Wind power cluster short-term power prediction method and system, electronic equipment and medium

Abstract

The invention discloses a wind power cluster short-term power prediction method and system, electronic equipment and a medium, and relates to the technical field of wind power prediction. The method comprises the following steps: constructing an improved wind angle field feature conversion method based on an improved Grubrum angle field; for the target wind power plant cluster, generating a WAF matrix by using an improved wind angle field feature conversion method; converting the WAF matrix into a scale space by using an SIFT (Scale Invariant Feature Transform) identification method, identifying continuous wind process mutation points and time sequence correlation points, and determining a space-time association relationship between the wind power plants; and constructing a graph structure of the target wind power plant cluster based on a dynamic graph attention network and the space-time association relationship, and performing dynamic prediction by using the graph structure and daily NWP meteorological information to obtain a power prediction result of the target wind power plant cluster. The short-term power prediction precision can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wind power prediction, in particular to a wind power cluster short-term power prediction method and system, an electronic device and a medium. BACKGROUND

[0002] Wind power generation essentially depends on natural wind resources. Meteorological factors such as wind speed and wind direction show high variability and uncertainty in time and space dimensions, resulting in strong random fluctuations and intermittent characteristics of wind power output. The large-scale integration of wind power has brought unprecedented challenges to the safe and stable operation of power systems, the economic efficiency of dispatching planning, and the orderly competition of power markets.

[0003] Under this background, achieving accurate and rapid short-term prediction of wind power is not only a key technical support for ensuring the grid accommodation capacity and improving the flexibility and reliability of system operation, but also a core prerequisite for optimizing dispatching decisions, reducing standby costs, promoting wind power consumption and participating in market transactions, especially for large-scale wind power clusters.

[0004] At present, various technical paths have been developed for the research of wind power prediction, which can be mainly summarized as follows: statistical methods are based on historical data and use time series analysis models for prediction. The advantage is that the model structure is relatively simple and the calculation is convenient. However, wind power series usually shows high nonlinearity and non-stationarity, and traditional statistical models are difficult to fully capture these complex dynamics.

[0005] Physical methods mainly rely on numerical weather prediction (NWP) systems, which simulate atmospheric physical processes to output meteorological parameters such as wind speed and direction at future time points, and then convert them into power prediction values combined with the physical characteristics of wind turbines. Physical methods can reflect the influence of meteorological conditions on wind power from a mechanistic perspective and have a certain theoretical basis, but the main bottleneck is that the NWP model itself requires a large amount of calculation and high computational resources, and the temporal and spatial resolution of the prediction results is relatively coarse, making it difficult to fully meet the needs of fine-grained operation of wind farms. Machine learning and deep learning methods, represented by support vector machine (SVM), artificial neural network (ANN), long short-term memory (LSTM), graph neural network (GNN), etc., have shown great potential in wind power prediction due to their strong non-linear mapping ability and the ability to learn complex patterns from massive data, and have become the mainstream direction of current research. These methods can effectively mine deep features and association rules hidden in historical data.

[0006] However, existing machine learning and deep learning methods still face several key challenges when applied to power prediction of large-scale wind farm clusters: First, most methods rely on pre-set, location-based static topological structures or use simplified similarity measures (such as Euclidean distance, Pearson correlation coefficient, etc.) when dealing with spatial correlations between wind farms. Such methods ignore the fact that the spatio-temporal correlation between wind farms changes dynamically with the evolution of upstream meteorological systems (such as the movement of wind belts and the passage of fronts). Static or quasi-static topological descriptions cannot truly capture the complex and time-varying interaction mechanisms within the wind farm cluster, limiting the model's understanding of the overall output characteristics of the cluster. Second, for large-scale wind farm clusters, the number of stations is large, and the interaction network between them is more complex. Simply stacking or independently modeling single-site prediction models cannot fully utilize the synergistic effects and information complementarity within the cluster, making it difficult to significantly improve prediction accuracy and model robustness. Third, traditional time series feature extraction methods may not fully leverage advanced concepts from other fields such as computer vision and image processing to reveal the underlying dynamic patterns of wind power output sequences and their potential associations with other station sequences from a deeper and more multi-dimensional perspective.

