Wind power cluster short-term power prediction method, device and equipment and storage medium

Abstract

The application provides a wind power cluster short-term power prediction method, device, equipment and storage medium. It relates to the technical field of wind power cluster power short-term prediction. The method comprises: obtaining wind speed data of each wind farm in the cluster to form a first data set, and dividing the wind farm into multiple sub-clusters through clustering; determining a short-term correlation relationship based on the distance between wind farms; regarding the wind farms in the same sub-cluster as the same type of node, and constructing an isomorphic graph to represent the isomorphic space-time relationship; regarding the wind farms in different sub-clusters as different types of nodes, and constructing a heterogeneous graph to represent the heterogeneous space-time relationship; dynamic graph network design: constructing a dynamic graph attention network, including an isomorphic module and a heterogeneous module, the isomorphic module extracts a space-time feature vector of each wind farm based on the isomorphic graph; the heterogeneous module fuses the heterogeneous graph and the space-time feature vector to output a power prediction result. The application accurately captures the complex space-time correlation of the wind farm group through clustering division and isomorphic / heterogeneous double-graph joint modeling, and improves the prediction accuracy.

Description

Technical Field

[0001] This application relates to the field of short-term power prediction technology for wind power clusters, and in particular to a method, apparatus, equipment and storage medium for short-term power prediction of wind power clusters. Background Technology

[0002] With the rapid development of the wind power industry, the construction model of wind farms is gradually evolving from early decentralized, small-scale, and low-penetration methods to centralized, large-scale, and high-penetration clustered layouts. The power output characteristics of these large-scale wind power clusters are not only significantly affected by macro-weather systems but also face challenges from the inherent biases in the spatiotemporal accuracy and prediction magnitude of numerical weather prediction (NWP). Unlike isolated wind farms, wind farms within a cluster exhibit more complex spatiotemporal correlations. Traditional prediction models based on single wind farm information or simple aggregation are insufficient to effectively capture the dynamic interactions within the cluster and the complex mechanisms by which global meteorological conditions affect overall power output, thus failing to meet increasingly sophisticated operational needs. How to fully utilize multi-dimensional NWP data to deeply mine and accurately characterize the complex spatiotemporal dependencies within large-scale wind power clusters has become one of the core technical bottlenecks that urgently need to be addressed in the field of wind power prediction.

[0003] To address the power prediction problem of wind power clusters, existing technical approaches mainly include the holistic method, the cumulative method, and the statistical upscaling method. The holistic method directly predicts the total power of the cluster by averaging the characteristics of all wind farms within the cluster; however, this approach coarsens the spatiotemporal differences and dynamic relationships between individual wind farms. While the cumulative method models each wind farm independently and then superimposes the power, it ignores the dynamic impact of interactions between wind farms. The statistical upscaling method relies on data from a small number of benchmark wind farms to infer the power output of the entire region; however, this inference method based on sparse, single-point meteorological information cannot accurately reflect the spatiotemporal evolution of the weather system in the entire cluster area. To further improve prediction accuracy, researchers have introduced multiple-input multiple-output (MIMO) models, such as convolutional neural networks (CNNs) and multilayer perceptrons (MLPs). While CNNs can extract local spatiotemporal features, their fixed receptive field limits their ability to capture global information and long-range dependencies; MLPs, when processing high-dimensional spatiotemporal data, are prone to losing key spatiotemporal structural information due to feature vectorization. While attention mechanisms (AM) enhance a model's ability to model long sequences and multi-objective associations by assigning different weights to different spatiotemporal features, their applicability and effectiveness in handling the complex interactions between nodes with different characteristics (i.e., heterogeneous nodes) in large-scale wind power clusters still need improvement. In recent years, graph neural networks (GNNs), with their unique advantages in processing non-Euclidean space data and modeling complex topological relationships, have been introduced into the field of wind power prediction, providing a new perspective for addressing the aforementioned challenges. However, current research based on GNNs mostly focuses on the application of isomorphic graphs. In such graphs, nodes (wind farms) are assumed to have single or similar attributes, and edge relationships are relatively fixed. This setting can only effectively represent the spatiotemporal associations within small-scale, similarly characterized wind farm sub-clusters, and is insufficient to effectively address the heterogeneity problem prevalent in large-scale wind power clusters—that is, different wind farms may exhibit significantly different output characteristics and response patterns due to factors such as geographical location, micro-meteorological conditions, equipment models, and operation and maintenance strategies.

[0004] Existing methods lack a systematic theoretical framework and effective technical solutions for constructing and utilizing heterogeneous graphs to capture the complex and dynamic heterogeneous relationships between different types of wind farms and different sub-clusters, resulting in the failure to fully explore and utilize this valuable heterogeneous information.

[0005] In summary, existing wind power forecasting technologies still have significant limitations when dealing with large-scale, high-density wind farm clusters, especially in accurately characterizing the complex dynamic spatiotemporal correlations between homogeneous and heterogeneous wind farms within the cluster. Therefore, there is an urgent need to develop an advanced forecasting method that can effectively integrate homogeneous and heterogeneous information within the cluster and dynamically capture and characterize their complex interactions, in order to significantly improve the accuracy and reliability of short-term power forecasting for large-scale wind farm clusters. Summary of the Invention

[0006] This application provides a method, apparatus, equipment, and storage medium for short-term power prediction of wind power clusters, in order to overcome the problems in the prior art such as insufficient characterization of complex spatiotemporal correlations within wind farm clusters and inability to dynamically adapt to meteorological changes.

[0007] In a first aspect, this application provides a short-term power prediction method for wind power clusters, including:

[0008] Wind speed data from each wind farm within the wind power cluster are acquired to form a first dataset. This first dataset is then clustered to obtain multiple sub-wind power clusters. The first dataset is represented as X = {x1, x2, ..., x...}. n}, x1, x2 and x n These represent the wind speed data for the first, second, and nth wind farms, respectively, and the sub-wind power cluster includes at least one wind farm.

[0009] The short-term correlation between wind farms is determined based on the distance between them;

[0010] All wind farms in the same sub-wind power cluster are treated as nodes of the same output type, and the spatiotemporal relationship of nodes of the same output type is taken as the isomorphic spatiotemporal relationship to construct the isomorphic graph of the wind power cluster.

[0011] Wind farms in different sub-wind power clusters are treated as nodes of different output types, and the spatiotemporal relationships of nodes of different output types are used as heterogeneous spatiotemporal relationships to construct a heterogeneous graph of the wind power cluster.

[0012] Based on the short-term correlation between wind farms, the adjacency matrices of the isomorphic and heteromorphic graphs of wind power clusters are determined.

