Modeling method and system considering wind power output correlation

Abstract

The invention discloses a modeling method and system considering wind power output correlation, and the method comprises the steps: obtaining the prediction error historical data of a target wind power plant, and selecting different edge distribution functions according to the prediction error historical data size of the target wind power plant, and the edge distribution functions comprise a first edge distribution function and a second edge distribution function; establishing a function model based on R-vine Copula according to the historical data of the prediction error of the target wind power plant, obtaining correlation coefficients among different variables in the historical data of the prediction error of the target wind power plant, and selecting an optimal tree structure by using a maximum spanning tree algorithm; according to the optimal tree structure, all sequence values in the function model of the R-vine Copula are calculated in combination with the AIC criterion, and modeling is conducted according to the sequence values. According to the modeling method and system, the correlation between the wind power plants can be reflected more accurately, so that the output prediction precision of the wind power plants is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of modeling considering wind power output correlation, and in particular to a modeling method and system considering wind power output correlation. Background Art

[0002] In the operation and management of wind farms, accurate wind power output forecasting is crucial for improving their economic benefits and grid stability. However, wind power output is affected by numerous factors, such as wind speed, direction, and temperature, making it highly random and uncertain. Furthermore, there may be correlations between different wind farms; for example, geographically close wind farms may be affected by the same climatic conditions. Therefore, modeling methods that account for wind power output correlations are crucial for improving forecast accuracy.

[0003] Some existing methods use statistical and time series analysis to predict wind power output, but these methods often ignore the correlation between wind farms. Other methods, while taking into account the correlation between wind farms, are complex in modeling and computationally intensive, making them difficult to widely apply in practice. Summary of the Invention

[0004] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of this application to avoid obscuring the purpose of this section, the abstract and the title of the invention, and such simplifications or omissions should not be used to limit the scope of the present invention.

[0005] In view of the above existing problems, the present invention is proposed.

[0006] Therefore, the present invention provides a modeling method and system that considers the correlation of wind power output, which can solve the problems mentioned in the background technology.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In a first aspect, the present invention provides a modeling method that considers wind power output correlation, comprising:

[0009] Acquire historical data of target wind farm prediction errors, and select different marginal distribution functions according to the amount of the historical data of target wind farm prediction errors, wherein the marginal distribution functions include a first marginal distribution function and a second marginal distribution function;

[0010] Establishing a function model based on R-vine Copula according to the target wind farm prediction error historical data, obtaining correlation coefficients between different variables in the target wind farm prediction error historical data, and selecting an optimal tree structure using a maximum spanning tree algorithm;

[0011] According to the optimal tree structure and in combination with the AIC criterion, all sequence values ​​in the function model of Rvine Copula are calculated, and modeling is performed based on the sequence values.

[0012] As a preferred solution of the modeling method considering wind power output correlation described in the present invention, wherein: the selecting of different marginal distribution functions according to the amount of historical data of target wind farm prediction errors includes:

[0013] When the amount of historical data is greater than a first preset amount, a first marginal distribution function is adopted;

[0014] When the amount of historical data is not greater than the first preset amount, the second marginal distribution function is used.

[0015] As a preferred solution of the modeling method considering wind power output correlation described in the present invention, wherein: the selecting of different marginal distribution functions according to the amount of historical data of target wind farm prediction errors also includes:

[0016] The steps of the first marginal distribution function are as follows:

[0017] First, given the target wind farm prediction error history data w, n discrete data sets {w1,w2...w n};

[0018] Then, the data in the n discrete data sets of the given target wind farm prediction error history data w are sorted in ascending order to obtain w'={w'1,w'2,...w' n};

[0019] Finally, the first marginal distribution function u and the inverse function of the first marginal distribution function of the variable w are calculated according to the following formula:

[0020]

[0021] Wherein, j=round(u*n), and round(·) is a rounding operation.

[0022] As a preferred solution of the modeling method considering wind power output correlation described in the present invention, wherein: the selecting of different marginal distribution functions according to the amount of historical data of target wind farm prediction errors also includes:

[0023] The steps to use the second marginal distribution function are as follows:

[0024] First, the target wind farm prediction error historical data interval [w min ,w max ] is divided into m segments, each segment is represented by [w k ,w k+1 ], where i = 0, 1, ... m-1;

[0025] Then use the linear function F k (x) = a k x+b k Fit the empirical cumulative probability distribution of each segment;

[0026] Finally, we get a according to the following formula k ,b k ,k=0,...m-1;

[0027]

[0028] Where l=1,...K represents the point number in each segment, and p(x) represents the empirical cumulative probability distribution expression.

[0029] As a preferred embodiment of the modeling method considering wind power output correlation described in the present invention, the method of establishing a function model based on R-vine Copula according to the historical data of target wind farm prediction errors, obtaining correlation coefficients between different variables in the historical data of target wind farm prediction errors, and selecting the optimal tree structure using a maximum spanning tree algorithm includes:

[0030] Initialization: Set the first layer, at which point the historical data of the forecast error of each target wind farm is considered an independent node;

[0031] Calculate the correlation coefficient: Use the improved Kendall correlation coefficient to calculate the correlation coefficient between all possible pairs of target wind farm prediction error historical data;

[0032] Select the maximum correlation coefficient: select the target wind farm prediction error historical data pair with the largest improved Kendall correlation coefficient to form an edge;

[0033] Constructing a tree structure: merging the selected target wind farm prediction error historical data pairs into a new node and entering the next layer;

[0034] Repeat the above steps for the remaining target wind farm prediction error historical data until all layers are constructed, completing the establishment of the function model based on R vine Copula.