[0007] Therefore, there is an urgent need to develop a new short-term power prediction method. SUMMARY

[0008] The application aims to provide a wind power cluster short-term power prediction method, system, electronic device and medium, aiming to solve or improve at least one of the above technical problems.

[0009] To achieve the above-mentioned purpose, the application provides the following solutions.

[0010] A wind power cluster short-term power prediction method comprises the following steps:

[0011] An improved wind angle field feature conversion method is constructed based on an improved Gram angle field; the improved wind angle field feature conversion method generates a polar coordinate by processing time series values as radii, taking the sum of the normalized arccosine value of the wind speed sequence and the wind direction value as the angle;

[0012] For a target wind power cluster, an improved wind angle field feature conversion method is used to generate a WAF matrix;

[0013] The SIFT recognition method is used to convert the WAF matrix into a scale space, and to identify continuous wind process mutation points and time series correlation points, so as to determine the space-time correlation relationship between wind power plants;

[0014] A graph structure of the target wind power cluster is constructed based on a dynamic graph attention network and the space-time correlation relationship, and the graph structure and daily NWP meteorological information are used for dynamic prediction to obtain a power prediction result of the target wind power cluster.

[0015] Optionally, the improved wind angle field feature conversion method is used to generate a WAF matrix for a target wind power cluster, specifically comprising:

[0016] First, the wind speed sequence is normalized, and then the arccosine value of the wind speed sequence is calculated:

[0017]

[0018] Then, the arccosine value of the wind speed sequence is added to the wind direction value to obtain an angle:

[0019] cos((φ i,s +φ i,d )+(φ j,s +φ j,d ))=cos(φ i +φ j )

[0020] Then, the coordinates in the dispatching period are point-by-point inner product to obtain a WAF matrix capable of representing the wind power plant dispatching period:

[0021]

[0022] Wherein, x i,sx represents the normalized wind speed sequence of the i th wind farm to be predicted day, i,d φ represents the normalized wind direction sequence of the i th wind farm to be predicted day; i,s φ represents the inverse cosine value of the normalized wind speed sequence of the i th wind farm; i,d φ represents the conversion of the wind direction sequence of the i th wind farm into radians; j,s φ represents the inverse cosine value of the normalized wind speed sequence of the j th wind farm; j,d φ represents the conversion of the wind direction sequence of the j th wind farm into radians; i φ represents the angle of the i th wind farm; j φ represents the angle of the j th wind farm; t φ 1,..., φ N represent the radius in the polar coordinate system obtained by time series transformation; n φ represents the characteristic value of different time sequences; t represents the time sequence point number, N represents the total length of the constructed window, and n represents the n th time point position.

[0023] Optionally, the SIFT recognition method is used to convert the WAF matrix into a scale space, identify continuous wind process mutation points and time sequence correlation points, and determine the space-time correlation relationship between wind farms, specifically including:

[0024] Firstly, the WAF matrix is subjected to Gaussian blur processing, and the key points of the WAF matrix are detected through Gaussian difference calculation between different scales;

[0025] Then, the mutation point feature description vector is generated according to the gradient direction and amplitude distribution around the key points, and the Euclidean distance between each mutation point feature description vector is calculated, and the space-time correlation relationship between wind farms is determined based on the distance ratio test method of David Lowe.

[0026] Optionally, the space-time correlation relationship between wind farms is determined based on the distance ratio test method of David Lowe, specifically including:

[0027] Based on the distance ratio test method of David Lowe, for each mutation point feature description vector, the nearest two matching points to be matched in the WAF matrix are found, and the nearest neighbor distance and the next nearest neighbor distance are calculated, and if the ratio of the nearest neighbor distance to the next nearest neighbor distance is less than a set threshold, it is considered that the two wind farms have a space-time correlation relationship.

[0028] Optionally, the graph structure of the target wind farm cluster is constructed based on the dynamic graph attention network and the space-time correlation relationship, specifically including:

[0029] An initial graph structure and an adjacency matrix are constructed according to the spatio-temporal correlation relationship; the initial graph structure is represented as: G=(V, E), wherein V and E are a node set and an edge set, respectively; and the adjacency matrix is used to describe the relationship between any two connected nodes.

[0030] Based on the initial graph structure and the adjacency matrix, a dynamic graph attention network is used to dynamically calculate the relevant interaction coefficients between nodes, so as to obtain the graph structure of the target wind farm cluster.