[0013] A dynamic graph attention network is constructed. The dynamic graph attention network includes a homogeneous module and a heterogeneous module. The homogeneous module calculates the spatiotemporal feature vector of each wind farm based on the homogeneous graph of the wind power cluster and the wind speed data of each wind farm in each sub-wind power cluster. The heterogeneous module outputs the predicted power of each wind farm based on the heterogeneous graph of the wind power cluster and the spatiotemporal feature vector of each wind farm.

[0014] A second dataset is constructed, and the dynamic graph attention network is trained using the second dataset to achieve short-term power prediction of wind power clusters; wherein, the second dataset includes wind speed data and wind power data of each wind farm.

[0015] In one possible design, the first dataset is clustered to obtain multiple sub-wind power clusters, including:

[0016] Set k initial centroids;

[0017] The Euclidean distances between the wind farm and each centroid are calculated using the following formula:

[0018]

[0019] In the formula, d(x) i ,c m () represents the wind speed data x of the i-th wind farm. i With the m-th centroid c m The Euclidean distance of the wind speed data, where t represents the time point and x it c represents the wind speed data of the i-th wind farm at time t. mt c represents the m-th centroid. m Wind speed data at time point t;

[0020] Based on the Euclidean distances between the wind farm and each centroid, the wind farm is allocated using the following formula:

[0021]

[0022] In the formula, S m Let m be the m-th cluster, and arg min means to take the minimum value;

[0023] The centroid is updated using the following formula:

[0024]

[0025] The process of repeatedly assigning and updating centroids continues until a set condition is met, which is expressed as:

[0026]

[0027] or

[0028] t'≥T' (5)

[0029] In the formula, max represents the maximum value function. ε This indicates the set threshold. This represents the wind speed data of the m-th centroid at the (t'+1)-th and t'-th iterations, where t' represents the current iteration number and T' represents the maximum iteration number.

[0030] In one possible design, the short-term correlation between wind farms is determined based on the distance between them, including:

[0031] The Euclidean distance between two wind farms is calculated using the following formula:

[0032]

[0033] In the formula, d(W1,W2) represents the Euclidean distance between two wind farms W1 and W2, and v 1t and v 2t This represents the wind speed data of two wind farms, W1 and W2, at time t.

[0034] Based on the Euclidean distance between the two wind farms, normalization is performed using the following formula:

[0035]

[0036] In the formula, d nom d represents the normalized distance value. max This represents the maximum value of the Euclidean distance between all wind farms;

[0037] Based on the set distance threshold, the STC relationship between two wind farms is determined using the following formula:

[0038]

[0039] In the formula, STC(W1,W2) represents the short-term correlation between wind farms W1 and W2. STC(W1,W2) = 1 indicates that wind farms W1 and W2 have an STC relationship, and STC(W1,W2) = 0 indicates that wind farms W1 and W2 do not have an STC relationship.

[0040] In one possible design, the isomorphic graph of the wind power cluster is represented as G. Ho ={G Ho,1 ,...,G Ho,c ,...,G Ho,C}, where G Ho,1 G Ho,c and G Ho,C Let G represent the isomorphic graphs of the first, c-th, and C-th sub-wind power clusters, respectively. Ho,c = (V, E), where V and E are the sets of nodes and edges in the isomorphic graph of the c-th sub-wind power cluster, respectively, and V is described as {V1, V2, ..., V...}. i}, V1, V2 and V i These represent the first, second, and i-th nodes in the same sub-wind power cluster, respectively. E ij C represents the directed edge relationship between node i and node j in the same sub-wind power cluster. n This represents the set of nodes in a specific sub-wind power cluster.

[0041] The heterogeneous graph of the wind power cluster is represented as G. He =(Φ C ,V',E'),Φ CLet V' and E' represent the meta-path, where V' and E' are the sets of nodes and edges in the heterogeneous graph, respectively, and V' = {V'1, ..., V'}. c ,...,V' C}, where V'1 and V' c and V' C Let represent the node sets within the first, c-th, and C-th sub-wind power clusters, respectively. Where E' ij C represents the directed edge relationship between node i and node j within two different sub-wind power clusters. i and C j This represents the set of nodes within two different sub-wind power clusters.

[0042] In one possible design, based on the short-term correlation between wind farms, the adjacency matrices of the isomorphic and heteromorphic graphs of the wind farm cluster are determined using the following formula:

[0043] Adjw i,j =STC i,j i=1,...,n; j=1,...,n-1 (9)

[0044] In the formula, Adjw i,j STC represents the adjacency matrix of nodes i and j. i,j This represents the short-term correlation between node i and node j, where n is the number of nodes or wind farms. If node i and node j are located in the same sub-wind farm cluster, then Adjw i,j Given an adjacency matrix of an isomorphic graph, if node i and node j are located in different sub-wind power clusters, then Adjw i,j is the adjacency matrix of the heterogeneous graph.

[0045] In one possible design, the isomorphic module calculates the spatiotemporal feature vector of each wind farm based on the isomorphic graph of the wind power cluster and the wind speed data of each sub-wind power cluster, including the following methods:

[0046] The attention coefficient for each sub-cluster is calculated using the following formula:

[0047] e(x Ci ,x Cj ) = a T [Wx Ci P Wx Cj j∈N Ci (13)

[0048] In the formula, e(x) Ci ,x Cj ) represents the attention coefficient of node j after nonlinear transformation of node i in the sub-wind power cluster C; LeakyReLU(·) is a nonlinear transformation function; aT (·) represents the attention coefficient; P is the feature concatenation operation; x Ci and x Cj are the input feature vectors of nodes i and j, respectively, and W is the shared parameter matrix of the linear transformation; N Ci γ is the total number of neighboring nodes of node i; γ is the positive slope of the negative region of the function;

[0049] The attention coefficients are normalized to obtain attention weights, and then a spatiotemporal vector is obtained by weighted summation. The calculation process is as follows:

[0050]

[0051]

[0052] In the formula, softmax j (·) is the attention coefficient e of the normalized neighbor node j. Cij The function is σ(·), which is the sigmoid activation function; K is the number of heads in the multi-head attention mechanism, and k is the head index of the multi-head attention mechanism; exp represents the empirical function; α Cij Indicates attention weight; N represents a spacetime vector; i express;

[0053] Based on the spatiotemporal vector, the spatiotemporal feature vector is calculated using the following formula:

[0054]

[0055]

[0056] In the formula, f CAM (·) indicates the unified extraction of spatial and temporal features of wind power clusters, f MLP Let AvgPool represent the nonlinear transformation function guided by the multilayer perceptron, and MaxPool represent the average pooling layer and the max pooling layer, respectively. This represents the feature vector obtained after average pooling. f represents the feature vector obtained after max pooling. CAM (x Ci ) represents the eigenvector set of the original wind farm, and e represents the Hama operator. This represents the spatiotemporal feature vector.

