index multi-angle observation algorithm, that is, the interval power coefficient mean value obtained based on the instantaneous power calculation and the maximum value C PmaxI , the mean of the interval power coefficient calculated based on the cumulative power and the maximum value C PmaxA Four evaluation indicators are used to observe the on-site historical evolution of aerodynamic performance of large wind turbines.

2. The method for analyzing the on-site historical evolution behavior of aerodynamic performance of a large wind turbine according to claim 1 is characterized in that: The implementation process of step 1 is as follows: the lift and drag acting on the blade element are written as dL and dD respectively, and according to the principle of force composition and decomposition, they are written as normal force and tangential force dF n 、dF t ; The torque dM acting on a blade ring of the wind wheel is: Where B is the number of wind wheel blades, ρ is the air density, v0 is the synthetic air velocity, C t is the tangential force coefficient, r is the radius of the ring where the current blade element is located; The axial force dT acting on a blade ring of the wind wheel can be written as: Where c is the chord length of the blade section, C n is the normal force coefficient; Assume that the rotor radius is R and the rotor rotation speed is ω. According to equations (1) and (2), the torque M, axial force T and rotor shaft power P acting on the entire rotor are: rt for: P rt =Mω (5) Wind rotor power coefficient C p for According to formula (6), the wind rotor power coefficient is calculated by using power and wind speed data and air density information. In a wind farm, the power coefficient is a basic indicator for analyzing the aerodynamic performance of wind rotor blades.

