Meteorological equipment abnormal data detection method and system based on graph neural network

Abstract

The invention relates to a meteorological equipment abnormal data detection method and system based on a graph neural network, and belongs to the field of meteorological data processing. The method comprises the following steps: S1, data collection and primary processing; s2, data preprocessing; s3, constructing a network; s4, carrying out space-time diagram convolutional network modeling; s5, performing model training; and S6, abnormal judgment. According to the method, the space-time diagram convolutional network is adopted to carry out joint modeling on meteorological data, deep representation learning on time sequence and space dependence is realized, a self-supervised reconstruction error strategy is introduced at the output end of the model, and a self-supervised reconstruction error is obtained by learning a space-time mode of normal data. Accurate diagnosis of potential abnormal points can be completed without a large number of labeled abnormal samples, and dependence on abnormal labeled data can be effectively reduced.

Description

Technical Field

[0001] The present invention belongs to the field of meteorological data processing, and in particular relates to a method for detecting abnormal data of meteorological equipment based on graph neural network. Background Art

[0002] At present, meteorological equipment plays an indispensable role in monitoring meteorological elements (such as temperature, humidity, air pressure, wind speed, etc.) and environmental changes. However, with the continuous expansion of the monitoring scope and the improvement of the degree of automation, meteorological equipment is extremely susceptible to various factors such as sensor failure, external environmental interference and network transmission during long-term operation, thus generating various types of abnormal data. In traditional anomaly detection methods, threshold rules, statistical analysis or machine learning techniques are often used to diagnose meteorological data. However, threshold-based methods usually rely on experience or prior knowledge and lack flexibility for potential diversified anomalies; and traditional machine learning models often have difficulty in accurately capturing abnormal features when there are not enough anomaly labeled samples. Especially when the spatiotemporal distribution of observation data is complex and there is a significant spatial dependency between multiple sites, it is difficult to guarantee the detection effect by relying solely on time series or geographic location for anomaly detection. In recent years, with the development of deep learning technology, graph neural networks (GNNs) have shown significant potential in processing data with complex topological structures, providing a new solution for anomaly detection based on spatiotemporal associations. By treating meteorological sites as nodes in the graph and constructing adjacency relationships based on sensor observations and their spatial and meteorological correlations, the dependency structure between sites can be more comprehensively characterized. If time series information can be combined, it is expected that potential abnormal patterns can be discovered in the dynamic changes of multi-dimensional meteorological variables. However, there are still some shortcomings in the current application of graph neural networks in the meteorological field. For example, it is impossible to take into account the dynamic interaction between multiple sites, and it relies on a large number of labeled abnormal samples during training. It is difficult to adapt to meteorological application scenarios with changeable environments and diverse abnormal distributions. Therefore, there is an urgent need for a solution that can capture complex spatiotemporal dependencies and effectively learn normal meteorological data patterns when abnormal annotations are insufficient. Summary of the invention

[0003] In order to solve the above problems, the present invention provides a method for detecting abnormal data of meteorological equipment based on graph neural network.

[0004] In order to achieve the above object, the present invention is implemented through the following technical solutions: The present invention provides a method for detecting abnormal data of meteorological equipment based on graph neural network, comprising the following steps: S1. Data collection and preliminary processing: Collection The multidimensional meteorological observation data and geographic information of each meteorological station are obtained to obtain the multidimensional meteorological observation data time series of each meteorological station. , geographic information features and the abnormal annotation set ; S2. Data preprocessing: Preprocess the time series of multi-dimensional meteorological observation data for each meteorological station to obtain the time series of preprocessed multi-dimensional meteorological observation data for each meteorological station ; S3. Network construction: Based on meteorological stations, the time series of preprocessed multi-dimensional meteorological observation data for each meteorological station and the geographical information features construct a meteorological station network to obtain the node features at time step t and the adjacency matrix ; S4. Spatiotemporal graph convolutional network modeling: Input the node features and the adjacency matrix into the spatiotemporal graph convolutional network model for feature fusion through temporal convolutional operations to obtain a high-dimensional embedding representation , and the high-dimensional embedding representation passes through the decoder to obtain the reconstructed meteorological features , and further obtain the reconstruction error of the meteorological station at time step t; S5. Model training: Use the reconstruction error loss function with a regularization term to iteratively update the model parameters to obtain a trained model; S6. Abnormality determination: Use the trained model to predict and reconstruct the newly input data, calculate the reconstruction error, and if the reconstruction error exceeds a predetermined threshold, it is determined as an abnormal point.