[0057] In one possible design, the heterogeneous module outputs the predicted power of each wind farm based on the heterogeneous graph of the wind power cluster and the spatiotemporal feature vector of each wind farm, including the following methods:

[0058] Based on the heterogeneous graph of the wind power cluster, a node pair (i,j) and the meta-path Φ connecting the node pair are obtained, and the asymmetric node attention coefficient is calculated using the following formula:

[0059]

[0060] In the formula, It is a transformation matrix that processes the i-th sub-wind power cluster into the same feature space; Φ i It is the i-th meta-path; It is an asymmetric node attention coefficient; For the metapath Φ i The total number of wind farms under; For the metapath Φ i The number of adjacent wind farms of wind farm i;

[0061] Normalization using softmax(·) yields the node-level attention vector based on the meta-path, calculated as follows:

[0062]

[0063]

[0064] In the formula, This is a node-level attention vector; Is node i traversed via metapath Φ i The output feature vector; For the metapath Φ i The total number of wind farms under; For the metapath Φ i The number of adjacent wind farms of wind farm i; K is the total number of attention heads; k is the current attention head.

[0065] Based on the node-level attention vector, the predicted power of the wind farm is calculated using the following formula:

[0066]

[0067] In the formula, W L A is the weight matrix; A is a biased vector; q is the type-level attention vector. It is a type-level attention vector; i is the i-th wind farm; Φ i denoted as the metapath; C is the total number of wind farms in the sub-cluster; tanh is a nonlinear function; b is the weighting coefficient. These are type-level attention weights; x pre P is the predicted power of the wind farm; P is the number of wind farms under the current metapath. It is the feature vector output at the node level.

[0068] Secondly, this application provides a short-term power prediction device for wind power clusters, the device comprising:

[0069] The data acquisition module is configured to acquire wind speed data from each wind farm within the wind power cluster, forming a first dataset, and then cluster the first dataset to obtain multiple sub-wind power clusters; wherein, the first dataset is represented as X = {x1, x2, ..., x...} n}, x1, x2 and x n These represent the wind speed data for the first, second, and nth wind farms, respectively, and the sub-wind power cluster includes at least one wind farm.

[0070] The correlation calculation module is configured to determine the short-term correlation between wind farms based on the distance between them.

[0071] The isomorphic graph construction module is configured to treat all wind farms in the same sub-wind power cluster as nodes of the same output type, and use the spatiotemporal relationship of nodes of the same output type as the isomorphic spatiotemporal relationship to construct the isomorphic graph of the wind power cluster.

[0072] The heterogeneous graph construction module is configured to treat wind farms in different sub-wind power clusters as nodes of different output types, and use the spatiotemporal relationship of nodes of different output types as heterogeneous spatiotemporal relationship to construct a heterogeneous graph of the wind power cluster.

[0073] The adjacency matrix construction module is configured to determine the adjacency matrices of isomorphic and heteromorphic graphs of wind farm clusters based on the short-term correlation between wind farms.

[0074] The graph attention network module is configured to construct a dynamic graph attention network. The dynamic graph attention network includes a homogeneous module and a heterogeneous module. The homogeneous module calculates the spatiotemporal feature vector of each wind farm based on the homogeneous graph of the wind power cluster and the wind speed data of each wind farm in each sub-wind power cluster. The heterogeneous module outputs the predicted power of each wind farm based on the heterogeneous graph of the wind power cluster and the spatiotemporal feature vector of each wind farm.

[0075] The network training module is configured to construct a second dataset and use the second dataset to train the dynamic graph attention network, so as to realize short-term power prediction of wind power clusters; wherein, the second dataset includes wind speed data and wind power data of each wind farm.

[0076] Thirdly, embodiments of this application provide an electronic device, including: at least one processor and a memory; the memory stores computer execution instructions; the at least one processor executes the computer execution instructions stored in the memory, causing the at least one processor to perform the short-term power prediction method for wind power clusters as described in the first aspect and various possible designs of the first aspect.

[0077] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions. When a processor executes the computer-executable instructions, it implements the short-term power prediction method for wind power clusters as described in the first aspect and various possible designs of the first aspect.

[0078] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the short-term power prediction method for wind power clusters as described in the first aspect and various possible designs of the first aspect.

[0079] The short-term power prediction method, apparatus, equipment, and storage medium for wind power clusters provided in this application have at least the following beneficial effects:

[0080] This application improves the accuracy and reliability of wind farm cluster power prediction by fusing homogeneous and heterogeneous information to dynamically construct a graph structure, overcoming the problems in existing technologies such as insufficient characterization of complex spatiotemporal correlations within wind farm clusters and inability to dynamically adapt to meteorological changes. Attached Figure Description

[0081] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0082] Figure 1 A flowchart illustrating a short-term power prediction method for wind power clusters provided in this application embodiment;

[0083] Figure 2 This is a flowchart illustrating the clustering process of the first dataset provided in an embodiment of this application.

[0084] Figure 3 A flowchart illustrating the determination of short-term correlations between wind farms provided in this application embodiment;

[0085] Figure 4 The process of constructing homogeneous and heterogeneous graphs provided in the embodiments of this application;

[0086] Figure 5 This is a schematic diagram illustrating the structure and data processing flow of a dynamic graph attention network provided in an embodiment of this application.

[0087] Figure 6 A time series diagram comparing wind farm prediction results provided in the embodiments of this application;

[0088] Figure 7 This is a structural diagram of the short-term power prediction device for wind power clusters provided in an embodiment of this application.

[0089] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0090] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0091] The collection, storage, use, processing, transmission, provision, and disclosure of financial data or user data involved in the technical solution of this application all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0092] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.

[0093] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0094] Based on this, embodiments of this application provide a short-term power prediction method for wind power clusters, such as... Figure 1 The diagram shows a flowchart of a short-term power prediction method for wind power clusters provided in an embodiment of this application. The short-term power prediction method for wind power clusters includes the following steps S100-S700.

[0095] S100: Obtain wind speed data of each wind farm within the wind power cluster to form the first dataset, and cluster the first dataset to obtain multiple sub-wind power clusters.

[0096] In this embodiment, the first dataset is represented as X = {x1, x2, ..., x...} n}, x1, x2 and x n These represent the wind speed data for the first, second, and nth wind farms, respectively. A sub-wind farm cluster includes at least one wind farm.