[0035] As a preferred solution of the modeling method considering wind power output correlation described in the present invention, the improved Kendall correlation coefficient is used to calculate the correlation coefficient between all possible pairs of target wind farm prediction error historical data, including:

[0036]

[0037] Where P is the number of consistent pairs, i.e., the number of data with the same change trend of the prediction errors of the two target wind farms; Q is the number of inconsistent pairs, i.e., the number of data with opposite change trends of the prediction errors of the two target wind farms; T and U are the number of incomparable pairs of the first and second variables, i.e., the number of missing values ​​or outliers.

[0038] As a preferred solution of the modeling method considering wind power output correlation described in the present invention, wherein: the calculation of all sequence values ​​in the function model of Rvine Copula based on the optimal tree structure and combined with the AIC criterion includes:

[0039] For each edge in the optimal tree structure, the best binary copula function is selected using the AIC criterion, the AIC value of each candidate binary copula function is calculated, and the binary copula function with the smallest AIC value is selected;

[0040] Use the sequential estimation method to estimate the parameters of the binary copula on each edge separately from the first layer, use the conversion rule to convert the first layer values ​​to the second layer, and then estimate the parameters of the binary copula of the second layer. And so on, estimate the parameters of all layers.

[0041] In a second aspect, the present invention provides a modeling system that considers wind power output correlation, comprising:

[0042] a distribution function selection module, configured to obtain historical data of target wind farm prediction errors and select different marginal distribution functions according to the amount of the historical data of target wind farm prediction errors, wherein the marginal distribution functions include a first marginal distribution function and a second marginal distribution function;

[0043] An optimal tree structure establishment module is used to establish a function model based on R-vine Copula according to the historical data of the target wind farm prediction error, obtain the correlation coefficients between different variables in the historical data of the target wind farm prediction error, and select the optimal tree structure using the maximum spanning tree algorithm;

[0044] The model building module is used to calculate all sequence values ​​in the function model of Rvine Copula according to the optimal tree structure and the AIC criterion, and to perform modeling according to the sequence values.

[0045] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-described method when executing the computer program.

[0046] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the method described above when the computer program is executed by a processor.

[0047] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention proposes a modeling method and system that considers the correlation of wind power output, obtains the historical data of the target wind farm prediction error, and selects different marginal distribution functions according to the data volume of the historical data of the target wind farm prediction error, the marginal distribution function includes a first marginal distribution function and a second marginal distribution function; establishes a function model based on the R-vine Copula according to the historical data of the target wind farm prediction error, obtains the correlation coefficient between different variables in the historical data of the target wind farm prediction error, and selects the optimal tree structure using the maximum spanning tree algorithm; according to the optimal tree structure, combined with the AIC criterion, calculates all sequence values ​​in the function model of the R-vine Copula, and performs modeling according to the sequence values. The modeling method and system of the present invention can more accurately reflect the correlation between wind farms, thereby improving the accuracy of wind farm output prediction. Specifically, by using the improved Kendall correlation coefficient to calculate the correlation coefficient between pairs of historical data of target wind farm prediction errors, it is possible to effectively handle missing values ​​or outliers in the data, thereby improving the accuracy of the correlation coefficient. In addition, the use of the maximum spanning tree algorithm to select the optimal tree structure can ensure that the established model can capture the main correlations between wind farms, avoiding the redundant correlation problems that may occur in traditional methods.

[0048] In practical applications, the modeling method and system of the present invention can provide more reliable forecasting data for wind farm operation and scheduling, thereby helping to improve the wind farm's power generation efficiency and economic benefits. For example, during the daily operation of a wind farm, accurate forecasting of wind farm output can better plan the wind farm's power generation, reduce wind curtailment, and improve wind farm utilization. Furthermore, during the wind farm's scheduling process, the modeling method and system of the present invention can provide dispatchers with more accurate forecasting data, helping them make more reasonable scheduling decisions, thereby improving the stability and reliability of the entire power system.

[0049] In summary, the modeling method and system considering the correlation of wind power output proposed in the present invention have significant technical advantages and application value. They can provide more accurate prediction data for the operation and scheduling of wind farms, help improve the power generation efficiency and economic benefits of wind farms, and also help improve the stability and reliability of the entire power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort. Among them:

[0051] Figure 1 A flow chart of a modeling method and system considering wind power output correlation provided by one embodiment of the present invention;

[0052] Figure 2 An n-dimensional vine copula structure diagram of a modeling method and system considering wind power output correlation provided by one embodiment of the present invention;

[0053] Figure 3 A schematic diagram of the structure of a 5-dimensional R-vine Copula of a modeling method and system considering wind power output correlation provided by one embodiment of the present invention;

[0054] Figure 4 A modeling method and system for considering wind power output correlation provided by one embodiment of the present invention shows the marginal distribution of five wind farms in January (diagonal bar chart), scatter plots between each of them (lower left corner), and Pearson correlation coefficients (upper right corner);

[0055] Figure 5 A modeling method and system considering wind power output correlation provided by one embodiment of the present invention shows the edge distribution of five wind farms in June (diagonal bar chart), scatter plots between each other (lower left corner), and Pearson correlation coefficients (upper right corner);

[0056] Figure 6 (a) R-vine Copula, (b) D-vine Copula, and (c) C-vine Copula structure diagrams of a modeling method and system considering wind power output correlation provided by one embodiment of the present invention;

[0057] Figure 7 A cumulative probability distribution curve of 100 data points of a modeling method and system considering wind power output correlation provided by one embodiment of the present invention;

[0058] Figure 8 An enlarged diagram of the empirical probability distribution in the interval [-65, -30] of a modeling method and system considering wind power output correlation provided by one embodiment of the present invention;

[0059] Figure 9 An internal structural diagram of a computer device of a modeling method and system considering wind power output correlation provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0060] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0061] Example 1

[0062] Reference Figures 1-9 , which is the first embodiment of the present invention, provides a modeling method and system considering wind power output correlation, including:

[0063] Before introducing the embodiments of the present application in detail, some related concepts are first explained for the sake of clarity.