[0005] Furthermore, step S1 specifically includes: S11. For each meteorological station, at discrete time steps collect multi-dimensional meteorological observation data and uniformly represent the multi-dimensional meteorological observation data as a vector to obtain the time series of multi-dimensional meteorological observation data of each meteorological station at time step , and the formula is as follows: , where represents the dimension of meteorological variables, including temperature, humidity, air pressure, and wind speed, represents the observation value of the th meteorological variable at meteorological station at time step , and integrate the multi-dimensional meteorological observation data of each meteorological station at time steps to obtain the time series of multi-dimensional meteorological observation data of each meteorological station​ , which is expressed by the following formula: ; S12. Collect the geographical location information and meta-information of each meteorological station to obtain geographical information features , which is expressed by the following formula: , where respectively represent the longitude, latitude and altitude of the meteorological station; S13. Combine the sensor self-check logs and the results of manual inspection to obtain a set of labeled known abnormal data , which is expressed by the following formula: , where represents the meteorological station with abnormal data, represents the time step corresponding to the abnormal data, represents the time series of multi-dimensional meteorological observation data with anomalies.

[0006] Furthermore, step S2 specifically includes: S21. Standardization processing: Let the mean of the th meteorological variable at all stations and all time steps be , and the standard deviation be . Standardize the time series of multi-dimensional meteorological observation data of the th meteorological station at time step to obtain the standardized time series of multi-dimensional meteorological observation data, which is expressed by the following formula: , , S22. Data alignment: Perform time alignment operations on the standardized time series of multi-dimensional meteorological observation data. If the observed data is missing at a certain time step, replace it with the most recent observed value to obtain the aligned time series of multi-dimensional meteorological observation data, which is expressed by the following formula: , where represents the observed value of the th aligned meteorological variable at meteorological station at time step ; Finally, obtain the time series of preprocessed multi-dimensional meteorological observation data for each meteorological station, which is expressed by the following formula: .

[0007] Furthermore, step S3 specifically includes: S31.Node definition: Each meteorological station As a node, the node features include the time series of multidimensional meteorological observation data preprocessed by each meteorological station and geographic information features ; S32. Edge and adjacency relationship construction: define the weather station network at each time step The adjacency matrix of , a dynamic neighbor construction method based on meteorological variable correlation is used to calculate the meteorological variable correlation coefficient between each station in the sliding window: , in the window size Calculate weather stations within range and weather stations Pearson correlation coefficient of multidimensional meteorological observation data time series , the formula is as follows: , in, Indicates weather station At time step to The meteorological observation subsequence, Indicates weather station At time step to The meteorological observation subsequence, Represents the complete calculation process of the Pearson correlation coefficient, and the result is a standardized correlation index in the range of [-1,1]. At each time step The adjacency matrix of Zhongwei and Add edges and assign weights , otherwise it is 0. The formula is as follows: , in, Represents each time step The adjacency matrix of middle and The weight between Represents an indicator function, which is 1 when the condition in the brackets is met, otherwise it is 0. Indicates the threshold for setting the adjacency matrix weights; S33. Building dynamic graph structure: Based on each time step The adjacency matrix of , combined with the aligned multi-dimensional meteorological observation data time series at the current time step constitute a series of dynamic graph representations, which are expressed by the following formula: , , , where, represents the dynamic graph structure, represents the node set of meteorological stations, which is used to represent all nodes in the graph, represents the total number of time steps, represents the node features at time step t, represents the th meteorological station located in the aligned multi-dimensional meteorological observation data time series at time step t, represents the th adjacency matrix at time step t.

[0008] Furthermore, the threshold in step S32 takes the 95% quantile of the distribution.

