Power transmission channel abnormal condition monitoring and early warning method and system

Abstract

The invention provides a power transmission channel abnormal condition monitoring and early warning method and system. The method comprises the following steps: acquiring channel data based on a power grid resource service platform, acquiring regional data based on a power grid power utilization information acquisition system, and acquiring historical meteorological grid data based on a meteorological center; constructing a channel-to-region mapping relation model and a different-to-different matching relation model based on the channel data and the region data; performing time sequence alignment and space alignment on abnormal event data recorded in the historical meteorological raster data and the channel data, performing data preprocessing and feature engineering operation on the historical meteorological raster data and the corresponding abnormal event data to form an input sequence, and training and optimizing a GD-Informer model based on the input sequence to obtain an optimal model; and performing real-time prediction on the new meteorological data based on the optimal model to obtain a prediction result, and performing result analysis and abnormal condition early warning based on the prediction result and the two relation models.

Description

Technical Field

[0001] The present invention relates to the technical field of early warning of abnormal conditions in transmission channels, and particularly to a method and system for monitoring and early warning of abnormal conditions in transmission channels. Background Art

[0002] As an important part of the power system, transmission lines undertake the task of transmitting electric energy from power plants to user terminals, and their operating status is related to the safe and stable operation of the power system and the sustainable and healthy development of the economy and society.

[0003] When various types of faults occur in transmission lines, they will seriously threaten the safe and stable operation of the power system. The current correlation monitoring method based on a single data source and the early warning method of coarse-grained and regional-wide full coverage have limitations such as low flexibility and high omission rate in the inspection and monitoring process, resulting in problems such as insufficient prediction refinement, frequent false alarms and missed alarms. In addition, the current power grid is facing increasingly severe challenges from meteorological disasters. Accurately monitoring and early warning abnormal problems of multi-source and complex transmission lines has become a severe test for ensuring the stable operation of the power system. Therefore, there is an urgent need for a method and system for monitoring and early warning of abnormal conditions in transmission channels with strong systematicness, good versatility, high standardization and excellent accuracy. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for monitoring and early warning of abnormal conditions in transmission channels, which comprehensively uses deep learning technology and data processing and analysis technology, and effectively improves the accuracy, real-time performance and reliability of monitoring abnormal conditions in transmission channels.

[0005] In the first aspect, the present invention provides a method for monitoring and early warning of abnormal conditions in transmission channels, specifically including: Obtaining channel data based on the power grid resource business middle platform, where the channel data includes transmission channel spatial data and recorded abnormal event data, obtaining regional data based on the power grid power consumption information collection system, and obtaining historical meteorological grid data based on the meteorological center, where the meteorological grid data includes meteorological data and grid data; Constructing a channel-to-region mapping relationship model and an abnormal-to-abnormal matching relationship model based on the channel data and the regional data; After performing time series alignment and spatial alignment on the historical meteorological grid data and the abnormal event data recorded in the channel data, performing data preprocessing and feature engineering operations on the abnormal event data in the historical meteorological grid data and the corresponding channel data to form an input sequence, and training and tuning the GD-Informer model based on the input sequence to obtain a trained GD-Informer model; Based on the trained GD-Informer model, real-time prediction is performed on new meteorological data to obtain prediction results, and result analysis and early warning of abnormal situations are carried out based on the prediction results, the channel-to-region mapping relationship model, and the anomaly-to-anomaly matching relationship model.

[0006] Further, the meteorological data includes temperature, humidity, wind speed, wind direction, and precipitation, and the grid data includes prediction time, report time, longitude, and latitude.

[0007] Further, constructing the channel-to-region mapping relationship model and the anomaly-to-anomaly matching relationship model based on channel data and region data includes: Based on channel data and region data, region matching and anomaly pairing are performed to construct the channel-to-region mapping relationship model and the anomaly-to-anomaly matching relationship model.

[0008] Further, constructing the channel-to-region mapping relationship model based on channel data and region data includes: Based on channel data and region data, by identifying key points and channels, analyzing the power grid topology relationship, establishing an influence matrix and a parametric relationship, the channel-to-region mapping relationship model is constructed.