[0064] R Vine Copula Model: The term copula, coined by Sklar, comes from the Latin word for "link" or "bond," meaning to connect. This function represents the joint probability density distribution of multivariate random variables as a "connection" of their marginal distributions. It is a powerful tool for constructing the joint distribution of multivariate correlated random variables.

[0065] Theorem 1 (Sklar's Theorem): Let F be the marginal distribution F1, F2, ... F n The random variables x1,x2,...x n The joint probability distribution function of , then there exists a Copula function C, for any x∈R:

[0066] F(x1,x2,...x n )=C(F1(x1),F2(x2),...F n (x n ))

[0067] If F1, F2, ... F n are all continuous, then C is unique. On the contrary, if F1, F2, ... F n is a univariate distribution function, then F determined by (1) is a random variable x1, x2, ...x n The joint probability distribution of F1. -1 ,F2 -1 ,...F n -1 They are F1, F2, ... F n The pseudo-inverse function of C is n Both:

[0068] F(u1,u2,...u n )=C(F1 -1 (u1),F2 -1 (u2),...F n -1 (u n ))

[0069] The density function of the N-dimensional joint distribution function F can be obtained by the density function c of the n-ary Copula function C and the n marginal distribution functions F1, F2, ... F n It is concluded that:

[0070]

[0071] Among them, the Copula probability density function is:

[0072]

[0073] f i (x i ) are F i (x i ) corresponds to the density function.

[0074] Copula functions can be generated in a variety of ways. Commonly used copula families include the ellipsoidal and Archimedean families. The ellipsoidal family includes the Gaussion copula and t-Copula, while the Archimedean family includes the Gumbel copula, Clayton copula, and Frank copula. Different copula functions are suitable for different types of correlation. For example, the Gumbel copula is suitable for cases with strong correlation in the upper tail, the Clayton copula is suitable for cases with strong correlation in the lower tail, and the Frank copula is suitable for cases with symmetric correlation in the tails.

[0075] Sklar's theorem states that once the marginal distributions and copula functions of multivariate random variables are determined, the joint probability distribution of these random variables can be obtained. Copula functions impose no restrictions on marginal distributions, thus having a wide range of applicability. Sklar's theorem not only proves the existence of copula functions but also provides a modeling method. The modeling process consists of two steps: constructing the marginal distributions of each variable; and finding a suitable copula function and determining its parameters to describe the correlation structure between the variables.

[0076] Copula modeling requires parameter estimation in both marginal distribution and copula function estimation. There are two main methods for parameter estimation: maximum likelihood estimation and moment estimation.

[0077] Existing research results show that the difference between the two-stage maximum likelihood estimation and the one-step simultaneous estimation results is not significant. Therefore, the two-stage maximum likelihood method can be used to estimate the parameters. The steps are as follows:

[0078] For parameter θ i Solve

[0079]

[0080] Solve for parameter a.

[0081]

[0082] The two-stage maximum likelihood method greatly simplifies the parameter estimation process of the multivariate copula model.

[0083] Rvine Copula Model Structure: Although Copula offers many advantages as a tool for describing correlations between random variables, its applicability becomes limited when the dimension of the random variables exceeds two. The primary reason is that the types of Copulas that can be extended to higher dimensions are very limited. Furthermore, the correlations between multidimensional random variables can be even more complex, making them difficult to describe using a single Copula function.

[0084] Definition 1 (Rvine): If V=(T1, T2, ... T n-1 ) is an R-vine structure with n elements, then V needs to meet the following conditions:

[0085] T1 is a tree structure, whose nodes are represented by N1={1,…,n} and the set of edges is represented by E1;

[0086] For tree T i , i=2,…n-1,the number of nodes in the tree is N i =E i-1 , the set of edges is represented by E i ;

[0087] In T i ,i=1,…,in the n tree, {a,b}∈E i , where a={a1,a2}, b={b1,b2}, a, b must satisfy #(a∩b)=1, # indicates the number of elements taken.

[0088] R-vine Copula decomposition is to decompose the probability density function of an n-dimensional variable into a form consisting of n*(n-1) / 2 binary copulas. The n-dimensional vine copula structure diagram is as follows Figure 2 As shown:

[0089] Vine Copula can be represented by n-1 trees (tree, T), where the jth tree T j Contains n+1-j nodes (N) and nj edges (e), and each edge corresponds to a set of Copula functions. j The edges in the tree T j+1 becomes a node in the tree T; if the corresponding edge j Share a node, then T j+1 The two nodes in are connected by an edge. Therefore, the n-dimensional vine structure can be decomposed into n(n-1) / 2 edges.

[0090] When the information of R-vine Copula (F, V, B) is known, the density function of the multivariate variable is:

[0091]

[0092] Where x=(x1,x2,...x n ) is a random variable, e={a,b} means an edge consists of variables a and b, C e,* Represents the elements connected on the edge*, Denotes the set D e The elements in f i (·) is F i The density function of (·).

[0093] R-vine doesn't specify a uniform tree structure; instead, its structure is ultimately determined based on the actual relationships between all variables. Therefore, the R-vine tree structure encompasses not only the C-vine and D-vine models, but also represents a general form of vine structure.

[0094] The comparison between the structure of C vine, D vine and the copula structure of R vine is as follows: Figure 3 As shown. A, B, C, and D all represent different vine structures. A is a D vine structure, in which each variable in each layer is directly connected to at most two variables. B is a C vine structure, in which each layer has a central variable, and other variables are connected to the central variable. It is suitable for depicting situations where there are obvious guiding variables between multiple variables. C and D are R vine structures, which have no obvious rules and are more complex in structure. Therefore, they can flexibly depict nonlinear correlation structures between high-dimensional variables. At the same time, the ability to flexibly depict complex nonlinear correlation structures can be further enhanced by selecting the Copula function between each pair of variables.

