Wind farm data reconstruction method based on graph attention mechanism and manifold autoencoder

Abstract

The application belongs to the technical field of data reconstruction, and particularly relates to a wind farm data reconstruction method based on a graph attention mechanism and a manifold autoencoder, which comprises the following steps: firstly, a graph structure of a wind farm site is constructed, the site is represented as a node, node features include wind speed, wind direction, temperature and power generation, and edges between nodes are defined based on geographical proximity or feature similarity; then, a graph attention encoder is used to dynamically allocate weights of neighbor nodes through an attention mechanism to generate a low-dimensional embedding representation of each site; a manifold constraint is introduced in the low-dimensional embedding space to ensure local neighborhood relationship and global geometric consistency; finally, the low-dimensional embedding is mapped back to a high-dimensional data space through a graph decoder to realize reconstruction of missing data. The total loss function proposed in the application combines data reconstruction error, manifold constraint loss and a regularization term, and model training is completed by optimizing the total loss function.

Description

Technical Field

[0001] This invention belongs to the field of data reconstruction technology, specifically relating to a wind farm data reconstruction method based on graph attention mechanism and manifold autoencoder. Background Technology

[0002] Wind energy is a crucial component of renewable energy, and the real-time operation of wind farms relies on accurate monitoring of site data such as wind speed, wind direction, temperature, and power generation. However, due to sensor malfunctions, data transmission problems, or other reasons, some sites may experience data gaps, affecting the operational efficiency and prediction accuracy of wind farms. Traditional data imputation methods (such as interpolation and statistical regression models) struggle to fully utilize the spatial correlations between multiple sites and the nonlinear characteristics of the data.

[0003] In recent years, Graph Neural Networks (GNNs) have been applied in multi-site collaborative prediction due to their ability to model graph-structured data. Meanwhile, Manifold Autoencoders (MAGs) have advantages in capturing the low-dimensional manifold properties of data. However, existing technologies have failed to effectively combine the advantages of both to dynamically construct weight relationships between nodes and address the data gap problem by utilizing the spatial and manifold properties of multi-site data. Summary of the Invention

[0004] To address the technical problem of missing wind farm site data in the prior art, this invention provides a wind farm data reconstruction method based on graph attention mechanism and manifold autoencoder, which fully utilizes the spatial correlation between sites and the data manifold structure to achieve high-precision filling of missing data.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0006] The wind farm data reconstruction method based on graph attention mechanism and manifold autoencoder includes the following steps:

[0007] S1, Graph Construction Module: Represents multiple sites of a wind farm as a graph. , where: node set This represents the sites in a wind farm, where each node's characteristics include wind speed, wind direction, and temperature; edge set. Represents the relationships between sites, defining edge weights based on geographical proximity or feature similarity;

[0008] S2, Graph Attention Encoder: Extracts site features using graph attention mechanism and generates low-dimensional embedding representations of nodes by dynamically assigning weights to neighboring nodes;

[0009] S3, Manifold Constraint Module: Applies local and global manifold constraints to nodes in a low-dimensional embedding space, preserving the geometric properties of the data;

[0010] S4, Graph Decoder: Maps low-dimensional embedding representations back to high-dimensional data space, reconstructs node features, including filling in missing values;

[0011] S5. Loss Function Optimization: Define the total loss function, including data reconstruction loss, local manifold constraint loss, global manifold constraint loss, and regularization loss;

[0012] S6. Model Training and Inference: Train the model by optimizing the loss function; use the trained model to reconstruct data with missing values ​​in the wind farm.

[0013] The edge set in the graph construction module of S1 The weights are defined based on both geographical proximity and feature similarity, where feature similarity is determined by calculating the Pearson correlation coefficient of the node feature vectors; by dynamically updating the adjacency matrix, the modeling of time series data is realized to adapt to the time-varying characteristics of the relationships between sites.

[0014] The weight calculation formula in S2 is as follows:

[0015]

[0016]

[0017] in, and They are nodes and nodes eigenvectors, For nodes For nodes Attention weights Represents a node The set of neighboring nodes; This represents the weight matrix, used for linear transformation of node features; This represents the attention weight vector, used to calculate the correlation between nodes; This indicates a feature concatenation operation, which concatenates two vectors along their dimensions. Represents a node For nodes The unnormalized attention score represents the correlation between the two. This represents the activation function. This represents an exponential function.