[0009] Further, constructing the anomaly-to-anomaly matching relationship model based on channel data and region data includes: Based on channel data and region data, by determining the fault type and the type of adverse impact, establishing a relationship matrix and a parametric relationship, the anomaly-to-anomaly matching relationship model is constructed.

[0010] Further, after performing temporal alignment and spatial alignment on the abnormal event data recorded in the historical grid meteorological data and channel data, data preprocessing and feature engineering operations are performed on the abnormal event data in the historical meteorological grid data and the corresponding channel data to form an input sequence, and based on the input sequence, the GD-Informer model is trained and optimized to obtain the trained GD-Informer model, including: Performing temporal alignment operation and spatial alignment operation on the abnormal event data recorded in the historical meteorological grid data and channel data; Dividing the abnormal event data recorded in the historical meteorological grid data and channel data into training samples and test samples; Performing data preprocessing and feature engineering operations on the abnormal event data in the historical meteorological grid data and the corresponding channel data to form an input sequence; Based on the input sequence, the GD-Informer model is trained, tested, and optimized to obtain the trained GD-Informer model.

[0011] Further, performing temporal alignment operation and spatial alignment operation on the abnormal event data recorded in the historical meteorological grid data and channel data includes: The time series alignment operation includes timestamp merging, finding common timestamps, and data alignment;

[0012] The spatial alignment operation performs regional division and matching based on the location data in the raster data and the abnormal event data.

[0013] Furthermore, data preprocessing is performed on the abnormal event data in the historical meteorological raster data and the corresponding channel data, including: Performing data cleaning, normalization, and data splitting operations on the abnormal event data in the historical meteorological raster data and the corresponding channel data.

[0014] Furthermore, feature engineering operations are performed on the abnormal event data in the historical meteorological raster data and the corresponding channel data, including: Performing feature selection, feature construction, and sequence construction operations on the abnormal event data in the historical meteorological raster data and the corresponding channel data.

[0015] Furthermore, based on the trained GD-Informer model, real-time prediction is performed on new meteorological data to obtain prediction results, and result analysis and abnormal situation warning are performed based on the prediction results, the channel-to-region mapping relationship model, and the abnormal-to-abnormal matching relationship model, including: Performing real-time prediction on the obtained meteorological data based on the trained GD-Informer model to obtain prediction results; After obtaining the prediction results, the channel-to-region mapping relationship model and the abnormal-to-abnormal matching relationship model are used for result analysis and abnormal situation warning.

[0016] Furthermore, after obtaining the prediction results, the channel-to-region mapping relationship model and the abnormal-to-abnormal matching relationship model are used for result analysis and abnormal situation warning, including: Result analysis is to form a heat map and a case report based on the prediction results using the channel-to-region mapping relationship model and the abnormal-to-abnormal matching relationship model. The heat map shows the predicted distribution of abnormal situations in different regions of the transmission channels, and the case report prepares a detailed case report for the predicted event, recording the event process and subsequent actions; The abnormal situation warning is to perform threshold warning for different abnormal event impact degrees. When the impact degree reaches the preset threshold, the corresponding emergency procedure is started, and real-time push is performed through the SMS and enterprise-level message push platform to notify the duty personnel to take corresponding measures. The push content includes time, weather, transmission channels, affected areas, and impact degrees.

[0017] In a second aspect, the present invention also provides a monitoring and warning system for abnormal situations in transmission channels, specifically including: The data acquisition module obtains channel data based on the power grid resource business middle platform, obtains regional data based on the power grid power consumption information acquisition system, and obtains historical meteorological grid data based on the meteorological center; The relationship model construction module constructs a mapping relationship model of channels to regions and a different-to-different matching relationship model based on the obtained channel data and regional data; The model training module performs temporal alignment and spatial alignment on the historical meteorological grid data and the abnormal event data recorded in the channel data, performs data preprocessing and feature engineering operations on the abnormal event data in the historical meteorological grid data and the corresponding channel data to form an input sequence, and trains and tunes the GD-Informer model based on the input sequence to obtain a trained GD-Informer model; The result determination and output module performs real-time prediction on new meteorological data based on the trained GD-Informer model to obtain a prediction result, then analyzes the result based on the prediction result, the channel-to-region mapping relationship model, and the different-to-different matching relationship model, and warns of the abnormal situation of the transmission channel and the regional abnormal situation through text messages and the enterprise-level message push platform.