[0095] Compared to C-vine and D-vine Copula functions, the R-vine Copula function model structure is more flexible, but it also requires more complex computational processes. In addition to the marginal distributions and the form of each pair-copula function required by the Copula function, the R-vine Copula model calculation also involves selecting an appropriate R-vine tree structure.

[0096] There are some issues with related technologies. For example, in wind farm prediction models, traditional statistical methods often fail to fully account for the complex correlations between wind farms, resulting in inaccurate predictions. Furthermore, historical wind farm prediction error data often contains missing values ​​or outliers, further affecting model accuracy.

[0097] This application provides an effective solution to the above-mentioned problems. Next, we will describe in detail how to implement the modeling method that considers the correlation of wind power output in combination with multiple embodiments.

[0098] Figure 1 A flow chart of a modeling method and system considering wind power output correlation is shown, including:

[0099] S101, obtaining historical data of target wind farm prediction errors, and selecting different marginal distribution functions according to the amount of the historical data of target wind farm prediction errors, where the marginal distribution functions include a first marginal distribution function and a second marginal distribution function;

[0100] In the embodiment of the present application, different marginal distribution functions are selected according to the amount of historical data of target wind farm prediction errors, including:

[0101] When the amount of historical data is greater than a first preset amount, a first marginal distribution function is adopted;

[0102] When the amount of historical data is not greater than the first preset amount, the second marginal distribution function is used.

[0103] In an optional embodiment, the first preset data volume may be a value manually set by a user or technician based on experience, a value set based on expert experience, or a value set by the user based on specific working conditions. The setting may also be performed according to the following steps:

[0104] First, review the size, quality, and distribution characteristics of the historical data. Check for missing values ​​or outliers.

[0105] Initial settings: Based on preliminary analysis, set a reasonable initial value. For example, assume the initial setting is 500 data points.

[0106] Verification: Experimentally verify whether the preset data size is appropriate. Techniques such as cross-validation can be used to evaluate model performance to see if it improves with increasing data size.

[0107] Adjustment: If you find that the model performance improves significantly with the increase in data volume, you can appropriately increase the preset data volume; conversely, if the increase in data volume has little impact on model performance, you can reduce the preset data volume.

[0108] Final setting: After several iterations, determine the first preset data volume that meets business needs and makes full use of existing data.

[0109] For example, when the data volume is less than 500, the model performance is unstable and easily affected by outliers. When the data volume is greater than 500, the model performance gradually stabilizes and the impact of outliers is weakened. In this case, the first preset data volume can be set to 500. If the data volume is less than 500, the second marginal distribution function is used; if the data volume is greater than 500, the first marginal distribution function is used.

[0110] In the embodiment of the present application, selecting different marginal distribution functions according to the amount of historical data of target wind farm prediction errors also includes:

[0111] The steps of the first marginal distribution function are as follows:

[0112] First, given the target wind farm prediction error history data w, n discrete data sets {w1,w2...w n};

[0113] Then, the data in the n discrete data sets of the given target wind farm prediction error history data w are sorted in ascending order to obtain w'={w'1,w'2,...w' n};

[0114] Finally, the first marginal distribution function u and the inverse function of the first marginal distribution function of the variable w are calculated according to the following formula:

[0115]

[0116] Wherein, j=round(u*n), and round(·) is a rounding operation.

[0117] Different marginal distribution functions are selected based on the historical data volume of target wind farm prediction errors, including:

[0118] The steps to use the second marginal distribution function are as follows:

[0119] First, the target wind farm prediction error historical data interval [w min ,w max] is divided into m segments, each segment is represented by [w k ,w k+1 ], where i = 0, 1, ... m-1;

[0120] Then use the linear function F k (x) = a k x+b k Fit the empirical cumulative probability distribution of each segment;

[0121] Finally, we get a according to the following formula k ,b k ,k=0,...m-1;

[0122]

[0123] Where l=1,...K represents the point number in each segment, and p(x) represents the empirical cumulative probability distribution expression.

[0124] In an optional embodiment, if the user selects a distribution function other than the first marginal distribution function and the second marginal distribution function, the steps for sampling the R-vine Copula model are as follows:

[0125] The Latin hypercube method is used to sample the n-dimensional [0, 1] independent uniformly distributed random number w. i ,i=1,...,n。

[0126] Will w i ,i=1,...,n is substituted into the inverse function of the conditional probability function of the following formula to obtain the random variable u i ,i=1,...,n sampling values. (u1,u2,...u n ) is the probability sampling result with uniform marginal distribution, that is,

[0127]

[0128] in,

[0129]

[0130] Use the inverse function x of the marginal distribution function i =F -1 (u i ) Convert u into the actual wind power sampling value.

[0131] For example, the sampling steps are described by taking three wind farms as an example.

[0132] First, simulate and obtain independent 3D [0, 1] uniformly distributed samples w i,i=1,2,3; then let u1=w1; then according to the known parameters, we can calculate F(u2|u1)=h(u2,u1,θ 11 ).therefore

[0133] u2=h -1 (w2,u1,θ 11 )

[0134] at last,

[0135] u3=F(u3|u1,u2)=h(F(u3|u1),F(u2|u1),θ 21 )

[0136] =h(h(u3,u1,θ 12 ),F(u2,u1),θ 21 )

[0137] Simplifying the above formula, we can get:

[0138] u3=h -1 (h -1 (w3,h(u2,u1,θ 11 ),θ 21 ),u1,θ 12 ).

[0139] It should be noted that the benefit of S101 lies in its ability to flexibly select an appropriate marginal distribution function based on the historical data of actual wind farm forecast errors, thereby improving the model's forecast accuracy and stability. This approach can better simulate the randomness and uncertainty of wind power, providing a more accurate reference for wind farm operation and scheduling.