[0018] The S2 graph attention encoder uses a multi-head attention mechanism to calculate dynamic weights, as shown in the formula:

[0019]

[0020] in, For the number of attention heads, Indicates the first Attention weight based on size.

[0021] The formulas for the local and global manifold constraints in S3 are as follows:

[0022] Local manifold constraints:

[0023] Global manifold constraints:

[0024] in, and For nodes and nodes The low-dimensional embedding representation, This represents the actual distance between nodes; the local constraints in the manifold constraint module are jointly defined by the weight of the node's neighborhood and the low-dimensional embedding distance, prioritizing the preservation of the geometric relationships between nodes within the neighborhood.

[0025] The global geometric consistency constraint in the manifold constraint module of S3 is based on the actual distance of high-dimensional data. It is calculated using geodesic distance or Euclidean distance.

[0026] The S4 graph decoder uses a fully connected layer to map the low-dimensional embedding representation Z back to the high-dimensional feature space, and uses the activation function σ to ensure the nonlinear reconstruction capability of the data.

[0027] The method for reconstructing node features in S4 is as follows:

[0028]

[0029] in, For the reconstruction of high-dimensional data, This represents the activation function. Here is the weight matrix of the decoder. It is a low-dimensional embedding representation, generated by a graph encoder.

[0030] The total loss function in S5 is:

[0031]

[0032] in, , This represents the total loss function, used to measure the difference between the model's predictions and the actual data, as well as the degree to which the constraints are met. Represents a node The original high-dimensional feature vector, Represents a node Reconstructing high-dimensional feature vectors; , , For weight hyperparameters, These are the weighting coefficients for local constraints. These are the weighting coefficients for global constraints. The regularization strength; Indicates the data reconstruction loss. This represents the loss due to local manifold constraints. This represents the global manifold constraint loss.

[0033] The S6 model initializes model parameters through self-supervised learning during training, improving reconstruction performance on datasets with high missing rates.

[0034] Compared with the prior art, the beneficial effects of this invention are:

[0035] This invention provides a method for reconstructing multi-site wind farm data based on graph attention and manifold autoencoders. The graph attention mechanism adaptively adjusts the weights between sites to accurately capture spatial correlations. Combined with the constraints of manifold autoencoders, it ensures that low-dimensional embedding preserves the geometric properties of the data. It exhibits strong robustness to noise and extreme data loss. Compared to traditional algorithms, this invention significantly improves the accuracy and robustness of wind farm data imputation. Attached Figure Description

[0036] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0037] The structures, proportions, sizes, etc. illustrated in this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed herein, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0038] Figure 1 This is an overall model architecture diagram provided in Embodiment 1 of the present invention;

[0039] Figure 2 This is a schematic diagram of the construction of the wind farm site map structure provided in Embodiment 1 of the present invention;

[0040] Figure 3 This is an architecture diagram of the graph attention encoder provided in Embodiment 1 of the present invention;

[0041] Figure 4 This is a schematic diagram of low-dimensional embedded manifold constraints provided in Embodiment 1 of the present invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. These descriptions are only for further illustrating the features and advantages of the present invention, and not for limiting the claims of the present invention. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0043] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0044] Example 1: Wind speed data reconstruction at a single wind farm site

[0045] Background: A wind farm has 50 wind turbine sites, each recording wind speed data. However, due to sensor malfunctions, approximately 20% of the site data is randomly missing. This embodiment reconstructs the missing data using the method of the present invention. Figure 1 As shown, it includes the following steps:

[0046] Step 1: Data Preparation: The data source is historical wind speed data from wind farm sites, spanning 30 days, recorded every 10 minutes. Missing data will be filled with initial placeholders (such as zero values).

[0047] Step 2, Graph Construction: (e.g.) Figure 2 As shown, construct the graph , where: node set This represents 50 stations, where each node is characterized by its wind speed record. (Edge set) Based on geographical proximity, the distance between each station and its five nearest neighbors is calculated; the smaller the distance, the larger the edge weight. Node features are standardized so that their mean is 0 and their standard deviation is 1.