[0097] In one exemplary embodiment, the K-means clustering method is used to cluster wind farms within a wind power cluster based on 96 wind speed data points at 15-minute resolution for the day. First, wind speed data measured every 15 minutes for each wind farm within the cluster are collected to form a dataset, where each row represents a wind farm and each column represents the wind speed at a specific time point, resulting in 96 columns. Let the first dataset be X = {x1, x2, ..., x...} n}, where x i =(x i1 ,x i2 ,...,x i96 Let represent the wind speed data of the i-th wind farm, and n be the number of wind farms. Next, the K-means clustering process is initialized, selecting k initial centroids C = {c1, c2, ..., c...}. k}, where c m =(c m1 ,c m2 ,...,c m96 () represents the m-th centroid. See also Figure 2 Clustering of the first dataset is achieved through the following iterative process:

[0098] S101: Assignment cluster.

[0099] For each wind farm x i Calculate its relationship with each centroid c. j Euclidean distance between them:

[0100]

[0101] In the formula, d(x) i ,c m () represents the wind speed data x of the i-th wind farm. i With the m-th centroid c m The Euclidean distance of the wind speed data, where t represents the time point and x it c represents the wind speed data of the i-th wind farm at time t. mt c represents the m-th centroid. m Wind speed data at time point t;

[0102] x i Assigned to the nearest centroid c m The corresponding cluster S m In, that is:

[0103]

[0104] In the formula, S m Let m be the m-th cluster, and arg min represent taking the minimum value;

[0105] S102: Update the centroid.

[0106] Calculate S for each cluster m The mean of wind speed data from all wind farms is used to update the centroid c. m :

[0107] Update the centroid using the following formula:

[0108]

[0109] S103: Repeat the above steps of assigning clusters and updating centroids until the set conditions are met.

[0110] The condition is that the change in the centroid is less than a set threshold. ε Or it can reach the maximum number of iterations T', that is:

[0111]

[0112] or

[0113] t'≥T' (5)

[0114] In the formula, max represents the maximum value function. ε This indicates the set threshold. This represents the wind speed data of the m-th centroid at the (t'+1)-th and t'-th iterations, where t' represents the current iteration number and T' represents the maximum iteration number.

[0115] Ultimately, each wind farm will be assigned to a specific cluster. Wind farms within the same cluster will have a high degree of similarity in their wind speed change trends on the same day, thus achieving effective clustering of wind farms within the wind power cluster.

[0116] S200: Determine the short-term correlation between wind farms based on the distance between them;

[0117] In one exemplary embodiment, the distance between wind farms is calculated based on Euclidean distance, and it is determined whether they possess a Short-Term Correlation (STC) relationship. The presence or absence of an STC relationship characterizes the short-term correlation between wind farms. Please refer to [link to relevant documentation]. Figure 3 The short-term correlation between wind farms can be determined through the following steps:

[0118] S201: Calculate the Euclidean distance.

[0119] Suppose the wind speed data of two wind farms W1 and W2 in a wind power cluster on a certain day are two 96-dimensional vectors v1 = [v 11 ,v 12 ,...,v 196 ] and v2 = [v 21 ,v 22 ,...,v 296 [The image shows a graph with each element representing a wind speed value at a 15-minute resolution. The Euclidean distance between the two values ​​can be calculated using the following formula:]

[0120]

[0121] In the formula, d(W1,W2) represents the Euclidean distance between two wind farms W1 and W2, and v 1t and v 2t This represents the wind speed data of two wind farms, W1 and W2, at time t.

[0122] Formula (6) calculates the square root of the sum of squares of the wind speed differences between two wind farms at 96 time points.

[0123] S202: Normalization process.

[0124] To eliminate the influence of data scale, the calculated Euclidean distance needs to be normalized. Normalization methods typically involve mapping the distance values ​​to a range of 0 to 1.

[0125]

[0126] In the formula, d nom d represents the normalized distance value. max This represents the maximum value of the Euclidean distance between all wind farms.

[0127] S203: Determining the STC relationship.

[0128] When the normalized distance value d norm When the threshold θ = 0.25 is less than the threshold value, the two wind farms are considered to have an STC relationship, and STC = 1 is defined; otherwise, STC = 0 is defined. The judgment formula is as follows:

[0129]

[0130] In the formula, STC(W1,W2) represents the short-term correlation between wind farms W1 and W2. STC(W1,W2) = 1 indicates that wind farms W1 and W2 have an STC relationship, and STC(W1,W2) = 0 indicates that wind farms W1 and W2 do not have an STC relationship.

[0131] Through the above steps S201-S203, the short-term correlation between wind farms within a wind power cluster can be effectively determined, providing technical support for the subsequent adjacency matrix.

[0132] S300: Treat all wind farms in the same sub-wind power cluster as nodes of the same output type, and construct an isomorphic graph of the wind power cluster based on the spatiotemporal relationship of nodes of the same output type.

[0133] S400: Treat wind farms in different sub-wind power clusters as nodes of different output types, and use the spatiotemporal relationship of nodes of different output types as heterogeneous spatiotemporal relationship to construct a heterogeneous graph of wind power clusters.

[0134] Traditional large-scale wind power cluster power prediction typically divides wind farms into several sub-clusters with similar output habits based on their historical output characteristics, and then models and predicts for each sub-cluster. To comprehensively construct the spatiotemporal topology of large-scale wind power clusters, this embodiment further refines the different node attributes and relationship attributes of the wind power clusters after partitioning them. Two spatiotemporal relationships applicable to both large-scale wind power clusters and sub-clusters are designed. In the first sub-cluster, all wind farms are nodes of the same output type, and the spatiotemporal relationship between nodes of the same output type is a homogeneous spatiotemporal relationship. In the second sub-cluster, wind farms are nodes with different output types, and the spatiotemporal relationship between nodes of different output types is a heterogeneous spatiotemporal relationship. The interaction diagram between homogeneous and heterogeneous spatiotemporal relationships is shown below. Figure 4 As shown, the topologies of large-scale wind power clusters based on these two types of associations are homogeneous and heterogeneous, respectively.

[0135] Based on the above basic principles, this embodiment constructs the isomorphic and heteromorphic graphs of the wind power cluster through steps S300 and S400, respectively.

[0136] In an exemplary embodiment, the isomorphic graph of a sub-wind cluster of a wind power cluster can be represented as G. Ho,c = (V, E), where V and E are the sets of nodes and edges, respectively. The isomorphic graph constructed in parallel by multiple sub-wind power clusters is G. Ho ={G Ho,1 ,...,G Ho,c ,...,G Ho,C}, C is the total number of sub-wind power clusters, and G Ho,c The interior of V can be described as {V1,V2,...,V} i}, where i∈[1,C i [i] refers to i nodes within the same sub-wind power cluster. Similarly, Where E ij This represents the directed edge relationship between node i and node j in the same sub-wind power cluster.