[0018] In the above technical solution, the technical effects and advantages provided by the present invention are as follows: The present invention comprehensively applies deep learning technology and data processing and analysis technology, effectively improves the accuracy, real-time performance, and reliability of monitoring the abnormal situation of transmission channels, and helps to promptly grasp the abnormal situation of transmission channels; effectively integrates multi-source data such as transmission channel spatial data, power grid regional data, and historical meteorological grid data to provide more comprehensive, accurate, and in-depth information analysis, and improves the prediction accuracy of the model; uses the self-built G-DAE model to replace the input layer of the Informer model to build the GD-Informer model, effectively learning robust and representative latent features from time series data, not only reducing the data dimension, reducing the influence of noise and outliers, but also improving the generalization ability of the model, preventing overfitting, and at the same time making the model more focused on the key information of the data; based on GD-Informer for time series prediction, effectively capturing the time dependence in the data and the interaction between variables, reducing the number of parameters to reduce the risk of overfitting, and providing comprehensive, accurate, and efficient prediction capabilities; combining the self-built mapping relationship model of transmission channels and power grid regions and the different-to-different matching relationship model, comprehensively analyzing and outputting accurate monitoring and warning results of the abnormal situation of transmission channels; real-time pushing through text messages and the enterprise-level message push platform, enabling staff to take corresponding preventive measures in a timely manner, and at the same time being able to conduct retrospective analysis and dynamic adjustment. The present invention can accurately and timely warn of the abnormal situation of transmission channels and regional abnormal situations, can reduce the impact of bad weather on power grid equipment and regions, reduce the possibility of major accidents, and reduce social and economic losses. Description of the Drawings

[0019] Figure 1 This is the overall flowchart of a method for monitoring and warning abnormal situations in a power transmission channel according to the present invention; Figure 2 This is the overall structure diagram of the GD-Informer model of the present invention. Specific implementation manners

[0020] In order to more clearly elaborate the purpose, technical solutions and their advantages of the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be described in detail below. Obviously, the provided embodiments only represent a part of the implementation situations of the present invention, not all. According to the embodiments described in the present invention, any other embodiments conceived by any person skilled in the art without creative efforts should be included in the protection scope of the present invention. Unless otherwise specified, the technical terms or scientific terms used herein should be interpreted according to the general meanings understood by those of ordinary skill in the art. The expressions such as "including" used herein are intended to indicate that the elements or objects described cover the elements or objects listed subsequently, as well as their equivalents, but do not exclude other elements or objects.

[0021] The embodiments of the present invention provide a method for monitoring and warning abnormal situations in a power transmission channel. Refer to Figure 1 , including the following steps: S1. Obtain channel data based on the power grid resource business middle platform. The channel data includes power transmission channel spatial data and recorded abnormal event data. Obtain regional data based on the power grid power consumption information acquisition system. The regional data includes interval abnormal data. Obtain historical meteorological grid data based on the meteorological center. The meteorological grid data includes meteorological data and grid data; S2. Construct a channel-to-region mapping relationship model and an abnormal-to-abnormal matching relationship model based on the channel data and the regional data; S3. After performing time series alignment and spatial alignment on the historical meteorological grid data and the abnormal event data recorded in the channel data, perform data preprocessing and feature engineering operations on the abnormal event data in the historical meteorological grid data and the corresponding channel data to form an input sequence. Train and optimize the GD-Informer model based on the input sequence to obtain a trained GD-Informer model; S4. Perform real-time prediction on new meteorological data based on the trained GD-Informer model to obtain a prediction result, and perform result analysis and abnormal situation warning based on the prediction result, the channel-to-region mapping relationship model and the abnormal-to-abnormal matching relationship model.

[0022] In some embodiments, when performing step S1, the meteorological data includes temperature, humidity, wind speed, wind direction, and precipitation, and the grid data includes prediction time, report time, longitude, and latitude.