[0140] S102, establishing a function model based on R-vine Copula according to the historical data of target wind farm prediction errors, obtaining correlation coefficients between different variables in the historical data of target wind farm prediction errors, and selecting an optimal tree structure using a maximum spanning tree algorithm;

[0141] In the embodiment of the present application, a function model based on R-vine Copula is established according to the historical data of the target wind farm prediction error, and the correlation coefficients between different variables in the historical data of the target wind farm prediction error are obtained, and the optimal tree structure is selected using the maximum spanning tree algorithm, including:

[0142] Initialization: Set the first layer, at which point the historical data of the forecast error of each target wind farm is considered an independent node;

[0143] Calculate the correlation coefficient: Use the improved Kendall correlation coefficient to calculate the correlation coefficient between all possible pairs of target wind farm prediction error historical data;

[0144] Select the maximum correlation coefficient: select the target wind farm prediction error historical data pair with the largest improved Kendall correlation coefficient to form an edge;

[0145] Constructing a tree structure: merging the selected target wind farm prediction error historical data pairs into a new node and entering the next layer;

[0146] Repeat the above steps for the remaining target wind farm prediction error historical data until all layers are constructed, completing the establishment of the function model based on R vine Copula.

[0147] In the embodiment of the present application, the improved Kendall correlation coefficient is used to calculate the correlation coefficient between all possible pairs of target wind farm prediction error historical data, including:

[0148]

[0149] Where P is the number of consistent pairs, i.e., the number of data with the same change trend of the prediction errors of the two target wind farms; Q is the number of inconsistent pairs, i.e., the number of data with opposite change trends of the prediction errors of the two target wind farms; T and U are the number of incomparable pairs of the first and second variables, i.e., the number of missing values ​​or outliers.

[0150] For example, assume there are three wind farms (referred to as wind farms A, B, and C), each with 10 prediction error data points. It is necessary to establish an R-vine Copula-based function model based on the historical prediction error data of these data points.

[0151] The historical data of wind farm A’s prediction error is: [-0.1, 0.3, -0.5, 0.2, 0.6, -0.3, 0.1, 0.4, -0.2, 0.5];

[0152] The historical prediction error data of wind farm B is: [0.2, -0.3, 0.4, -0.1, 0.6, 0.3, -0.2, 0.1, 0.5, -0.4];

[0153] The historical data of wind farm C’s prediction error is [0.3, 0.1, -0.2, 0.4, -0.1, 0.6, 0.2, -0.2, 0.5, -0.4];

[0154] Step 1: Initialization. The historical forecast error data for each wind farm is treated as an independent node. Therefore, in the first layer, there will be 30 nodes: A1, A2, ..., A10, B1, B2, ..., B10, C1, C2, ..., C10. Each node represents a forecast error data point for a wind farm.

[0155] Step 2: Calculate the improved Kendall correlation coefficient. Use the improved Kendall correlation coefficient formula to calculate the correlation coefficients between all possible pairs of historical data of target wind farm prediction errors.

[0156] Step 3: Select the maximum correlation coefficient. Assuming that the correlation coefficients between all wind farm pairs are calculated, select the wind farm pair with the largest improved Kendall correlation coefficient. For example, if the correlation coefficient between wind farms A and B is found to be the largest, then select A1 and B1 as the edges in the first step.

[0157] Step 4: Construct the tree structure and merge A1 and B1 into a new node A1B1. Go to the next level and you will now have nodes A1B1 and C1.

[0158] Repeat the above steps to calculate the correlation coefficient between nodes A1B1 and C1, select the pair with the largest correlation coefficient, form a new edge, and merge the nodes. Assuming that the correlation coefficient between A1B1 and C1 is the largest, select A1B1 and C1 as the next edge. Merge A1B1 and C1 into a new node A1B1C1, and so on, until all nodes have been merged into one node.

[0159] It should be noted that the benefit of step S102 is that it accurately captures the correlations between different variables in the target wind farm's historical forecast error data, thereby improving the accuracy and reliability of wind power forecasts. By selecting the optimal tree structure using the maximization spanning tree algorithm, the correlations between wind farms are properly reflected when simulating wind power, providing a more scientific decision-making basis for wind farm operation and scheduling.

[0160] S103, according to the optimal tree structure and in combination with the AIC criterion, all sequence values ​​in the function model of the R vine Copula are calculated, and modeling is performed based on the sequence values.

[0161] In the embodiment of the present application, according to the optimal tree structure and in combination with the AIC criterion, all sequence values ​​in the function model of the Rvine Copula are calculated, including:

[0162] For each edge in the optimal tree structure, the AIC criterion is used to select the best binary copula function, the AIC value of each candidate binary copula function is calculated, and the binary copula function with the smallest AIC value is selected;

[0163] Use the sequential estimation method to estimate the parameters of the binary copula on each edge separately from the first layer, use the conversion rule to convert the first layer values ​​to the second layer, and then estimate the parameters of the binary copula of the second layer. And so on, estimate the parameters of all layers.

[0164] In the embodiment of the present application, the optimal binary Copula function is selected. For each edge e, all functions C in the candidate binary Copula function set C are considered;

[0165] Calculate the AIC value of each binary Copula function C. Select the binary Copula function C with the smallest AIC value * as the optimal function.

[0166] For each binary copula function C on edge e: Assume C has k parameters. Assume the maximum likelihood estimate of C is It has been calculated. Then the AIC value is:

[0167]

[0168] Furthermore, for each edge e in the first layer: assuming that edge e connects nodes i and j, use maximum likelihood estimation to estimate the parameter θ of the binary copula function C;

[0169] For the binary Copula function C on edge e: let u i and u j are the marginal probabilities converted from marginal distributions. The log-likelihood function of the maximum likelihood estimate is:

[0170]

[0171] Where c is the density function of the binary copula function C. n is the number of observations.