[0137] In an exemplary embodiment, the heterogeneity graph of a wind power cluster can be represented as G. He =(Φ C ,V',E'), where Φ C Let V' and E' represent the metapath, where V' and E' are sets of nodes and edges, respectively. The nodes in V' are described as V' = {V'1, ..., V'...}. c ,...,V' C}, where is the set of nodes within the c-th sub-wind power cluster, i∈[1,N]. c ] is containing N c V of each node c The i-th node in the cluster contains n sub-wind power clusters. Similarly, Where E' ij This represents the directed edge relationship between nodes i and j within two different sub-clusters. Different types of nodes are used to construct different edge relationships between wind farms using a bipartite graph.

[0138] In homogeneous graphs, the attributes of nodes and edges are relatively uniform, allowing for feature extraction and relation modeling using homogeneous GNNs. In heterogeneous graphs, the attributes of nodes and edges have different output and error characteristics. Processing heterogeneous graphs requires heterogeneous GNNs to consider the characteristics of each type of node and edge, as well as their interactions.

[0139] S500: Determine the adjacency matrices of isomorphic and heteromorphic graphs of wind farm clusters based on the short-term correlation between wind farms.

[0140] Adjacency matrix Adj∈R N×N Used to describe the relationship between any two connected nodes. Note that if the graph relationship identification result is STC... i,j =1 between wind farm WF i and WF j If there is an edge between two nodes, then there is an edge between them, denoted as Adjw. i,j =1. Otherwise, Adjw i,j =0 indicates that there is no connection between them. This ensures the sparsity of the adjacency matrix, which helps to reduce the interference of unnecessary noise.

[0141] In an exemplary embodiment, the adjacency matrices of the isomorphic and heteromorphic graphs of the wind power cluster are determined by the following formula:

[0142] Adjw i,j =STC i,j i=1,...,n; j=1,...,n-1 (9)

[0143] In the formula, Adjw i,j STC represents the adjacency matrix of nodes i and j.i,j This represents the short-term correlation between node i and node j, where n is the number of nodes or wind farms. If node i and node j are located in the same sub-wind farm cluster, then Adjw i,j Given an adjacency matrix of an isomorphic graph, if node i and node j are located in different sub-wind power clusters, then Adjw i,j is the adjacency matrix of the heterogeneous graph.

[0144] For the proposed isomorphic and heteromorphic graphs, determining the spatial connectivity of nodes under multiple modes is key to HHGAT's learning of spatiotemporal relationships. Simultaneously, the dynamics of the graph structure require more attention, rather than being highly dependent on predefined structures, especially for intermittent wind power generation. Since the spatiotemporal dependencies between different sites evolve over time, in order to enable the constructed graph to dynamically capture these time-varying relationships, in some implementations, the adjacency matrix is ​​updated using the actual update frequency of the NWP (daily update) to adapt to dynamic and practical engineering applications. Finally, the expressions for the adjacency matrix are given below (Formulas (10) and (11)), where cluster a is the isomorphic graph of the a-th sub-wind power cluster, cluster ao is the heteromorphic graph with edge attribute ij between the a-th and o-th sub-wind power clusters, and I is the identity matrix.

[0145]

[0146] S600: Construct a dynamic graph attention network; wherein, the dynamic graph attention network includes a homogeneous module and a heterogeneous module. The homogeneous module is based on the homogeneous graph of the wind power cluster and calculates the spatiotemporal feature vector of each wind farm according to the wind speed data of each wind farm in the sub-wind power cluster. The heterogeneous module outputs the predicted power of each wind farm based on the heterogeneous graph of the wind power cluster and the spatiotemporal feature vector of each wind farm.

[0147] In this embodiment, the weight change process between wind farms s at different times is fitted by homogeneous and heterogeneous modules, thereby fully extracting the interaction relationship between spatiotemporal relationship nodes.

[0148] In an exemplary embodiment, the structure of the dynamic graph attention network and its data processing flow are illustrated as follows: Figure 5 As shown.

[0149] Homogeneous modules extract local correlation features. To construct the spatiotemporal relationships within sub-wind power clusters, a Dynamic Graph Attention Network (DGAT) is used to compute local correlation features for each sub-cluster based on the constructed homogeneous graph. The specific process can be described as follows:

[0150]

[0151] e(xCi ,x Cj ) = a T [Wx Ci P Wx Cj j∈N Ci (13)

[0152] In the formula, e(x) Ci ,x Cj ) represents the attention coefficient of node j after nonlinear transformation of node i in the sub-wind power cluster C; LeakyReLU(·) is a nonlinear transformation function; a T (·) represents the attention coefficient; P is the feature concatenation operation; x Ci and x Cj These are the input feature vectors of node i and node j, respectively (obtained by transforming the wind speed data of node i and node j), and W is the shared parameter matrix of the linear transformation; N Ci γ is the total number of neighboring nodes of node i; γ is the positive slope of the negative region of the function.

[0153] Then, the obtained attention coefficients are normalized to obtain the attention weight α. ij The output vector is obtained by weighted summation.

[0154]

[0155]

[0156] In the formula, softmax j (·) is the attention coefficient e of the normalized neighbor node j. Cij The function is σ(·), which is the sigmoid activation function; K is the number of heads in the multi-head attention mechanism, and k is the head index of the multi-head attention mechanism; exp represents the empirical function; α Cij Indicates attention weight; N represents a spacetime vector; i This represents the total number of neighboring nodes of node i.

[0157] Finally, through f CAM (·) Unified extraction of spatial and temporal features of wind farm clusters. The key node features of the original wind farm s are obtained by performing average pooling and max pooling on the features of the wind farm s. After MLP and sigmoid activation operations, the feature vector set f representing the original wind farm s is obtained. CAM (x Ci ), and compare it with the spatiotemporal vector extracted by DGAT. The fusion process is performed to obtain the spatiotemporal feature vector of each wind farm through isomorphic modules. The calculation process is as follows:

[0158]

[0159] In the formula, f CAM (·) indicates the unified extraction of spatial and temporal features of wind power clusters, f MLP Let AvgPool represent the nonlinear transformation function guided by the multilayer perceptron, and MaxPool represent the average pooling layer and the max pooling layer, respectively. This represents the feature vector obtained after average pooling. f represents the feature vector obtained after max pooling. CAM (x Ci ) represents the eigenvector set of the original wind farm, and e represents the Hama operator. This represents the spatiotemporal feature vector.

[0160] Heterogeneous modules extract global correlation features. Based on the constructed heterogeneous graph, the spatiotemporal relationships within the wind power cluster are built using the heterogeneous module HDGAT. The attention mechanism computation process of the heterogeneous module HDGAT consists of node-level attention mechanism and type-level attention mechanism.