[0009] Furthermore, step S4 specifically includes: Input the node features at time step t and the adjacency matrix into the spatio-temporal graph convolutional network model, and perform feature fusion between adjacent time steps through temporal convolution. The propagation rule formula of the th layer is expressed as follows: , where, represents the node representation of the th layer, , represents the initial node features, , represents the identity matrix, is the degree matrix of, represents the trainable weights, represents the non-linear activation function; after the temporal convolution is completed, the high-dimensional embedding representation at time step is finally obtained, and the formula is expressed as follows: , where, represents the number of temporal convolution layers, represents the final embedding dimension; the high-dimensional embedding representation is input into the decoder for reconstruction to obtain the reconstructed meteorological features , , Indicates the multi-dimensional meteorological observation data of the th reconstructed meteorological station, and the decoder is a multi-layer perceptron.

[0010] Furthermore, in step S4, the meteorological station The reconstruction error at time step t is .

[0011] Furthermore, step S5 specifically includes: The formula of the reconstruction error loss function is expressed as follows: , Adding a regularization term to the reconstruction error loss function , and obtaining the training objective function, the formula is expressed as follows: , where Represents the hyperparameter for adjusting the regularization strength, Represents the overall set containing all trainable weights. The trainable weights include the trainable weights in the temporal convolution and the internal parameters of the multi-layer perceptron. The regularization term Adopts norm.

[0012] Furthermore, step S6 specifically includes: The predetermined threshold is set by the standard deviation method. Calculate the mean and standard deviation of the reconstruction error in the training set, and the threshold is set to: , where Is the proportionality coefficient, and the value is 3.

[0013] The present invention also provides a meteorological equipment abnormal data detection system based on a graph neural network, which executes the above-mentioned meteorological equipment abnormal data detection method based on a graph neural network, including: Data collection and preliminary processing module: used to collect multi-dimensional meteorological observation data and geographical information of meteorological stations, and obtain the multi-dimensional meteorological observation data time series, geographical information features and abnormal annotation sets of each meteorological station; Data preprocessing module: used to preprocess the multi-dimensional meteorological observation data time series of each meteorological station to obtain the preprocessed multi-dimensional meteorological observation data time series of each meteorological station; Network construction module: used to construct a meteorological station network based on meteorological stations, the preprocessed multi-dimensional meteorological observation data time series of each meteorological station and geographical information features, and obtain node features and an adjacency matrix; Spatio-temporal graph convolutional network modeling unit: It is used to input node features and adjacency matrix into the spatio-temporal graph convolutional network model, perform feature fusion through temporal convolutional operation to obtain high-dimensional embedding representation, input the high-dimensional embedding representation into the decoder to obtain reconstructed meteorological features, and then obtain the reconstruction error; Model training unit: It is used to iteratively update the model parameters by using the reconstruction error loss function with regularization terms to obtain a trained model; Abnormality determination unit: It is used to use the trained model to predict and reconstruct the newly input data, calculate the reconstruction error, and if the reconstruction error exceeds a predetermined threshold, it is determined as an abnormal point.

[0014] The advantages of the present invention are as follows: The present invention proposes a method for detecting abnormal data of meteorological equipment based on graph neural network, introduces a dynamic adjacency relationship construction strategy, can capture real meteorological relevance in spatio-temporal dimensions, and compared with the abnormal detection schemes based only on time series or single-site information, the present invention can more comprehensively identify the abnormalities generated in the interaction of multiple sites; the present invention uses a spatio-temporal graph convolutional network to jointly model meteorological data, realizes deep representation learning of temporal and spatial dependencies, introduces a self-supervised reconstruction error strategy at the output end of the model, and through learning the spatio-temporal patterns of normal data, can complete the accurate diagnosis of potential abnormal points without a large number of labeled abnormal samples, effectively reducing the dependence on abnormal labeled data. Description of the drawings

[0015] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification, and are used to explain the present invention together with the embodiments of the present invention, and do not constitute a limitation to the present invention.

[0016] Figure 1 It is the step flow chart of the method of the present invention; Figure 2 It is the experimental result graph of the method of the present invention and the traditional method. Detailed implementation manners