[0048] Step 3, Model Design: (e.g.) Figure 3 As shown, when designing the Graph Attention Encoder (GAT Encoder), a multi-head attention mechanism is used to calculate the dynamic weights between nodes. The low-dimensional embedding representation of each node is generated using the following formula:

[0049]

[0050] in, Indicates attention weights. These are learnable parameters. It's an activation function. Number of attention heads. .

[0051] like Figure 4 As shown, local and global manifold constraints are applied in the low-dimensional embedding space. The formula for the local manifold constraint is:

[0052]

[0053] The global manifold constraint formula is:

[0054]

[0055] When designing the graph decoder, the low-dimensional embedding representation ZZZ is mapped back to the high-dimensional space to reconstruct the missing data. The decoding formula is as follows:

[0056]

[0057] Step 5, Model Training: The optimization objective is the total loss function.

[0058]

[0059] The parameter settings are as follows: , , We used the Adam optimizer with a learning rate of 0.001 and 500 training iterations.

[0060] Step Six: Result Evaluation: The reconstruction effect is evaluated using mean squared error (MSE) and mean absolute error (MAE). Compared with traditional interpolation methods (such as linear interpolation), the method in this embodiment reduces MSE by 35% and MAE by 28%.

[0061] Example 2: Multi-site collaborative wind speed and power generation data filling

[0062] Background: A wind farm has recorded wind speed, wind direction, and power generation data at 100 stations. Due to extreme weather, wind speed and power generation data are missing at some stations, with a missing rate of 25%. This embodiment utilizes the method of the present invention for multi-feature collaborative data imputation, including the following steps:

[0063] Step 1: Data Preparation: The data source consists of historical records of wind speed, wind direction, and power generation from 100 stations. The feature vector for each node includes wind speed, wind direction, and power generation.

[0064] Step 2: Graph Construction: Constructing the graph , where: node set Represents 100 stations. Edge set. Based on the joint definition of geographical proximity and feature similarity, feature similarity is determined by calculating the Pearson correlation coefficient of the node feature vectors.

[0065] Step 3, Model Design: When designing the Graph Attention Encoder (GAT Encoder), dynamic weight relationships are constructed, and multi-feature collaborative modeling is utilized; when designing the manifold constraint module, wind speed and power generation features are simultaneously embedded into the low-dimensional manifold, preserving the correlation between features and between sites; during the decoding process, the missing wind speed and power generation data are reconstructed based on the collaborative relationships between multiple features.

[0066] Step 4: Model Training: Parameter settings and optimization are similar to those in Example 1, except that the joint loss weights of wind speed and power generation are added.

[0067] Step 5, Result Evaluation: The MSE and MAE of the reconstructed data were reduced by 40% and 32% respectively compared with the traditional regression method.

[0068] The program portion of a technology can be considered a "product" or "artifact" existing in the form of readable code and / or related data, and is involved in or implemented through a computer-readable medium. Tangible, permanent storage media can include memory or storage used by any computer, processor, or similar device or related module. For example, various semiconductor memories, tape drives, disk drives, or any similar device capable of providing storage functionality for software.

[0069] All software, or parts thereof, may sometimes communicate via networks, such as the Internet or other communication networks. Such communication can load software from one computer device or processor to another. For example, loading software from a server or host computer of a video object detection device to a hardware platform of a computer environment, or another computer environment that implements the system, or a system with similar functionality related to providing the information needed for object detection. Therefore, another medium capable of transmitting software elements can also be used as a physical connection between local devices, such as light waves, radio waves, electromagnetic waves, etc., propagated through cables, fiber optic cables, or air. Physical media used for carrier waves, such as cables, wireless connections, or fiber optic cables, can also be considered as media carrying software. In this context, unless limited to tangible "storage" media, the term "readable medium" for a computer or machine refers to the medium involved in the execution of any instructions by the processor.

[0070] Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are merely for the purpose of helping to understand the method and core ideas of this invention. Those skilled in the art should understand that the various modules or steps of this invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. This invention is not limited to any specific combination of hardware and software.

[0071] The above description only illustrates the preferred embodiments of the present invention. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention, and all such changes should be included within the protection scope of the present invention.