[0172] The goal of maximum likelihood estimation is to minimize the negative log-likelihood function, that is:

[0173]

[0174] Furthermore, the marginal probabilities of the first layer are transferred to the second layer using a conversion rule. The parameter estimation step is repeated for each edge of the second layer.

[0175] For the edge e in the first layer connecting nodes i and j, the conversion rule is:

[0176]

[0177] Where: v ij is the marginal probability after transformation. i and u j is the marginal probability converted from the marginal distribution.

[0178] are the parameters of the maximum likelihood estimate.

[0179] Furthermore, for each edge of the second layer, repeat the above steps, and so on, until the parameters of all layers are estimated.

[0180] In an optional embodiment, the R-vine Copula parameters can also be estimated using a one-time estimation method, which writes a probability function based on the selected R-vine structure and uses the maximum likelihood estimation method to estimate all parameters at once.

[0181] In an optional embodiment, modeling based on sequence values ​​can be specifically performed as follows to construct a complete R-vine Copula model. The final R-vine Copula model can be constructed by combining the binary Copula functions and parameters of all layers. The entire model can be expressed using the following formula:

[0182]

[0183] Where: C is the final R vine Copula function. C i is the binary Copula function of the i-th layer. is the parameter of the binary Copula function of the i-th layer.

[0184] Once the final R-vine Copula model C is obtained, it can be used to simulate the joint distribution of wind power output. In order to simulate wind power output, samples can be sampled from the final R-vine Copula model C, and then the sampling results are converted back to the actual wind power output value through the inverse marginal distribution function.

[0185] Sampling process, sampling from C (u1,u2,…,u n ), for each u i , using the inverse marginal distribution function will u i Convert back to actual wind power output value x i :

[0186]

[0187] In summary, the present invention proposes a modeling method and system considering the correlation of wind power output, which obtains the historical data of the prediction error of the target wind farm, and selects different marginal distribution functions according to the data volume of the historical data of the prediction error of the target wind farm. The marginal distribution functions include the first marginal distribution function and the second marginal distribution function; a function model based on R-vine Copula is established according to the historical data of the prediction error of the target wind farm, and the correlation coefficients between different variables in the historical data of the prediction error of the target wind farm are obtained, and the optimal tree structure is selected by using the maximum spanning tree algorithm; according to the optimal tree structure, combined with the AIC criterion, all sequence values in the function model of R-vine Copula are calculated, and modeling is performed according to the sequence values. The modeling method and system of the present invention can more accurately reflect the correlation between wind farms, thereby improving the accuracy of wind farm output prediction. Specifically, by using the improved Kendall correlation coefficient to calculate the correlation coefficients between pairs of historical data of the prediction error of the target wind farm, missing values or outliers in the data can be effectively processed, thereby improving the accuracy of the correlation coefficients. In addition, using the maximum spanning tree algorithm to select the optimal tree structure can ensure that the established model can capture the main correlations between wind farms, avoiding the redundant correlation problems that may occur in traditional methods.

[0188] In practical applications, the modeling method and system of the present invention can provide more reliable prediction data for the operation and scheduling of wind farms, thereby helping to improve the power generation efficiency and economic benefits of wind farms. For example, in the daily operation of a wind farm, by accurately predicting the output of the wind farm, the power generation plan of the wind farm can be better arranged, the phenomenon of wind curtailment can be reduced, and the utilization rate of the wind farm can be improved. At the same time, in the scheduling process of the wind farm, the modeling method and system of the present invention can provide more accurate prediction data for the schedulers, helping them make more reasonable scheduling decisions, thereby improving the stability and reliability of the entire power system.

[0189] In summary, the modeling method and system considering the correlation of wind power output proposed by the present invention have significant technical advantages and application values, can provide more accurate prediction data for the operation and scheduling of wind farms, help improve the power generation efficiency and economic benefits of wind farms, and also help improve the stability and reliability of the entire power system.

[0190] This embodiment also provides a modeling system considering the correlation of wind power output, including:

[0191] A distribution function selection module, configured to obtain the historical data of the prediction error of the target wind farm, and select different marginal distribution functions according to the data volume of the historical data of the prediction error of the target wind farm. The marginal distribution functions include the first marginal distribution function and the second marginal distribution function;

[0192] The optimal tree structure establishment module is used to establish a function model based on R-vine Copula according to the historical data of target wind farm prediction errors, obtain the correlation coefficients between different variables in the historical data of target wind farm prediction errors, and select the optimal tree structure using the maximum spanning tree algorithm;

[0193] The model building module is used to calculate all sequence values ​​in the Rvine Copula function model based on the optimal tree structure and the AIC criterion, and to perform modeling based on the sequence values.

[0194] The above-mentioned unit modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the corresponding operations of the above-mentioned modules.

[0195] This embodiment also provides a computer device, which may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 9 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a modeling method that takes into account the correlation of wind power output is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0196] This embodiment further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the following steps are implemented:

[0197] Obtaining historical data of target wind farm prediction errors, and selecting different marginal distribution functions according to the amount of the historical data of target wind farm prediction errors, where the marginal distribution functions include a first marginal distribution function and a second marginal distribution function;

[0198] According to the historical data of target wind farm prediction errors, a function model based on R-vine Copula is established, and the correlation coefficients between different variables in the historical data of target wind farm prediction errors are obtained. The optimal tree structure is selected using the maximum spanning tree algorithm.

[0199] According to the optimal tree structure and combined with the AIC criterion, calculate all sequence values in the function model of the R-vine Copula, and perform modeling based on the sequence values.