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

[0018] Embodiment 1 In this embodiment, as Figure 1 shown, a method for detecting abnormal data of meteorological equipment based on graph neural network is provided, and the specific steps include: S1. Data collection and preliminary processing: Collect multidimensional meteorological observation data and geographical information of meteorological stations, and obtain the time series of multidimensional meteorological observation data for each meteorological station , geographical information features and the abnormal annotation set ; Specifically, S11. For each meteorological station, at discrete time steps collect multidimensional meteorological observation data, and uniformly represent the multidimensional meteorological observation data as vectors, obtaining the time series of multidimensional meteorological observation data for each meteorological station at time step , which is expressed by the formula as follows: , where represents the dimension of meteorological variables, including temperature, humidity, air pressure, and wind speed, represents the th meteorological variable's observed value at meteorological station at time step . Integrate the multidimensional meteorological observation data of time steps for each meteorological station to obtain the time series of multidimensional meteorological observation data for each meteorological station , which is expressed by the formula as follows: ; S12. Collect the geographical location information and meta-information of each meteorological station to obtain geographical information features , which is expressed by the formula as follows: , where represent the longitude, latitude, and altitude of the meteorological station respectively; S13. Combine the sensor self-check log and the manual inspection result to obtain the known abnormal data annotation set , which is expressed by the formula as follows: , where represents the meteorological station with abnormal data, represents the time step corresponding to the abnormal data, represents the time series of multidimensional meteorological observation data with anomalies.

[0019] S2. Data preprocessing: Preprocess the time series of multidimensional meteorological observation data for each meteorological station to obtain the preprocessed time series of multidimensional meteorological observation data for each meteorological station ; ​Specifically, S21. Standardization processing: Let the mean of the th meteorological variable over all stations and all time steps be , and the standard deviation be . Standardize the time series of multi-dimensional meteorological observation data of the th meteorological station at time step to obtain the standardized time series of multi-dimensional meteorological observation data , which is expressed by the formula as follows: , . S22. Data alignment: Perform time alignment on the standardized time series of multi-dimensional meteorological observation data . If there is missing observation data at a certain time step, replace it with the most recent observation value to obtain the aligned time series of multi-dimensional meteorological observation data , which is expressed by the formula as follows: , where represents the observation value of the th meteorological variable at meteorological station at time step . Finally, obtain the preprocessed time series of multi-dimensional meteorological observation data for each meteorological station , which is expressed by the formula as follows: .

[0020] S3. Network construction: Based on meteorological stations, the preprocessed time series of multi-dimensional meteorological observation data for each meteorological station and the geographic information features , construct a meteorological station network to obtain the node features and the adjacency matrix at time step t; Specifically, S31. Node definition: Take each meteorological station as a node, and the node features include the preprocessed time series of multi-dimensional meteorological observation data for each meteorological station and the geographic information features ; S32. Edge and adjacency relationship construction: Define the adjacency matrix of each time step of the meteorological station network as . Adopt a dynamic adjacency construction method based on the correlation of meteorological variables to calculate the correlation coefficient of meteorological variables between stations within a sliding window: For time step , within the window size Calculate meteorological stations within a range and meteorological stations Pearson correlation coefficient of the time series of multi-dimensional meteorological observation data , which is expressed by the formula as follows: , where, represents the meteorological observation subsequence of meteorological station from time step to . represents the meteorological observation subsequence of meteorological station from time step to . represents the complete calculation process of the Pearson correlation coefficient, and its result is a standardized correlation index, ranging from [-1, 1]. When , in the adjacency matrix at each time step for and add an edge and assign a weight , otherwise it is 0. The formula is expressed as follows: , where, represents the weight between and in the adjacency matrix at each time step , represents the indicator function, which is 1 when the condition in the parentheses is satisfied, otherwise it is 0, represents the threshold for setting the adjacency matrix weight; S33. Construct a dynamic graph structure: Based on the adjacency matrix at each time step , combined with the aligned multi-dimensional meteorological observation data time series at the current time step, a series of dynamic graph representations are formed. The formula is expressed as follows: , , , where, represents the dynamic graph structure, represents the set of nodes of meteorological stations, used to represent all nodes in the graph, represents the total number of time steps, represents the node features at time step t, represents the th aligned multi-dimensional meteorological observation data time series of the meteorological station at time step t, Indicates the adjacency matrix at the th time step.