[0031] The application further provides a wind farm cluster short-term power prediction system, comprising:

[0032] An algorithm construction unit is configured to construct an improved wind angle field feature conversion method based on an improved Gram angle field; the improved wind angle field feature conversion method processes time series values as radii, generates polar coordinates by taking the sum of the normalized arccosine values of the wind speed sequence and the wind direction values as the angle, and generates polar coordinates.

[0033] A matrix generation unit is configured to generate a WAF matrix for a target wind farm cluster by using the improved wind angle field feature conversion method.

[0034] A spatio-temporal correlation relationship determination unit is configured to convert the WAF matrix into a scale space by using a SIFT recognition method, recognize continuous wind process mutation points and time series correlation points, and determine the spatio-temporal correlation relationship between wind farms.

[0035] A power prediction unit is configured to construct a graph structure of the target wind farm cluster based on a dynamic graph attention network and the spatio-temporal correlation relationship, and perform dynamic prediction by using the graph structure and daily NWP meteorological information, so as to obtain a power prediction result of the target wind farm cluster.

[0036] The application further provides an electronic device comprising a memory and a processor, wherein the memory is configured to store a computer program, and the processor is configured to run the computer program to enable the electronic device to perform the wind farm cluster short-term power prediction method according to the above.

[0037] The application further provides a computer readable storage medium storing a computer program, wherein the computer program is executed by a processor to implement the wind farm cluster short-term power prediction method as described above.

[0038] According to the embodiments of the application, the following technical effects are achieved:

[0039] The application discloses a wind power cluster short-term power prediction method and system, electronic equipment and medium, the method comprises the following steps: constructing an improved wind angle field feature conversion method based on an improved Gram angle field; for a target wind power cluster, an improved wind angle field feature conversion method is used to generate a WAF matrix; the WAF matrix is converted into a scale space by using a SIFT identification method, and continuous wind process mutation points and time sequence correlation points are identified to determine the spatio-temporal correlation between wind power clusters; a graph structure of the target wind power cluster is constructed based on a dynamic graph attention network and the spatio-temporal correlation, and dynamic prediction is performed by using the graph structure and daily NWP meteorological information to obtain the power prediction result of the target wind power cluster. The application can improve the short-term power prediction accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings described below are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0041] Figure 1 The flowchart of the wind power cluster short-term power prediction method of the present application;

[0042] Figure 2 The correlation discrimination process between WAF matrices in the present embodiment;

[0043] Figure 3 The WAF graph of each wind power plant in the wind power cluster in the present embodiment for a day to be predicted;

[0044] Figure 4 The prediction power and actual power time series graph of the wind power cluster in the present embodiment. DETAILED DESCRIPTION

[0045] The technical solutions in the embodiments of the present application will be described in detail below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0046] The present application aims to provide a wind power cluster short-term power prediction method, system, electronic equipment and medium, which aims to solve or improve at least one of the above technical problems.

[0047] In order to make the above-mentioned purposes, features and advantages of the present application more apparent, further detailed description will be given below in combination with the drawings and specific embodiments.

[0048] As shown in Figure 1 The present application provides a wind power cluster short-term power prediction method, comprising:

[0049] Step 100: An improved wind angle field feature conversion method is constructed based on an improved Gram angle field; the improved wind angle field feature conversion method generates polar coordinates by processing time series values as radii, taking the sum of the normalized inverse cosine value of the wind speed sequence and the wind direction value as the angle.

[0050] Step 200: For the target wind farm cluster, the improved wind angle field feature conversion method is used to generate a WAF matrix.

[0051] Step 300: The WAF matrix is converted into a scale space using the SIFT recognition method, and continuous wind process mutation points and time series correlation points are identified to determine the spatio-temporal correlation relationship between wind farms.

[0052] Step 400: Based on the dynamic graph attention network and the spatio-temporal correlation relationship, a graph structure of the target wind farm cluster is constructed, and dynamic prediction is performed using the graph structure and daily NWP meteorological information to obtain the power prediction result of the target wind farm cluster.