[0200] Embodiment 2

[0201] Refer to Figure 2-Figure 8 , which is an embodiment of the present invention, provides a modeling method and system considering the correlation of wind power output. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0202] Use the data of wind farms #03579, #03693, #03902, #03945, #03978 released by the National Renewable Energy Laboratory (NREL) of the United States to test the modeling effect of the R-vine Copula. The marginal distributions and Pearson correlation coefficients of the historical output of the five wind farms in January are as Figure 4 and 5 shown:

[0203] By comparing Figure 4 and Figure 5 , it can be found that there are significant differences in the marginal distributions and correlations of the prediction errors of the five wind farms in January and June. Therefore, historical data of the same period should be selected when modeling the output uncertainty.

[0204] Use the R-vine Copula, C-vine Copula, D-vine Copula, and Gaussian Copula to model the prediction error data of the five wind farms in January. Take the wind power prediction errors in January of 2004, 2005, and 2006 as input data, a total of 2162 points. Select pair-Copula functions from Gaussian Copula, Student t Copula, Clayton Copula, Gumbel Copula, Frank Copula, and Joe Copula. When estimating the parameters of the vine Copula, the sequential estimation method is used for estimation. The following table lists the log-likelihood values of the R-vine Copula, C-vine Copula, D-vine Copula, and Gaussian Copula functions (only including the binary copula connection function part).

[0205] Table 1 Comparison of log-likelihood values of different models

[0206]

[0207] It can be seen from the log-likelihood values in Table 1 that the fitting effect of the R-vine Copula is better than that of other Copulas. Since the Gaussian Copula has only one functional form, there is a large gap in the fitting accuracy compared with the vine Copula.

[0208] Figure 6 The tree structures of Rvine Copula, Cvine Copula and Dvine Copula are listed. Figure 6 The structure shows that the sum of the Pearson correlation coefficients of the first layer of the R vine copula is 0.54(1,4)+0.6(4,5)+0.52(2,5)+0.54(2,3)=2.2; the sum of the Pearson coefficients of the first layer of the D vine copula is 0.5(1,2)+0.54(2,3)+0.50(3,4)+0.6(4,5)=2.14; and the sum of the Pearson coefficients of the first layer of the C vine copula is 0.5(1,4)+0.5(3,4)+0.53(2,4)+0.6(4,5)=2.13. That is, the sum of the Pearson coefficients of the first layer of the R vine is greater than the sum of the Pearson coefficients of the first layers of the D vine copula and the C vine copula. The sum of the Kendall correlation coefficients for the first level of the R-vine Copula is 0.38(1,4)+0.42(4,5)+0.38(2,5)+0.39(2,3)=1.57; the sum of the Pearson coefficients for the first level of the D-vine Kendall Copula is 0.35(1,2)+0.39(2,3)+0.34(3,4)+0.42(4,5)=1.5; and the sum of the Pearson coefficients for the first level of the C-vine Kendall Copula is 0.38(1,4)+0.34(3,4)+0.36(2,4)+0.42(4,5)=1.5. That is, the sum of the Kendall coefficients for the first level of the R-vine Copula is greater than the sum of the Kendall coefficients for the first levels of the D-vine and C-vine Copula. This shows that the R-vine Copula selects structures by maximizing the correlation between variables, matching highly correlated variables, while the C-vine and D-vine are restricted by fixed structures and have limited freedom in variable combination.

[0209] Table 2 lists the pair-copula types, parameters, and τ values ​​for the edges of the R-vine Copula tree. As can be seen, when the tree structure has a relatively low number of layers, the correlation between variables is high (τ is around 0.4 for the first layer). This correlation gradually decreases with increasing layers (τ is 0.09 for the third layer), even approaching correlated independence. Therefore, when the data dimension is large, multivariate copulas can be used instead of binary vine copulas for higher-level tree structures. In this case, even independence of variables can be assumed to reduce computational complexity.

[0210] Table 2 Copula types and parameters on the edge of R vine Copula

[0211]

[0212]

[0213] Using the empirical distribution curve, the sample u is converted into the actual power value. When there is sufficient historical data, a high-precision solution can be obtained, but when there is insufficient historical data, it will lead to large errors.

[0214] Figure 7 The cumulative probability distribution graph drawn using 100 historical data points is enlarged to show the forecast error interval [-65, -30]. Figure 8 shown.

[0215] Assuming that the sampling value u=0.36, according to the inverse rule of the empirical distribution function, w is the 36th historical data, then F -1 (0.36) = -50 MW. Using piecewise fitting, we obtain the function expression for this segment as u = 0.0047w + 0.565. Substituting this into the fitting equation, we obtain w = -43.617 MW. Compared to the empirical distribution result of -50 MW, the result of -43.671 MW is more reasonable.

[0216] In summary, after analyzing the fitting performance and correlation analysis of different copula models, we can conclude that R-vine copula has a clear advantage when processing high-dimensional data. Its flexible tree structure can better capture the correlation between variables, thereby providing a more accurate joint distribution estimate.

[0217] However, while R-vine Copula performs well in many situations, its computational complexity should be considered in practical applications. As the data dimension increases, the number of layers and edges in the tree structure also increases, leading to a significant increase in computational effort. Therefore, in practical applications, a trade-off must be struck between model accuracy and computational efficiency.

[0218] To further improve computational efficiency, some approximation methods can be considered. For example, a multivariate copula can be used instead of a binary vine copula for a high-level tree structure, or in some cases, the independence of variables can be assumed. These methods can effectively reduce computational effort without significantly reducing model accuracy.

[0219] Furthermore, when historical data is insufficient, interpolation and extrapolation methods can be used to improve forecast accuracy. For example, existing historical data points can be used to estimate the distribution characteristics of missing data points through interpolation. Furthermore, external information, such as weather forecasts and seasonal factors, can also be considered to enhance the model's predictive capabilities.

[0220] In practical applications, model robustness is also important. Because data in areas such as financial and energy markets often exhibit volatility and uncertainty, models must be able to adapt to these changes. Robust statistical methods, such as median regression and robust estimation, can be introduced to improve the model's stability in the face of outliers and noisy data.