[0021] S4. Spatiotemporal Graph Convolutional Network Modeling: Input the node features and the adjacency matrix into the spatiotemporal graph convolutional network model, and perform feature fusion through temporal convolutional operations to obtain a high-dimensional embedding representation . The high-dimensional embedding representation passes through the decoder to obtain the reconstructed meteorological features , and then the reconstruction error of the meteorological station at time step t is obtained; Specifically, input the node features and the adjacency matrix at time step t into the spatiotemporal graph convolutional network model, perform feature fusion between adjacent time steps through temporal convolution, and the propagation rule formula of the th layer is expressed as follows: , where, represents the node representation of the th layer, , represents the initial node features, , represents the identity matrix, is the degree matrix of, represents the trainable weight, represents the non-linear activation function; after the temporal convolution is completed, the high-dimensional embedding representation at time step is finally obtained, and the formula is expressed as follows: , where, represents the number of temporal convolution layers, represents the final embedding dimension; the high-dimensional embedding representation is input into the decoder for reconstruction to obtain the reconstructed meteorological features , , represents the multi-dimensional meteorological observation data of the th reconstructed meteorological station, and the decoder is a multi-layer perceptron; The reconstruction error of the meteorological station at time step t is .

[0022] S5. Model Training: Use the reconstruction error loss function with a regularization term to iteratively update the model parameters to obtain a trained model; Specifically, the formula of the reconstruction error loss function is expressed as follows: , Add a regularization term to the reconstruction error loss function , and obtain the training objective function, which is expressed by the following formula: , where represents the hyperparameter for adjusting the regularization strength represents the overall set containing all trainable weights. The trainable weights include the trainable weights in the temporal convolution and the internal parameters of the multi-layer perceptron. The regularization term adopts norm

[0023] S6. Anomaly determination: Use the trained model to predict and reconstruct the newly input data, calculate the reconstruction error. If the reconstruction error exceeds a predetermined threshold, it is determined as an anomaly point

[0024] Specifically, the predetermined threshold is set by the standard deviation method. Calculate the mean and standard deviation of the reconstruction error in the training set, and the threshold is set as: , where is the proportionality coefficient

[0025] Example 2 This example is to verify the effectiveness of the meteorological equipment anomaly data detection method based on the graph neural network of the present invention. Design a comparative experiment to simulate the time series data of a real meteorological station. The data includes multi-dimensional meteorological observation variables (such as temperature, humidity, air pressure, etc.) and labeled anomaly points. As Figure 2 shown. The experiment compares the method of the present invention with the following algorithms: the baseline model (Z-Score based on the sliding window), the machine learning model (LSTM and GRU time series models), and the graph neural network model (static GCN); the machine learning model is a method based only on the time series information of a single station, and the graph neural network model is a method of multi-station static information. The comparison results under the four indicators of accuracy, precision, recall, and F1-score are shown in the figure. The experimental results show that compared with the anomaly detection schemes based only on time series or single-station information, all evaluation indicators of the method of the present invention are higher than the other two methods, proving that the present invention can more comprehensively identify the anomalies generated in the multi-station interaction; the present invention combines the dynamic adjacency matrix and the spatio-temporal graph convolutional network, and can simultaneously capture the time changes and spatial correlations of meteorological stations, significantly superior to the comparative algorithms

[0026] Example 3 This embodiment provides a meteorological equipment abnormal data detection system based on a graph neural network, which executes the meteorological equipment abnormal data detection method based on the graph neural network described in Embodiment 1, including: Data collection and preliminary processing module: It is used to collect multi-dimensional meteorological observation data and geographical information of meteorological stations, and obtain the multi-dimensional meteorological observation data time series, geographical information features, and abnormal annotation sets of each meteorological station; Data preprocessing module: It is used to preprocess the multi-dimensional meteorological observation data time series of each meteorological station to obtain the preprocessed multi-dimensional meteorological observation data time series of each meteorological station; Network construction module: It is used to construct a meteorological station network based on meteorological stations, the preprocessed multi-dimensional meteorological observation data time series of each meteorological station, and geographical information features, and obtain node features and an adjacency matrix; Spatio-temporal graph convolutional network modeling unit: It is used to input the node features and the adjacency matrix into the spatio-temporal graph convolutional network model to perform feature fusion through temporal convolutional operations to obtain a high-dimensional embedding representation, and input the high-dimensional embedding representation into a decoder to obtain reconstructed meteorological features, and further obtain a reconstruction error; Model training unit: It is used to iteratively update the model parameters using a reconstruction error loss function with a regularization term to obtain a trained model; Abnormality determination unit: It is used to use the trained model to predict and reconstruct newly input data, calculate the reconstruction error, and if the reconstruction error exceeds a predetermined threshold, it is determined as an abnormal point.