[0023] In some embodiments, when performing step S2, it specifically includes: S2-1. Based on the channel data and regional data, construct a channel-to-region mapping relationship model by identifying key points and channels, analyzing the power grid topology relationship, establishing an influence matrix, and parameterizing the relationship; S2-2. Based on the channel data and regional data, construct a different-to-different matching relationship model by determining the fault type and the type of adverse impact, establishing a relationship matrix, and parameterizing the relationship.

[0024] Specifically, when performing step S2-1, based on the channel data and regional data, construct a channel-to-region mapping relationship model by identifying key points and channels, analyzing the power grid topology relationship, establishing an influence matrix, and parameterizing the relationship, including: S2-1-1. Determine the key nodes and key channels in the spatial channel according to the transmission spatial channel data and interval anomaly data; S2-1-2. Analyze the topological connection between the spatial channel and the affected area, and determine the current direction and possible fault propagation path; S2-1-3. Establish an influence matrix, where the rows represent each interval of the spatial channel, the columns represent each sub-region of the affected area, and the matrix elements represent the degree of influence of the state of a certain channel interval on a certain sub-region; S2-1-4. Assign specific values to the influence matrix according to the transmission spatial channel data and interval anomaly data, and parameterize the relationship to construct a channel-to-region mapping relationship model; Specifically, when performing step S2-2, based on the channel data and regional data, construct a different-to-different matching relationship model by determining the fault type and the type of adverse impact, establishing a relationship matrix, and parameterizing the relationship, including: S2-2-1. Determine the channel fault type and the type of adverse impact generated in different regions. The channel fault types include short circuit, open circuit, grounding, overvoltage, overload, insulation fault, etc., and the influence types include economic influence, ecological environment influence, safety hazard influence, etc.; S2-2-2. Establish a relationship matrix, where the rows represent the channel fault types, the columns represent the types of adverse impacts generated in different regions, and the matrix elements represent the degree of entanglement between different channel fault types and various influence types; S2-2-3. Assign specific values to the influence matrix according to the transmission spatial channel data and interval anomaly data, parameterize the relationship, and construct a different-to-different matching relationship model.

[0025] In some embodiments, when performing step S3, after performing temporal alignment and spatial alignment on the abnormal event data recorded in the historical meteorological grid data and channel data, data preprocessing and feature engineering operations are performed on the abnormal event data in the historical meteorological grid data and corresponding channel data to form an input sequence, and the GD-Informer model is trained and tuned based on the input sequence to obtain a trained GD-Informer model, including: S3-1. Perform temporal alignment operations and spatial alignment operations on the abnormal event data recorded in the historical meteorological grid data and channel data; S3-2. Divide the abnormal event data recorded in the historical meteorological grid data and channel data into training samples and test samples; S3-3. Perform data preprocessing and feature engineering operations on the abnormal event data in the historical meteorological grid data and corresponding channel data to form an input sequence; S3-4. Train, test, and tune the GD-Informer model based on the input sequence to obtain a trained GD-Informer model.

[0026] Specifically, when performing step S3-1, performing temporal alignment operations and spatial alignment operations on the abnormal event data recorded in the historical meteorological grid data and channel data, including: S3-1-1. Perform temporal alignment operations such as timestamp merging, finding common timestamps, and data alignment on the historical meteorological grid data and abnormal event data; S3-1-2. Perform spatial alignment by performing regional division and matching operations on the bit field data in the historical meteorological grid data and abnormal event data; Specifically, when performing step S3-3, performing data preprocessing and feature engineering operations on the abnormal event data in the historical meteorological grid data and corresponding channel data to form an input sequence, including: S3-3-1. Perform data preprocessing operations such as data cleaning, normalization, and data splitting operations on the abnormal event data in the historical meteorological grid data and corresponding channel data; S3-3-2. Perform feature engineering operations such as feature selection, feature construction, and sequence construction on the abnormal event data in the historical meteorological grid data and corresponding channel data; Specifically, when performing step S3-3-2, performing feature engineering operations such as feature selection, feature construction, and sequence construction on the abnormal event data in the historical meteorological grid data and corresponding channel data, including: S3-3-2-1. Feature selection: Based on the abnormal event data in the historical meteorological grid data and corresponding channel data, evaluate the correlation between each meteorological feature and the abnormal event, and eliminate redundant or irrelevant features to improve the prediction performance and interpretability of the model; S3-3-2-2, Feature Construction: Create interaction terms by combining different features based on historical meteorological grid data; S3-3-2-3, Sequence Construction: For historical meteorological grid data, define a time window and then slide this window over the sequence to create fixed-length subsequences as input sequences for subsequent training of the GD-Informer model; Specifically, when performing step S3-4, the GD-Informer model is trained, tested, and tuned based on the input sequence to obtain a trained GD-Informer model, including: S3-4-1, Build a G-DAE denoising autoencoder based on GRU, using GRU as the encoder (encode) and decoder (decode), effectively learning robust and representative latent features from time series data, not only reducing the data dimension, reducing the impact of noise and outliers, but also improving the generalization ability of the model, preventing overfitting, and making the model more focused on the key information of the data.