[0161] The calculation process for node-level attention is as follows:

[0162] After processing different sub-clusters of wind farms into the same feature space, a self-attention mechanism is used to learn weights for different types of nodes. Based on the constructed heterogeneous graph, given a node pair (i,j) and the meta-path Φ connecting them, the node-level attention is calculated as follows:

[0163]

[0164] In the formula, It is a transformation matrix that processes the i-th sub-wind power cluster into the same feature space; Φ i It is the i-th meta-path; It is an asymmetric node attention coefficient; yes; Yes; T is the matrix transpose; yes.

[0165] Normalization is performed using softmax(·) to obtain the neighbor weight parameters based on the metapath. And the meta-path Φ through node i is calculated based on the neighbor weight parameters. i The output feature vector is calculated as shown in formulas (20) and (21).

[0166]

[0167] In the formula, This is a node-level attention vector; Is node i traversed via metapath Φi The output feature vector; For the metapath Φ i The total number of wind farms under; For the metapath Φ i The number of adjacent wind farms of wind farm i.

[0168] Node-level attention is used to capture the importance of wind farms with different output characteristics and reduce the weight of noisy nodes.

[0169] The calculation process for type-level attention is as follows:

[0170] Type-level attention aggregation aggregates the attention vectors of the C groups of nodes obtained from the previous step based on multiple meta-paths. As input, a nonlinear function is used to transform the node embeddings of different output feature subclusters. These edges are then aggregated according to their weights to obtain the final representation x. pre .

[0171]

[0172] In the formula, W L A is the weight matrix; A is a biased vector; q is the type-level attention vector. It is a type-level attention vector; i is the i-th wind farm; Φ i Let be the i-th metapath; C is the total number of wind farms in the sub-cluster; tanh is a nonlinear function; b is a weighting coefficient; These are type-level attention weights; x pre P is the predicted power of the wind farm; P is the number of wind farms under the current metapath. It is the feature vector output at the node level.

[0173] Type-level attention learns the weights of neighboring subclusters of different output feature types and weights the feature vectors accordingly.

[0174] S700: Construct a second dataset and use it to train a dynamic graph attention network to achieve short-term power prediction for wind power clusters; the second dataset includes wind speed data and wind power data for each wind farm.

[0175] In this embodiment, the data included in the second dataset is categorized into NWP data. And wind power data p=[p1,p2,...,p n ] TIn this model, n represents the number of samples, and q represents the number of NWP meteorological elements used. Based on engineering experience, wind speed, wind direction, temperature, and pressure are commonly used as explanatory variables in wind power forecasting, directly inputting them into the statistical model. p, on the other hand, is the explained variable in the statistical model, representing the forecast target. Specifically, the forecast wind speed in X will be used below as a key meteorological element for classifying wind processes, denoted as v. The dataset D = (X, p), consisting of the explanatory variable X and the corresponding explained variable p, is divided into a training set and a test set according to a certain ratio. The training set data (historical data) is used to fit the model parameters, while the test set data serves as independent data to verify the model's predictive performance.

[0176] During training, both the homogeneous attention vector α and the heterogeneous attention vector q can be learned through gradient backpropagation. To date, spatiotemporal relationship modeling has integrated the spatiotemporal features of the entire large-scale wind power cluster, representing the future state of all nodes within the cluster.

[0177] The method proposed in this application can quickly and effectively identify and match wind processes, and perform targeted integrated optimization on different identified wind processes, thereby improving the accuracy of wind power prediction. To demonstrate that the method can achieve the above effects, this embodiment tests the proposed method with five existing models in a wind power cluster. The five models are as follows:

[0178] Model 1: The advanced model Transformer with a large number of parameters.

[0179] Model 2: G-GAT based on building an adjacency matrix according to geographic location.

[0180] Model 3: CBAM-LSTM based on convolutional module attention mechanism.

[0181] Model 4: TCN-LSTM based on temporal convolution.

[0182] Model 5: Linear Model LASSO.

[0183] Table 1. Wind Farm Prediction Indicators

[0184] Table 1 shows the prediction results for three wind farms. The prediction model provides prediction results for 96 points (at a 15-minute resolution) from 00:00 to 23:45 the following day at 8:00 each day. Three differentiated models (Models 1-5) are provided for each wind farm, and the "Proposed Optimized Model" represents the wind power prediction results based on the method presented in this paper. The results in Table 1 show that the proposed optimized model significantly improves prediction accuracy. To illustrate the effect more clearly, Figure 6Time series plots comparing wind farm prediction results are presented. It can be seen that the method in this paper effectively extracts the spatial characteristics of wind farms to a certain extent and applies them to predict power.

[0185] This application also provides a short-term power prediction device for wind power clusters, such as... Figure 7 As shown, the short-term power forecasting device for this wind power cluster includes:

[0186] Data acquisition module 701 is configured to acquire wind speed data from each wind farm within the wind power cluster, forming...

[0187] A first dataset is generated, and the first dataset is clustered to obtain multiple sub-wind power clusters; wherein, the first dataset is represented as X = {x1, x2, ..., x...} n}, x1, x2 and x n These represent the wind speed data for the first, second, and nth wind farms, respectively, and the sub-wind power cluster includes at least one wind farm.

[0188] The correlation calculation module 702 is configured to determine the short-term correlation between wind farms based on the distance between them.

[0189] The isomorphic graph construction module 703 is configured to treat all wind farms in the same sub-wind power cluster as nodes of the same output type, and use the spatiotemporal relationship of nodes of the same output type as the isomorphic spatiotemporal relationship to construct the isomorphic graph of the wind power cluster.

[0190] The heterogeneous graph construction module 704 is configured to treat wind farms in different sub-wind power clusters as nodes of different output types, and use the spatiotemporal relationship of nodes of different output types as heterogeneous spatiotemporal relationship to construct a heterogeneous graph of the wind power cluster.

[0191] The adjacency matrix construction module 705 is configured to determine the adjacency matrices of the isomorphic and heteromorphic graphs of the wind farm cluster based on the short-term correlation between wind farms.

[0192] Graph attention network module 706 is configured to construct a dynamic graph attention network; wherein, the dynamic graph attention network includes a homogeneous module and a heterogeneous module, the homogeneous module calculates the spatiotemporal feature vector of each wind farm based on the homogeneous graph of the wind power cluster and the wind speed data of each wind farm in the sub-wind power cluster, and the heterogeneous module outputs the predicted power of each wind farm based on the heterogeneous graph of the wind power cluster and the spatiotemporal feature vector of each wind farm;

[0193] The network training module 707 is configured to construct a second dataset and use the second dataset to train the dynamic graph attention network, so as to realize short-term power prediction of wind power clusters; wherein, the second dataset includes wind speed data and wind power data of each wind farm.