[0027] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for detecting abnormal data of meteorological equipment based on graph neural network, characterized in that: The following steps are involved: S1. Data collection and preliminary processing: Collection The multidimensional meteorological observation data and geographic information of each meteorological station are obtained to obtain the multidimensional meteorological observation data time series of each meteorological station. , geographic information features And the anomaly annotation collection ; S2. Data preprocessing: Multidimensional meteorological observation data time series of each meteorological station Preprocessing is performed to obtain the multidimensional meteorological observation data time series after preprocessing at each meteorological station ; S3. Network construction: based on Meteorological stations, multi-dimensional meteorological observation data time series after preprocessing at each meteorological station and geographic information features Construct a weather station network and obtain node characteristics at time step t and the adjacency matrix ; S4. Spatiotemporal graph convolutional network modeling: node features and the adjacency matrix Input into the spatiotemporal graph convolutional network model and perform feature fusion through temporal convolution operations to obtain a high-dimensional embedding representation , the high-dimensional embedding representation The reconstructed meteorological features are obtained through the decoder , and then get the weather station The reconstruction error at time step t; S5. Model training: The model parameters are iteratively updated using the reconstruction error loss function with a regularization term added to obtain a trained model; S6. Anomaly determination: Use the trained model to predict and reconstruct the newly input data, calculate the reconstruction error, and if the reconstruction error exceeds the predetermined threshold, it is identified as an anomaly point.

2. The method for detecting abnormal data of meteorological equipment based on graph neural network according to claim 1 is characterized in that: Step S1 specifically includes: S11. For each meteorological station, at discrete time steps Collect multidimensional meteorological observation data and represent them uniformly as vectors to obtain the time step of each meteorological station. Multidimensional meteorological observation data time series , the formula is as follows: ， in, Represents the dimensions of meteorological variables, including temperature, humidity, air pressure, and wind speed, Indicates Meteorological variables at the meteorological station At time step The observation values ​​of each meteorological station The multidimensional meteorological observation data of each time step is integrated to obtain the multidimensional meteorological observation data time series of each meteorological station. , the formula is as follows: ； S12. Collect the geographic location information and meta-information of each meteorological station to obtain geographic information features , the formula is as follows: ， in, Respectively represent the longitude, latitude and altitude of the meteorological station; S13. Combine the sensor self-test log and manual test results to obtain a known abnormal data annotation set , the formula is as follows: ， in, Indicates weather stations with abnormal data, represents the time step corresponding to the abnormal data, Represents a multidimensional time series of meteorological observation data with anomalies.

3. The method for detecting abnormal data of meteorological equipment based on graph neural network according to claim 2 is characterized in that: Step S2 specifically includes: S21. Standardization: Set The mean of the meteorological variables at all stations and all time steps is , the standard deviation is , for The weather station is at the time step Multidimensional meteorological observation data time series Standardize to obtain the standardized multidimensional meteorological observation data time series , the formula is as follows: ， ， S22. Data alignment: Standardized multidimensional meteorological observation data time series Perform time alignment. If observation data is missing at a certain time step, the most recent observation value is used to replace it to obtain an aligned multidimensional meteorological observation data time series. , the formula is as follows: ， in, Indicates the alignment Meteorological variables at the meteorological station At time step Finally, the multidimensional meteorological observation data time series after preprocessing at each meteorological station is obtained. , the formula is as follows: 。 4. The method for detecting abnormal data of meteorological equipment based on graph neural network according to claim 3 is characterized in that: Step S3 specifically includes: S31.Node definition: Each meteorological station As a node, the node features include the time series of multidimensional meteorological observation data preprocessed by each meteorological station and geographic information features ; S32. Edge and adjacency relationship construction: define the weather station network at each time step The adjacency matrix of , a dynamic neighbor construction method based on meteorological variable correlation is used to calculate the meteorological variable correlation coefficient between each station in the sliding window: , in the window size Calculate weather stations within range and weather stations Pearson correlation coefficient of multidimensional meteorological observation data time series , the formula is as follows: ， in, Indicates weather station At time step to The meteorological observation subsequence, Indicates weather station At time step to The meteorological observation subsequence, Represents the complete calculation process of the Pearson correlation coefficient, and the result is a standardized correlation index in the range of [-1,1]. At each time step The adjacency matrix of Zhongwei and Add edges and assign weights , otherwise it is 0. The formula is as follows: ， in, Represents each time step The adjacency matrix of middle and The weight between Represents an indicator function, which is 1 when the condition in the brackets is met, otherwise it is 0. Indicates the threshold for setting the adjacency matrix weights; S33. Building dynamic graph structure: Based on each time step The adjacency matrix of , combined with the aligned multidimensional meteorological observation data time series of the current time step Constitute a series of dynamic graph representations, the formula is as follows: ， ， ， in, Represents a dynamic graph structure, A node set representing a meteorological station, used to represent all nodes in the graph, represents the total number of time steps, represents the node features at time step t, Indicates The aligned multidimensional meteorological observation data time series of meteorological stations at time step t, Indicates The adjacency matrix for each time step.