[0027] Among them, the calculation process of the encoder (Encoder)-GRU: Reset gate: ; Update gate: ; Candidate hidden state: ; Hidden state: , where represents the output of the reset gate at time step t, represents the input vector to the reset gate weight matrix, represents the previous hidden state to the reset gate weight matrix, represents the bias vector of the reset gate, represents the output of the update gate at time step t, represents the input vector to the update gate weight matrix, represents the previous hidden state to the update gate weight matrix, represents the bias vector of the update gate, represents the candidate hidden state vector at time step t, represents the input vector to the candidate hidden state weight matrix, Denote the hidden state at the previous moment after the reset gate processing to the candidate hidden state weight matrix, Denote the bias vector of the candidate hidden state, is the sigmoid function, tanh is the hyperbolic tangent function, Denote element-wise multiplication; Final encoding: After T time steps of calculation, the final hidden state is the encoded representation of the input noise sequence, denoted as c = ; Among them, the calculation process of the decoder (Decoder)-GRU: Initial state: Use the final hidden state c of the encoder as the initial state of the decoder ; At each time step t of the decoder, according to the current hidden state generate the output , and the calculation process is similar to that of the encoder: Reset gate: ; Update gate: ; Candidate hidden state: ; Hidden state: ; Output: , after T time steps, obtain the reconstructed sequence ={ , ,…, }; Among them denotes the output of the reset gate at time step t in the d-th layer, denotes the input vector in the d-th layer to the reset gate weight matrix, denotes the input vector at time step t-1, denotes the hidden state at the previous moment in the d-th layer to the reset gate weight matrix, denotes the hidden state vector at time step t-1 in the d-th layer, denotes the bias vector of the reset gate in the d-th layer, denotes the output of the update gate at time step t in the d-th layer, denotes the input vector in the d-th layer to the update gate weight matrix, Denote the hidden state at the previous moment in the d-th layer to the update gate weight matrix of, Denote the bias vector of the update gate in the d-th layer, Denote the input vector in the d-th layer to the candidate hidden state weight matrix, Denote the hidden state at the previous moment after passing through the reset gate in the d-th layer to the candidate hidden state weight matrix of, Denote the bias vector of the candidate hidden state in the d-th layer, Denote the hidden state in the d-th layer to the output weight matrix of, Denote the bias vector of the output in the d-th layer.

[0028] S3-4-2. Initially train the G-DAE model based on the abnormal event data in the historical meteorological grid data and the corresponding channel data. In the original data ={ ,…, ,…, }, where ∈ , is the sequence length, d is the feature dimension of each time step. Add noise to the original data with a certain probability to obtain the noisy data ={ }; Perform encoding and decoding on to obtain the final result . During the whole process, minimize the difference between the reconstructed sequence and the original clean sequence as the goal, that is, min = + , denotes the regularization parameter, represents penalizing the sum of squares of the model parameters, denotes the Frobenius norm. Use the Adam optimizer to minimize the loss function . Update the parameters (weight matrices W, U and bias vector b) of the GRU in the encoder and decoder through the backpropagation algorithm, and continuously iterate the training until the loss function converges or reaches the predetermined number of training times; S3-4-3. Take G-DAE as the input layer and replace the embedding layer of the Informer model to form the GD-Informer model. The overall structure is asFigure 2 As shown, the trained GD-Informer model is obtained by training, testing, and tuning the GD-Informer model using the input sequence formed by processing the historical meteorological grid data and the abnormal event data in the corresponding channel data: Among them, the key component is ProbSparse Attention (reducing computational complexity): , In the formula ∈ , is the th query vector, K belongs to , K is the key matrix ( is the sequence length), represents the key vector, represents the dimension of the key vector, and the top-u queries are selected (u = , c is a constant); Distillation layer, reducing the sequence length: , where represents a one-dimensional convolution operation, represents a non-linear activation function, represents the original sequence data input to the operation sequence, and the sequence length is compressed to 1 / 2 through convolution; Model training and tuning: 1. Two-stage training strategy:

[0029] Pre-training G-DAE: = + , : L2 regularization coefficient, Joint fine-tuning of GD-Informer: = , , represents the prediction loss, represents the prediction range length, represents the actual observed value at time point t, represents the output prediction value of the model at time point t, represents the total loss function, : Reconstruction and prediction loss weight (initially 0.5, dynamically decaying), 2. Dynamic learning: , represents the learning rate obtained over time, represents the decay rate, , Represents the minimum value of the learning rate decay, and E is the total number of training rounds; In some embodiments, when performing step S4, real-time prediction is performed on new meteorological data based on the trained GD-Informer model to obtain a prediction result, and result analysis and early warning of abnormal situations are performed based on the prediction result, the channel-to-region mapping relationship model, and the different-to-different matching relationship model, including: S4-1. Perform real-time prediction on new meteorological data based on the trained GD-Informer model to obtain a prediction result; S4-2. After obtaining the prediction result, use the channel-to-region mapping relationship model and the different-to-different matching relationship model to perform result analysis and early warning of abnormal situations; Specifically, when performing step S4-2, after obtaining the prediction result, use the channel-to-region mapping relationship model and the different-to-different matching relationship model to perform result analysis and early warning of abnormal situations, including: S4-2-1. The result analysis is to analyze based on the prediction result and the two relationship models to form a heat map and a case report. The heat map mainly shows the predicted distribution of abnormal situations in transmission channels in different regions. The case report is compiled for important prediction events, recording the event process and subsequent actions; S4-2-2. The early warning of abnormal situations mainly performs threshold early warning for the impact degrees of different abnormal events. When the impact degree reaches the preset threshold, the corresponding emergency procedure is started, and real-time push is performed through the SMS and enterprise-level message push platforms to notify the on-duty personnel to perform corresponding processing. The push content includes time, weather, transmission channel, affected area, and impact degree, etc.

[0030] In an embodiment of the present invention, a monitoring and early warning system for abnormal situations in transmission channels is further provided, including: A data acquisition module, which obtains channel data based on the power grid resource business middle platform, obtains regional data based on the power grid power consumption information acquisition system, and obtains historical meteorological grid data based on the meteorological center; A relationship model construction module, which constructs a channel-to-region mapping relationship model and a different-to-different matching relationship model based on the obtained channel data and regional data; A model training module, which performs time series alignment and spatial alignment on the abnormal event data recorded in the historical meteorological grid data and the channel data, performs data preprocessing and feature engineering operations on the abnormal event data in the historical meteorological grid data and the corresponding channel data to form an input sequence, and trains and tunes the GD-Informer model based on the input sequence to obtain a trained GD-Informer model; Result determination and output module, which performs real-time prediction on new meteorological data based on the trained GD-Informer model to obtain prediction results, and then conducts result analysis based on the prediction results, channel-to-region mapping relationship model, and different-to-different matching relationship model, and issues early warnings about abnormal situations of transmission channels and regional abnormal situations through text messages and enterprise-level message push platforms.

[0031] Although the embodiments of the present invention have been described in detail above, it is obvious to those skilled in the art that various modifications and changes can be made to these embodiments. However, it should be understood that such modifications and changes fall within the scope and spirit of the present invention as described in the claims. Moreover, the present invention described herein can have other embodiments and can be implemented or realized in various ways.