[0053] As a specific embodiment, the present embodiment constructs a multi-feature conversion method to generate a wind angle field feature matrix (wind angle field, WAF) of each wind farm. In combination with the wind speed, wind direction and time gradient of the wind farm to be measured on the day, the SIFT recognition method in the image recognition field is used to identify whether there is a spatio-temporal correlation relationship between wind farms, and a graph structure is constructed accordingly. Specifically, the following steps are included:

[0054] (1) WAF matrix construction of wind farm:

[0055] In order to more accurately capture the spatio-temporal correlation between wind farms, the present application proposes a wind angle field feature conversion method based on an improved Gram angle field. The method generates polar coordinates by processing time series values as radii, taking the sum of the normalized inverse cosine value of the wind speed sequence and the wind direction value as the angle. Specifically, first, the wind speed sequence is normalized, formula (1) is the normalization formula, then the inverse cosine value is calculated, and then the angle is obtained by adding the wind direction value. Then, the coordinates in the dispatching period are pointwise inner product to obtain the wind angle field that can represent the dispatching period of the wind farm. This process not only considers the amplitude information of the time series, but also introduces the gradient relationship of the time series, thereby more comprehensively describing the spatio-temporal features of the wind field. The specific process is as follows:

[0056]

[0057]

[0058] cos((φ i,s +φ i,d )+(φ j,s +φ j,d ))=cos(φ i +φ j )(3)

[0059]

[0060] wherein the input is the normalized wind speed sequence x i,s and the wind direction sequence x i,d of the i th wind farm to be predicted on the day; φ i,d is the inverse cosine value of the normalized wind speed sequence x i,s of the i th wind farm; φ i,d is the wind direction sequence of the i th wind farm converted into radians. φ i and φ j are feature values at different time sequences; r n is the radius in the polar coordinate system obtained through time sequence transformation.

[0061] (2) According to the spatial and temporal correlation relationship between WAFs to distinguish

[0062] In order to identify the state mutation point in the image matrix, the SIFT method is introduced to convert the WAF matrix into a scale space to identify the continuous wind process mutation point and the time sequence correlation point. The specific process is shown in Figure 2 . First, the WAF is subjected to Gaussian blur processing to obtain the key points of the WAF matrix; the representation form of the Gaussian blur matrix is as follows:

[0063] L(x,y,σ)=G(x,y,σ)*I(x,y)(5)

[0064]

[0065] wherein L(x,y,σ) is the matrix converted into a scale space. G(x,y,σ) is a Gaussian kernel with a scale of σ, I(x,y) is a unit matrix, * represents convolution operation, and o and u are the horizontal and vertical coordinates of the center of the Gaussian kernel, respectively. Then, the key points are detected through Gaussian difference calculation between different scales.

[0066] D(x,y,σ)=[G(x,y,kσ)-G(x,y,σ)]*I(x,y)(7)

[0067] where k is a scale change factor. This process can effectively detect stable feature points in the WAF matrix. After determining the key points, a feature descriptor is generated according to the gradient direction and amplitude distribution around the key points. The calculation formula of the gradient direction and amplitude is formula (8)-formula (9).

[0068]

[0069]

[0070] where f 1,i and f 2,i are the i-th elements of the two descriptors, respectively. Generally speaking, the empirical value 0.75 is a commonly used threshold in the distance ratio test method of David Lowe. If the ratio of the nearest neighbor distance to the next nearest neighbor distance should be less than 0.75, it is considered that there is a spatiotemporal correlation relationship between WF i and WF j .

[0071] (3) Dynamic graph attention network

[0072] In order to fine-grained model the spatiotemporal relationship of the wind farm, the application introduces a dynamic graph attention network to calculate the action coefficient between the related wind farms. The dynamic graph attention network (DGAT) can fully utilize the prior spatiotemporal dependent information to construct the spatiotemporal relationship between nodes, so as to form the spatiotemporal relationship topology of the wind power cluster, and calculate the influence weight between the wind farms to construct the nonlinear relationship between the input features and the output. By virtue of the spatiotemporal correlation relationship between the wind farms identified in the above, a graph structure G=(V,E) of the wind power cluster is constructed, where V and E are node sets and edge sets, respectively, and the form of the adjacency matrix Adj is Adj∈R N×N , which is used to describe the relationship between any two connected nodes. If the wind farm i and the wind farm j have a spatiotemporal correlation relationship, Adjw i,j =1, otherwise Adjw i,j =0 indicates that there is no connection between them. The graph structure is dynamically updated according to the daily updated NWP meteorological information.