3. A method for analyzing the on-site historical evolution behavior of aerodynamic performance of a large wind turbine according to claim 2, characterized in that: The implementation process of step 2 is as follows: Step 2.1 Wind speed data compensation; In the SCADA system, the wind speed data comes from the wind speed sensor installed on the nacelle; the wind speed before the wind rotor is set to v1, and the wind speed when passing through the wind rotor is v d , the wind speed after passing through the wind wheel is set to v2, v1>v2, under relatively stable wind speed conditions: In the formula, P is the generator power, S is the wind rotor swept area; According to formula (7), the relationship between v1 and v2 is: Step 2.2 Wind speed data filtering: The relationship between wind speed and wind rotor speed is approximately described by a first-order inertial system, expressed as: Where W(s) and V(s) are the Laplace transform results of the wind rotor speed and wind speed respectively, and τ is the time constant of the system; A first-order inertial system is a filter system with a cutoff frequency f c (f c =1 / 2πτ) will be filtered out and will not have any effect on the change of wind rotor speed; filtering is performed when conducting relevant performance analysis of wind turbines based on wind speed data; the time constant of the wind rotor inertial system is determined by using the data-driven time constant determination method; In the SCADA data of wind farms, wind speed, rotation speed and power data are recorded synchronously. From different perspectives, there are different ways to divide the operating conditions of wind turbines. From the perspective of the "rotation speed-power" relationship, the operating conditions of wind turbines are divided into four stages. The first stage is transition zone 1, when the wind speed reaches the cut-in wind speed and the wind turbine starts to operate in the initial stage. The second stage is the maximum wind energy tracking area MPPT, where the tip speed ratio is maintained at the optimal value to obtain the maximum power coefficient; the third stage is the second transition area, after which the constant speed area is entered; in the constant speed area, even if the wind speed increases, the rotor speed remains basically unchanged at the set value, and the utilization rate of wind energy is adjusted by controlling the change of the pitch angle; Based on the analysis of each stage of the operating conditions, it can be seen that in the maximum wind energy tracking area MPPT, the tip speed ratio λ is basically unchanged and is defined as Where R is the radius of the wind wheel swept area; Let V = {v t1 ,v t2 ,v t3 ,···,v tn ,v t(n+1) ,v t(n+2) ,···} is a section of wind speed data in the maximum wind energy tracking area MPPT, the sampling time interval is 1 second, and the time constant τ=n, n is an integer, then: The mean filtered array is obtained from Start taking values, the new array becomes The array is represented by V = {v1,v2,v3,···,v m }, where m is the size of the array after the sliding mean filter processing; Substitute the filtered array into equation (10) to calculate the tip speed ratio, and we have λ={λ1,λ2,λ3,···,λ m } (12) In the formula, And so on; The array λ obtained when the filtering time constant is n is expressed as λ n , changing the value of n to obtain multiple different arrays, for a λ n , the standard deviation is: In the formula, is the mean; Taking multiple different time constants, we can get a standard deviation array of the tip speed ratio, that is, σ={σ1,σ2,σ3,···,σ q } (14) In the formula, the subscript q is the number of times the time constant takes value; Based on the control principle of wind turbines, in the maximum wind energy tracking area, the tip speed ratio should be a constant value. Under the most ideal time constant value, the calculated standard deviation should be the minimum, thus finding min(σ1,σ2,σ3,···,σ q ) is used as the time constant of the wind turbine, that is, 4. The method for analyzing the on-site historical evolution behavior of aerodynamic performance of a large wind turbine according to claim 3 is characterized in that: The implementation process of step 3 is as follows: selecting the wind rotor power coefficient within the comparable working condition; among the multiple operating conditions of the wind turbine generator set, the characteristic of the maximum wind energy tracking condition is to keep the power coefficient unchanged; according to the speed-power curve of the wind turbine generator set, the key to determining the calculation condition is to clarify the minimum speed and the maximum speed of the maximum wind energy tracking area; the speed-power curve is selected as the judgment basis because these two parameters are stable and slow-changing relative to the wind speed; Taking the maximum wind energy tracking area as the calculation condition area, two problems need to be solved: 1) converting the speed-power scatter plot into single value; 2) Clarify the minimum and maximum speeds in the maximum wind energy tracking area; Problem 1): Use data binning method; The speed is discretized into several equally spaced data intervals, namely, N = {N1(n1-Δn,n1+Δn),···,N m (n m -Δn,n m +Δn)}; For any small interval N n (n n -Δn,n n +Δn), and use the kernel density estimation method KDE to extract the power point with the highest probability, and take this point as the only power value in the interval; let the power point in the interval be P n1 , P n2 , P n3 ,···,P nq ; The kernel density estimate of the population density function f(P) at any point P is In the formula, K() is the kernel density function, q is the number of sample points, and h is the window width; The probability density set of all points in this interval is defined as Then there is Finally, the final value of the interval is: Problem 2): Adopt the "conservative" data comparison method; first select a conservative speed interval [ω1, ω2] within the maximum wind energy tracking area, that is, based on the operating characteristics of the wind turbine, the speed in this interval must be within the maximum wind energy tracking area; use the power-speed expression (18) of the maximum wind energy tracking area to perform data fitting; In the formula, λ opt is the optimal tip speed ratio, C Pmax is the best power factor; The above formula is abbreviated as P=kω 3 (19) In the formula, k is the proportionality coefficient; In the speed range [ω1,ω2], for a speed ω i , there is a corresponding power P i , denoted as (ω i ,P i );set up The number of discrete points in the interval is n, and the value of parameter k is obtained by the least squares method, that is: Then, let (ω c ,P c ) is a point outside the speed range [ω1,ω2], and the fitting curve P=k is obtained by substituting opt ω 3 , calculate the difference ΔP as: Set a critical value. When ΔP is greater than this critical value, it is considered that the point (ω c ,P c ) is no longer in the maximum wind energy tracking area; the value at the time of analysis gradually moves outward from the interval [ω1, ω2] port to both sides; when calculating, the critical value is selected from the fitting curve P = k opt ω 3 exist The maximum difference ΔP within the interval max , the calculation method is the same as formula (21), and the speed range of the wind turbine maximum wind energy tracking area is obtained as [ω c1 ,ω c2 ]; At this point, the division of the calculation condition area is completed.

5. The method for analyzing the on-site historical evolution behavior of aerodynamic performance of large wind turbines according to claim 4 is characterized in that: The implementation process of step 4 is as follows: the power coefficient based on instantaneous power and cumulative power is calculated according to formula (6) and formula (22), the incoming wind speed is calculated according to formula (8), and the wind speed filtering and power accumulation integration are performed according to the time constant τ (40s); Where, τ = t2-t1; Define the evaluation index C PI That is, the real-time value of the interval power coefficient calculated based on the instantaneous power and C PA That is, based on the real-time value of the interval power coefficient calculated based on the cumulative power, a "four-index" multi-angle observation algorithm is proposed, that is, the mean value of the interval power coefficient calculated based on the instantaneous power and the maximum value C PmaxI , the mean of the interval power coefficient calculated based on the cumulative power and the maximum value C PmaxA Four evaluation indicators are used to observe the on-site historical evolution of aerodynamic performance of large wind turbines.

Images