[0221] In summary, R-vine Copula has significant advantages in modeling high-dimensional data, but its application still requires attention to computational efficiency and robustness. By rationally selecting model structures, employing approximate methods, and introducing robust statistical methods, computational efficiency and the ability to cope with data fluctuations can be effectively improved while maintaining model accuracy.

[0222] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0223] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code. The scheme in the embodiment of the present application can be implemented in various computer languages, for example, object-oriented programming language Java and literal translation scripting language JavaScript, etc.

[0224] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0225] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0226] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0227] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0228] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. A modeling method considering wind power output correlation, characterized in that: include: Acquire historical data of prediction errors of a target wind farm, and select different marginal distribution functions according to the amount of historical data of prediction errors of the target wind farm, wherein the marginal distribution functions include a first marginal distribution function and a second marginal distribution function; Establishing a function model based on R-vine Copula according to the historical data of prediction errors of the target wind farm, obtaining correlation coefficients between different variables in the historical data of prediction errors of the target wind farm, and selecting an optimal tree structure using a maximum spanning tree algorithm; According to the optimal tree structure and in combination with the AIC criterion, all sequence values ​​in the function model of Rvine Copula are calculated, and modeling is performed based on the sequence values.

2. The modeling method considering wind power output correlation according to claim 1, characterized in that: The selecting of different marginal distribution functions according to the amount of historical data of target wind farm prediction errors includes: When the amount of historical data is greater than a first preset amount of data, a first marginal distribution function is used; When the amount of historical data is not greater than the first preset amount, the second marginal distribution function is used.

3. The modeling method considering wind power output correlation according to claim 2, characterized in that: The selecting of different marginal distribution functions according to the historical data amount of prediction error of the target wind farm also includes: The steps of the first marginal distribution function are as follows: First, given the target wind farm prediction error history data w, n discrete data sets {w1,w2...w n }; Then, the data in the n discrete data sets of the given target wind farm prediction error historical data w are sorted from small to large, and w' = {w'1, w'2, ... w' n }; Finally, the first marginal distribution function u and the inverse function of the first marginal distribution function of the variable w are calculated according to the following formula: Wherein, j=round(u*n), and round(·) is a rounding operation.

4. The modeling method considering wind power output correlation according to claim 3, characterized in that: The selecting of different marginal distribution functions according to the historical data amount of prediction error of the target wind farm also includes: The steps to use the second marginal distribution function are as follows: First, the target wind farm prediction error historical data interval [w min ,w max ] is divided into m segments, each segment is represented by [w k ,w k+1 ], where i = 0, 1, ... m-1; Then, using the linear function F k (x) = a k x+b k Fit the empirical cumulative probability distribution of each segment; Finally, we get a according to the following formula k ,b k ,k=0,...m-1; Where l=1,...K represents the point number in each segment, and p(x) represents the empirical cumulative probability distribution expression.

5. The modeling method considering wind power output correlation according to claim 4, characterized in that: The method of establishing a function model based on R-vine Copula according to the target wind farm prediction error historical data, obtaining correlation coefficients between different variables in the target wind farm prediction error historical data, and selecting an optimal tree structure using a maximum spanning tree algorithm includes: Initialization: Set the first layer, at which point the historical data of the prediction error of each target wind farm is considered as an independent node; Calculate the correlation coefficient: Use the improved Kendall correlation coefficient to calculate the correlation coefficient between all possible pairs of historical data of target wind farm prediction errors; Select the maximum correlation coefficient: select the target wind farm prediction error historical data pair with the maximum improved Kendall correlation coefficient to form an edge; Constructing a tree structure: merging the selected historical data pairs of target wind farm prediction errors into a new node and entering the next layer; Repeat the above steps for the remaining target wind farm prediction error historical data until all layers are constructed, completing the establishment of the function model based on R vine Copula.

6. The modeling method considering wind power output correlation according to claim 5, characterized in that: The improved Kendall correlation coefficient calculates the correlation coefficient between all possible target wind farm prediction error historical data pairs, including: Among them, P is the number of consistent pairs, that is, the number of data with the same change trend of the prediction errors of the two target wind farms, Q is the number of inconsistent pairs, that is, the number of data with opposite change trends of the prediction errors of the two target wind farms, and T and U are the number of incomparable pairs of the first variable and the second variable, respectively, that is, the number of missing values ​​or outliers.

7. The modeling method considering wind power output correlation according to claim 6, characterized in that: The calculation of all sequence values ​​in the function model of Rvine Copula based on the optimal tree structure and the AIC criterion includes: For each edge in the optimal tree structure, the best binary Copula function is selected using the AIC criterion, the AIC value of each candidate binary Copula function is calculated, and the binary Copula function with the smallest AIC value is selected; Use the sequential estimation method to estimate the parameters of the binary Copula on each edge separately from the first layer, use the conversion rule to convert the first layer values ​​to the second layer, and then estimate the parameters of the binary Copula of the second layer. And so on, estimate the parameters of all layers.

8. A modeling system considering wind power output correlation, characterized in that: include: A distribution function selection module, used to obtain historical data of prediction errors of a target wind farm, and select different marginal distribution functions according to the amount of historical data of prediction errors of the target wind farm, wherein the marginal distribution functions include a first marginal distribution function and a second marginal distribution function; An optimal tree structure establishment module is used to establish a function model based on R-vine Copula according to the historical data of the prediction error of the target wind farm, obtain the correlation coefficients between different variables in the historical data of the prediction error of the target wind farm, and select the optimal tree structure by using the maximum spanning tree algorithm; The model building module is used to calculate all sequence values ​​in the function model of Rvine Copula according to the optimal tree structure and in combination with the AIC criterion, and to perform modeling according to the sequence values.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images