[0073] Adjw i,j =STC i,j i=1,...,n;j=1,...,n-1(12)

[0074] where STC i,j represents the spatiotemporal relationship matrix between the wind farms.

[0075] To fully extract the spatio-temporal interaction between nodes, the embodiment adopts a dynamic graph attention network to dynamically calculate the relevant interaction coefficients between nodes. The specific process is as follows:

[0076]

[0077] e(x Ci ,x Cj )=a T [Wx Ci P Wx Cj ]j∈N Ci (14)

[0078]

[0079]

[0080] where e(x Ci ,x Cj ) represents the nonlinear transformation attention coefficient of node j (the adjacent node of node i) to node i in the wind power cluster; LeakyReLU(·) is a nonlinear transformation function; a T (·) is an attention coefficient; P is a feature splicing operation; x Ci and x Cj are input feature vectors of node i and node j, respectively. W is a shared parameter matrix of linear transformation; N Ci is the total number of neighbor nodes of node i; γ is the positive slope of the negative region of the function; then the attention weight α Cij is obtained by normalizing the obtained attention coefficient. The output vector is obtained by weighted sum of the multi-head single field power obtained after the feature vector is fitted by a nonlinear function. softmax(·) is a normalization function; softmax j (·) is a function of normalizing the attention coefficient of adjacent node j; σ(·) is a sigmoid activation function; the number of heads of the multi-head attention mechanism is K. e Cij represents e(x Ci ,x Cj ) is the attention coefficient between wind farms i and j in the cluster, and N i represents the total number of wind farms in the i-th wind power cluster.

[0081] Therefore, the application can not only realize the fast and effective matching of wind farms with spatio-temporal correlation in the wind power cluster, but also can guide the information exchange between multi-channel wind farms according to the prior constructed graph structure. As an example analysis result, Table 1 shows the example result of a wind power cluster in a certain province in Northeast China. The prediction model gives the prediction result of 96 points (with 15-minute resolution) from 00:00 to 23:45 of the next day at 8:00 every day. Among them, four differentiated models, i.e., models 1-4, are given, and the “proposed optimization model” is the result of the short-term power prediction of the wind power cluster based on the graph feature relationship matching. According to the result in the table, the proposed optimization model can significantly improve the prediction accuracy. To visually show the effect, Figure 3 The WAF of the test set of each wind farm in the wind power cluster for the specified prediction day is shown, Figure 4 The time series graph corresponding to the wind power cluster is given, and the optimization model can utilize the information complement of the wind farms to optimize the prediction result.

[0082] Table 1: Prediction index of wind power cluster

[0083]

[0084] Each embodiment in the specification is described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts of each embodiment can be referred to each other.

[0085] The principles and implementation modes of the application are described by applying specific examples in this paper, and the above embodiment description is only used to help understand the core idea of the application; at the same time, for those skilled in the art, according to the idea of the application, the specific implementation mode and application range will be changed. In view of the above, the content of the specification should not be understood as a limitation of the application.

Claims

1. A method for short-term power prediction of a wind power cluster, characterized in that: include: An improved wind angle field feature conversion method is constructed based on the improved Gram angle field; The improved wind angle field feature conversion method generates polar coordinates by processing the time series value into a radius and taking the sum of the normalized arc cosine value of the wind speed sequence and the wind direction value as the angle; For the target wind farm cluster, the WAF matrix is ​​generated using the improved wind angle field feature conversion method; The WAF matrix is ​​converted into a scale space using the SIFT recognition method, and continuous wind process mutation points and time series related points are identified to determine the spatiotemporal correlation relationship between wind farms; A graph structure of the target wind farm cluster is constructed based on a dynamic graph attention network and the spatiotemporal association relationship, and a dynamic prediction is performed using the graph structure and daily NWP meteorological information to obtain a power prediction result of the target wind farm cluster.