[0194] This application provides an electronic device. The electronic device may include a processor and a memory, wherein the processor and the memory can communicate; exemplarily, the processor and the memory communicate via a communication bus.

[0195] The processor executes computer execution instructions stored in memory, causing the processor to perform the scheme in the above embodiments. The processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0196] The communication bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The system bus can be divided into address bus, data bus, control bus, etc. Transceivers are used to enable communication between database access devices and other computers (e.g., clients, read-write libraries, and read-only libraries). Memory may include random access memory (RAM) and may also include non-volatile memory.

[0197] The electronic device provided in this application embodiment can be the terminal device described in the above embodiments.

[0198] This application also provides a computer-readable storage medium storing computer instructions. When the computer instructions are executed on a computer, the computer performs the technical solution of the wind power cluster short-term power prediction method described in the above embodiments.

[0199] This application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium. When the at least one processor executes the computer program, it can implement the technical solution of the short-term power prediction method for wind power clusters in the above embodiments.

[0200] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0201] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to implement the solution of this embodiment according to actual needs.

[0202] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The unit composed of the above modules can be implemented in hardware or in the form of hardware plus software functional units.

[0203] The integrated modules described above, implemented as software functional modules, can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods of the various embodiments of this application.

[0204] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0205] The memory may include high-speed RAM, and may also include non-volatile storage (NVM), such as at least one disk storage device, and may also be a USB flash drive, external hard drive, read-only memory, disk or optical disc, etc.

[0206] Buses can be Industry Standard Architecture (ISA) buses, Peripheral Component Interconnect (PCI) buses, or Extended Industry Standard Architecture (EISA) buses, etc. Buses can be categorized into address buses, data buses, control buses, etc.

[0207] The aforementioned storage medium can be implemented from any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The storage medium can be any available medium accessible to general-purpose or special-purpose computers.

[0208] An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Alternatively, the storage medium can be an integral part of the processor. The processor and storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and storage medium can exist as discrete components in an electronic control unit or main control device.

[0209] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0210] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for short-term power prediction of wind power clusters, characterized in that, The method includes: Wind speed data from each wind farm within the wind power cluster are acquired to form a first dataset. This first dataset is then clustered to obtain multiple sub-wind power clusters. The first dataset is represented as X = {x1, x2, ..., x...}. n }, x1, x2 and x n These represent the wind speed data for the first, second, and nth wind farms, respectively, and the sub-wind power cluster includes at least one wind farm. The short-term correlation between wind farms is determined based on the distance between them; All wind farms in the same sub-wind power cluster are treated as nodes of the same output type, and the spatiotemporal relationship of nodes of the same output type is taken as the isomorphic spatiotemporal relationship to construct the isomorphic graph of the wind power cluster. Wind farms in different sub-wind power clusters are treated as nodes of different output types, and the spatiotemporal relationships of nodes of different output types are used as heterogeneous spatiotemporal relationships to construct a heterogeneous graph of the wind power cluster. Based on the short-term correlation between wind farms, the adjacency matrices of the isomorphic and heteromorphic graphs of wind power clusters are determined. A dynamic graph attention network is constructed. The dynamic graph attention network includes a homogeneous module and a heterogeneous module. The homogeneous module calculates the spatiotemporal feature vector of each wind farm based on the homogeneous graph of the wind power cluster and the wind speed data of each wind farm in each sub-wind power cluster. The heterogeneous module outputs the predicted power of each wind farm based on the heterogeneous graph of the wind power cluster and the spatiotemporal feature vector of each wind farm. A second dataset is constructed, and the dynamic graph attention network is trained using the second dataset to achieve short-term power prediction of wind power clusters; wherein, the second dataset includes wind speed data and wind power data of each wind farm.

2. The short-term power prediction method for wind power clusters according to claim 1, characterized in that, Clustering the first dataset yields multiple sub-wind power clusters, including: Set k initial centroids; The Euclidean distances between the wind farm and each centroid are calculated using the following formula: In the formula, d(x) i ,c m () represents the wind speed data x of the i-th wind farm. i With the m-th centroid c m The Euclidean distance of the wind speed data, where t represents the time point and x it c represents the wind speed data of the i-th wind farm at time t. mt c represents the m-th centroid. m Wind speed data at time point t; Based on the Euclidean distances between the wind farm and each centroid, the wind farm is allocated using the following formula: In the formula, S m Let m be the m-th cluster, and arg min represent taking the minimum value; The centroid is updated using the following formula: The process of repeatedly assigning and updating centroids continues until a set condition is met, which is expressed as: or t'≥T' (5) In the formula, max represents the maximum value function. ε This indicates the set threshold. This represents the wind speed data of the m-th centroid at the (t'+1)-th and t'-th iterations, where t' represents the current iteration number and T' represents the maximum iteration number.

3. The short-term power prediction method for wind power clusters according to claim 1, characterized in that, Based on the distance between wind farms, determine the short-term correlation between them, including: The Euclidean distance between two wind farms is calculated using the following formula: In the formula, d(W1,W2) represents the Euclidean distance between two wind farms W1 and W2, and v 1t and v 2t This represents the wind speed data of two wind farms, W1 and W2, at time t. Based on the Euclidean distance between the two wind farms, normalization is performed using the following formula: In the formula, d nom d represents the normalized distance value. max This represents the maximum value of the Euclidean distance between all wind farms; Based on the set distance threshold, the STC relationship between two wind farms is determined using the following formula: In the formula, STC(W1,W2) represents the short-term correlation between wind farms W1 and W2. STC(W1,W2) = 1 indicates that wind farms W1 and W2 have an STC relationship, and STC(W1,W2) = 0 indicates that wind farms W1 and W2 do not have an STC relationship.

4. The short-term power prediction method for wind power clusters according to claim 1, characterized in that, The isomorphic graph of the wind power cluster is represented by G. Ho ={G Ho,1 ,...,G Ho,c ,...,G Ho,C }, where G Ho,1 G Ho,c and G Ho,C Let G represent the isomorphic graphs of the first, c-th, and C-th sub-wind power clusters, respectively. Ho,c = (V, E), where V and E are the sets of nodes and edges in the isomorphic graph of the c-th sub-wind power cluster, respectively, and V is described as {V1, V2, ..., V...}. i }, V1, V2 and V i These represent the first, second, and i-th nodes in the same sub-wind power cluster, respectively. E ij C represents the directed edge relationship between node i and node j in the same sub-wind power cluster. n This represents the set of nodes in a specific sub-wind power cluster. The heterogeneous graph of the wind power cluster is represented as G. He =(Φ C ,V',E'),Φ C Let V' and E' represent the meta-path, where V' and E' are the sets of nodes and edges in the heterogeneous graph, respectively, and V' = {V'1, ..., V'}. c ,...,V' C }, where V'1 and V' c and V' C Let represent the node sets within the first, c-th, and C-th sub-wind power clusters, respectively. Where E' ij C represents the directed edge relationship between node i and node j within two different sub-wind power clusters. i and C j This represents the set of nodes within two different sub-wind power clusters.