5. The method for detecting abnormal data of meteorological equipment based on graph neural network according to claim 4 is characterized in that: Step S4 specifically includes: The node features at time step t and the adjacency matrix Input into the spatiotemporal graph convolutional network model, and perform feature fusion between adjacent time steps through temporal convolution. The propagation rule formula of the layer is expressed as follows: ， in, Indicates The node representation of the layer, , represents the initial node characteristics, , represents the identity matrix, for The degree matrix of represents the trainable weights, represents a nonlinear activation function; after the temporal convolution is completed, the time step is finally obtained High-dimensional embedding representation of , the formula is as follows: ， in, represents the number of temporal convolution layers, Represents the final embedding dimension; high-dimensional embedding representation Input to the decoder for reconstruction to obtain the reconstructed meteorological characteristics , , Represents the reconstruction The decoder is a multi-layer perceptron.

6. The method for detecting abnormal data of meteorological equipment based on graph neural network according to claim 5 is characterized in that: In step S4, the weather station The reconstruction error at time step t is .

7. The method for detecting abnormal data of meteorological equipment based on graph neural network according to claim 6 is characterized in that: Step S5 specifically includes: The reconstruction error loss function formula is as follows: ， Add a regularization term based on the reconstruction error loss function , we get the training objective function, the formula is as follows: ， in, represents a hyperparameter that adjusts the regularization strength, Represents the overall set of all trainable weights, including the trainable weights in the temporal convolution and the internal parameters of the multilayer perceptron, the regularization term use Norm.

8. The method for detecting abnormal data of meteorological equipment based on graph neural network according to claim 7 is characterized in that: In step S32, the threshold Pick The 95% quantile of the distribution.

9. The method for detecting abnormal data of meteorological equipment based on graph neural network according to claim 8, characterized in that: Step S6 specifically includes: The predetermined threshold is set using the standard deviation method, and the mean of the reconstruction error in the training set is calculated. and standard deviation , the threshold is set to: ， in, is the proportional coefficient, the value is 3.

10. A meteorological equipment abnormal data detection system based on graph neural network, executing the meteorological equipment abnormal data detection method based on graph neural network as claimed in claim 1, characterized in that: include: Data collection and preliminary processing module: used to collect multidimensional meteorological observation data and geographic information of meteorological stations, and obtain the multidimensional meteorological observation data time series, geographic information features and abnormal annotation sets of each meteorological station; Data preprocessing module: used to preprocess the multidimensional meteorological observation data time series of each meteorological station to obtain the preprocessed multidimensional meteorological observation data time series of each meteorological station; Network construction module: used to construct a meteorological station network based on the meteorological station, the pre-processed multi-dimensional meteorological observation data time series of each meteorological station, and the geographic information features, and obtain node features and adjacency matrix; Spatiotemporal graph convolutional network modeling unit: used to input node features and adjacency matrix into the spatiotemporal graph convolutional network model, perform feature fusion through temporal convolution operation to obtain high-dimensional embedding representation, and input the high-dimensional embedding representation into the decoder to obtain reconstructed meteorological features, and then obtain reconstruction error; Model training unit: used to iteratively update model parameters using a reconstruction error loss function with a regularization term added to obtain a trained model; Anomaly determination unit: used to use the trained model to predict and reconstruct the newly input data, calculate the reconstruction error, and if the reconstruction error exceeds the predetermined threshold, it is identified as an anomaly point.

Images