2. The method for short-term power prediction of a wind power cluster according to claim 1, characterized in that: The WAF matrix is ​​generated by using an improved wind angle field feature conversion method for the target wind farm cluster, specifically including: First, normalize the wind speed sequence, and then calculate the arc cosine value of the wind speed sequence: Then add the arc cosine value of the wind speed sequence to the wind direction value to get the angle: cos((φ i,s +φ i,d )+(φ j,s +φ j,d ))=cos(φ i +φ j ) Then, the coordinates within the scheduling period are inner-producted point by point to obtain the WAF matrix that can characterize the wind farm scheduling period: Among them, x i,s represents the normalized wind speed sequence of the i-th wind farm on the forecast day, x i,d represents the normalized wind direction sequence of the i-th wind farm on the forecast day; φ i,s represents the inverse cosine value of the normalized wind speed sequence of the i-th wind farm; φ i,d Indicates converting the wind direction sequence of the i-th wind farm into radians; φ j,s represents the inverse cosine value of the normalized wind speed sequence of the j-th wind farm; φ j,d The wind direction sequence of the j-th wind farm is converted into radians; φ i represents the angle of the i-th wind farm; φ j represents the angle of the j-th wind farm; r t represents the radius in the polar coordinate system obtained by time series transformation; φ1,...,φ n Represents the characteristic values ​​of different time series; t represents the sequence number of the time series point; N represents the total length of the constructed window, and n represents the position of the nth time point.

3. The method for short-term power prediction of a wind power cluster according to claim 1, characterized in that: The WAF matrix is ​​converted into a scale space using the SIFT recognition method, and continuous wind process mutation points and time series related points are identified to determine the spatiotemporal correlation relationship between wind farms, specifically including: First, the WAF matrix is ​​Gaussian blurred, and the key points of the WAF matrix are detected by Gaussian difference calculation between different scales; Then, a mutation point feature description vector is generated according to the gradient direction and amplitude distribution around the key point, and the Euclidean distance between each mutation point feature description vector is calculated. The spatiotemporal correlation between wind farms is determined based on David Lowe's distance ratio test method.

4. The method for short-term power prediction of a wind power cluster according to claim 3, characterized in that: The distance ratio test method based on David Lowe's method for determining the spatiotemporal correlation between wind farms specifically includes: Based on David Lowe's distance ratio test method, for each mutation point feature description vector, the two nearest matching points to be matched in the WAF matrix are found, and the nearest neighbor distance and the next nearest neighbor distance are calculated. If the ratio of the nearest neighbor distance to the next nearest neighbor distance is less than a set threshold, it is considered that there is a temporal and spatial correlation relationship between the two wind farms.

5. The method for short-term power prediction of a wind power cluster according to claim 1, characterized in that: The step of constructing the graph structure of the target wind farm cluster based on the dynamic graph attention network and the spatiotemporal association relationship specifically includes: An initial graph structure and an adjacency matrix are constructed based on the spatiotemporal association relationship; the initial graph structure is represented as: G = (V, E), where V and E are a node set and an edge set, respectively; the adjacency matrix is ​​used to describe the relationship between any two connected nodes; Based on the initial graph structure and the adjacency matrix, a dynamic graph attention network is used to dynamically calculate the correlation interaction coefficients between the nodes to obtain the graph structure of the target wind farm cluster.

6. A short-term power prediction system for a wind power cluster, characterized in that: include: an algorithm construction unit, for constructing an improved wind angle field characteristic conversion method based on the improved Gram angle field; The improved wind angle field feature conversion method generates polar coordinates by processing the time series value into a radius and taking the sum of the normalized arc cosine value of the wind speed sequence and the wind direction value as the angle; A matrix generation unit is used to generate a WAF matrix for a target wind farm cluster using an improved wind angle field feature conversion method; a spatiotemporal correlation determination unit, configured to convert the WAF matrix into a scale space using a SIFT recognition method, identify continuous wind process mutation points and time series related points, and determine the spatiotemporal correlation between wind farms; A power prediction unit is used to construct a graph structure of the target wind farm cluster based on the dynamic graph attention network and the spatiotemporal correlation relationship, and to perform dynamic prediction using the graph structure and daily NWP meteorological information to obtain a power prediction result of the target wind farm cluster.

7. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the wind power cluster short-term power prediction method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that The computer program is stored therein, and when the computer program is executed by a processor, the method for short-term power prediction of a wind power cluster according to any one of claims 1 to 5 is implemented.

Images