5. The short-term power prediction method for wind power clusters according to claim 1 or 3, characterized in that, Based on the short-term correlation between wind farms, the adjacency matrices of the isomorphic and heteromorphic graphs of wind farm clusters are determined using the following formula: Adjw i,j =STC i,j i=1,...,n;j=1,...,n-1 (9) In the formula, Adjw i,j STC represents the adjacency matrix of nodes i and j. i,j This represents the short-term correlation between node i and node j, where n is the number of nodes or wind farms. If node i and node j are located in the same sub-wind farm cluster, then Adjw i,j Given an adjacency matrix of an isomorphic graph, if node i and node j are located in different sub-wind power clusters, then Adjw i,j is the adjacency matrix of the heterogeneous graph.

6. The short-term power prediction method for wind power clusters according to claim 1, characterized in that, The isomorphic module calculates the spatiotemporal feature vector of each wind farm based on the isomorphic graph of the wind power cluster and the wind speed data of each sub-wind power cluster. include: The attention coefficient for each sub-cluster is calculated using the following formula: e(x Ci ,x Cj )=a T [Wx Ci PWx Cj ]j∈N Ci (13) In the formula, e(x) Ci ,x Cj ) represents the attention coefficient of node j after nonlinear transformation of node i in the sub-wind power cluster C; LeakyReLU(·) is a nonlinear transformation function; a T (·) represents the attention coefficient; P is the feature concatenation operation; x Ci and x Cj are the input feature vectors of nodes i and j, respectively, and W is the shared parameter matrix of the linear transformation; N Ci γ is the total number of neighboring nodes of node i; γ is the positive slope of the negative region of the function; The attention coefficients are normalized to obtain attention weights, and then a spatiotemporal vector is obtained by weighted summation. The calculation process is as follows: In the formula, softmax j (·) is the attention coefficient e of the normalized neighbor node j. Cij The function is σ(·), which is the sigmoid activation function; K is the number of heads in the multi-head attention mechanism, and k is the head index of the multi-head attention mechanism; exp represents the empirical function; α Cij Indicates attention weight; N represents a spacetime vector; Ci It is the total number of neighboring nodes of node i; Based on the spatiotemporal vector, the spatiotemporal feature vector is calculated using the following formula: In the formula, f CAM (·) indicates the unified extraction of spatial and temporal features of wind power clusters, f MLP Let AvgPool represent the nonlinear transformation function guided by the multilayer perceptron, and MaxPool represent the average pooling layer and the max pooling layer, respectively. This represents the feature vector obtained after average pooling. f represents the feature vector obtained after max pooling. CAM (x Ci Let ) represent the eigenvector set of the original wind farm, and e represent the Hama operator. This represents the spatiotemporal feature vector.

7. The short-term power prediction method for wind power clusters according to claim 6, characterized in that, The heterogeneous module outputs the predicted power of each wind farm based on the heterogeneous graph of the wind power cluster and the spatiotemporal feature vector of each wind farm in the following ways: Based on the heterogeneous graph of the wind power cluster, a node pair (i,j) and the meta-path Φ connecting the node pair are obtained, and the asymmetric node attention coefficient is calculated using the following formula: In the formula, It is a transformation matrix that processes the i-th sub-wind power cluster into the same feature space; Φ i It is the i-th meta-path; It is an asymmetric node attention coefficient; It is the metapath Φ i The i-th wind farm; It is the metapath Φ i The j-th wind farm; T is the matrix transpose; It is the metapath Φ i The attention vector below; Normalization using softmax(·) yields the node-level attention vector based on the meta-path, calculated as follows: In the formula, This is a node-level attention vector; Is node i traversed via metapath Φ i The output feature vector; For the metapath Φ i The total number of wind farms under; For the metapath Φ i The number of adjacent wind farms of wind farm i; Based on the node-level attention vector, the predicted power of the wind farm is calculated using the following formula: In the formula, W L A is the weight matrix; A is a biased vector; q is the type-level attention vector. It is a type-level attention vector; i is the i-th wind farm; Φ i Let be the i-th metapath; C is the total number of wind farms in the sub-cluster; tanh is a nonlinear function; b is a weighting coefficient; These are type-level attention weights; x pre P is the predicted power of the wind farm; P is the number of wind farms under the current metapath. It is the feature vector output at the node level.

8. A short-term power prediction device for wind power clusters, characterized in that, The device includes: The data acquisition module is configured to acquire wind speed data from each wind farm within the wind power cluster, forming a first dataset, and then cluster the first dataset to obtain multiple sub-wind power clusters; wherein, the first dataset is represented as X = {x1, x2, ..., x...} n }, x1, x2 and x n These represent the wind speed data for the first, second, and nth wind farms, respectively, and the sub-wind power cluster includes at least one wind farm. The correlation calculation module is configured to determine the short-term correlation between wind farms based on the distance between them. The isomorphic graph construction module is configured to treat all wind farms in the same sub-wind power cluster as nodes of the same output type, and use the spatiotemporal relationship of nodes of the same output type as the isomorphic spatiotemporal relationship to construct the isomorphic graph of the wind power cluster. The heterogeneous graph construction module is configured to treat wind farms in different sub-wind power clusters as nodes of different output types, and use the spatiotemporal relationship of nodes of different output types as heterogeneous spatiotemporal relationship to construct a heterogeneous graph of the wind power cluster. The adjacency matrix construction module is configured to determine the adjacency matrices of isomorphic and heteromorphic graphs of wind farm clusters based on the short-term correlation between wind farms. The graph attention network module is configured to construct a dynamic graph attention network. The dynamic graph attention network includes a homogeneous module and a heterogeneous module. The homogeneous module calculates the spatiotemporal feature vector of each wind farm based on the homogeneous graph of the wind power cluster and the wind speed data of each wind farm in each sub-wind power cluster. The heterogeneous module outputs the predicted power of each wind farm based on the heterogeneous graph of the wind power cluster and the spatiotemporal feature vector of each wind farm. The network training module is configured to construct a second dataset and use the second dataset to train the dynamic graph attention network, so as to realize short-term power prediction of wind power clusters; wherein, the second dataset includes wind speed data and wind power data of each wind farm.

9. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the short-term power prediction method for wind power clusters as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the short-term power prediction method for wind power clusters as described in any one of claims 1-7.

Images