Claims

1. A wind farm data reconstruction method based on graph attention mechanism and manifold autoencoder, characterized in that, Includes the following steps: S1, Graph Construction Module: Represents multiple sites of a wind farm as a graph. , where: node set This represents the sites in a wind farm, where each node's characteristics include wind speed, wind direction, and temperature; edge set. Represents the relationships between sites, defining edge weights based on geographical proximity or feature similarity; S2, Graph Attention Encoder: Extracts site features using graph attention mechanism and generates low-dimensional embedding representations of nodes by dynamically assigning weights to neighboring nodes; The weight calculation formula in S2 is as follows: in, and They are nodes and nodes eigenvectors, For nodes For nodes Attention weights Represents a node The set of neighboring nodes; This represents the weight matrix, used for linear transformation of node features; This represents the attention weight vector, used to calculate the correlation between nodes; This indicates a feature concatenation operation, which concatenates two vectors along their dimensions. Represents a node For nodes The unnormalized attention score represents the correlation between the two. This represents the activation function. Represents an exponential function; S3, Manifold Constraint Module: Applies local and global manifold constraints to nodes in a low-dimensional embedding space, preserving the geometric properties of the data; The formulas for the local and global manifold constraints in S3 are as follows: Local manifold constraints: Global manifold constraints: in, and For nodes and nodes The low-dimensional embedding representation, This represents the actual distance between nodes; the local constraints in the manifold constraint module are jointly defined by the weight of the node's neighborhood and the low-dimensional embedding distance, prioritizing the preservation of the geometric relationships between nodes within the neighborhood; The global geometric consistency constraint in the manifold constraint module of S3 is based on the actual distance of high-dimensional data. It is calculated using geodesic distance or Euclidean distance; S4, Graph Decoder: Maps low-dimensional embedding representations back to high-dimensional data space, reconstructs node features, including filling in missing values; S5. Loss Function Optimization: Define the total loss function, including data reconstruction loss, local manifold constraint loss, global manifold constraint loss, and regularization loss; S6. Model Training and Inference: Train the model by optimizing the loss function; use the trained model to reconstruct data with missing values ​​in the wind farm.

2. The wind farm data reconstruction method based on graph attention mechanism and manifold autoencoder according to claim 1, characterized in that, The edge set in the graph construction module of S1 The weights are defined based on both geographical proximity and feature similarity, where feature similarity is determined by calculating the Pearson correlation coefficient of the node feature vectors; by dynamically updating the adjacency matrix, the modeling of time series data is realized to adapt to the time-varying characteristics of the relationships between sites.

3. The wind farm data reconstruction method based on graph attention mechanism and manifold autoencoder according to claim 1, characterized in that, The S2 graph attention encoder uses a multi-head attention mechanism to calculate dynamic weights, as shown in the formula: in, For the number of attention heads, This represents the activation function. Indicates the first Attention weight based on size.

4. The wind farm data reconstruction method based on graph attention mechanism and manifold autoencoder according to claim 1, characterized in that, The S4 graph decoder uses a fully connected layer to map the low-dimensional embedding representation Z back to the high-dimensional feature space, and uses the activation function σ to ensure the nonlinear reconstruction capability of the data.

5. The wind farm data reconstruction method based on graph attention mechanism and manifold autoencoder according to claim 4, characterized in that, The method for reconstructing node features in S4 is as follows: in, For the reconstruction of high-dimensional data, This represents the activation function. Here is the weight matrix of the decoder. It is a low-dimensional embedding representation, generated by a graph encoder.

6. The wind farm data reconstruction method based on graph attention mechanism and manifold autoencoder according to claim 1, characterized in that, The total loss function in S5 is: in, , This represents the total loss function, used to measure the difference between the model's predictions and the actual data, as well as the degree to which the constraints are met. Represents a node The original high-dimensional feature vector, Represents a node Reconstructing high-dimensional feature vectors; , , For weight hyperparameters, These are the weighting coefficients for local constraints. These are the weighting coefficients for global constraints. The regularization strength; Indicates the data reconstruction loss. This represents the loss due to local manifold constraints. This represents the global manifold constraint loss.

7. The wind farm data reconstruction method based on graph attention mechanism and manifold autoencoder according to claim 1, characterized in that: The S6 model initializes model parameters through self-supervised learning during training, thereby improving reconstruction performance on the dataset.

Images