Claims

1. A method for monitoring and early warning of abnormal conditions in a power transmission channel, characterized in that: The following steps are involved: Based on the power grid resource business middle station, channel data is obtained, and the channel data includes transmission channel spatial data and recorded abnormal event data; based on the power grid power consumption information collection system, regional data is obtained; based on the meteorological center, historical meteorological grid data is obtained, and the meteorological grid data includes meteorological data and grid data; Construct channel-to-region mapping relationship model and heterogeneous-to-heterogeneous matching relationship model based on channel data and regional data; After the abnormal event data recorded in the historical meteorological raster data and the channel data are aligned in time series and space, data preprocessing and feature engineering operations are performed on the abnormal event data in the historical meteorological raster data and the corresponding channel data to form an input sequence. The GD-Informer model is trained and tuned based on the input sequence to obtain a trained GD-Informer model; Based on the trained GD-Informer model, real-time prediction of new meteorological data is performed to obtain prediction results. Based on the prediction results, channel-to-region mapping relationship model and heterogeneous-to-different matching relationship model, result analysis and abnormal situation warning are carried out.

2. A method for monitoring and early warning abnormal conditions in a power transmission channel according to claim 1, characterized in that: The meteorological data include temperature, humidity, wind speed, wind direction, and precipitation, and the grid data include predicted time, reported time, longitude, and latitude.

3. A method for monitoring and early warning abnormal conditions in a power transmission channel according to claim 1, characterized in that: Based on the channel data and regional data, a channel-to-region mapping relationship model and a heterogeneous-to-heterogeneous matching relationship model are constructed, including: Based on channel data and regional data, a channel-to-region mapping relationship model is constructed by identifying key points and channels, analyzing power grid topology relationships, establishing influence matrices and parameterized relationships; Based on channel data and regional data, a heterogeneous matching relationship model is constructed by determining the fault type and adverse impact type, establishing a relationship matrix and parameterizing the relationship.

4. A method for monitoring and early warning abnormal conditions in a power transmission channel according to claim 1, characterized in that: After the abnormal event data recorded in the historical meteorological raster data and the channel data are aligned in time series and space, data preprocessing and feature engineering operations are performed on the abnormal event data in the historical meteorological raster data and the corresponding channel data to form an input sequence. The GD-Informer model is trained and tuned based on the input sequence to obtain a trained GD-Informer model, including: Perform temporal alignment and spatial alignment operations on the abnormal event data recorded in the historical meteorological raster data and channel data; The abnormal event data recorded in the historical meteorological raster data and channel data are divided into training samples and test samples; Performing data preprocessing and feature engineering operations on abnormal event data in historical meteorological raster data and corresponding channel data to form an input sequence, wherein the data preprocessing includes data cleaning, normalization and data segmentation operations, and the feature engineering operations include feature selection, feature construction and sequence construction; The GD-Informer model is trained, tested and tuned based on the input sequence to obtain a trained GD-Informer model.

5. A method for monitoring and early warning abnormal conditions in a power transmission channel according to claim 1, characterized in that: Based on the trained GD-Informer model, the new meteorological data is predicted in real time to obtain the prediction results. Based on the prediction results, the channel-to-region mapping relationship model and the heterogeneous-to-different matching relationship model, the results are analyzed and abnormal situation warnings are issued, including: Based on the trained GD-Informer model, the acquired meteorological data is predicted in real time to obtain the prediction results; After obtaining the prediction results, the channel-to-region mapping relationship model and the heterogeneous-to-different matching relationship model are used to analyze the results and warn of abnormal situations.

6. A method for monitoring and early warning abnormal conditions in a power transmission channel according to claim 4, characterized in that: Perform temporal alignment and spatial alignment operations on the abnormal event data recorded in the historical meteorological raster data and channel data, including: Timing alignment operations include timestamp merging, finding common timestamps, and data alignment; The spatial alignment operation performs region division and matching based on the bit field data in the raster data and the abnormal event data.

7. A method for monitoring and early warning abnormal conditions in a power transmission channel according to claim 5, characterized in that: After obtaining the prediction results, the channel-to-region mapping relationship model and the heterogeneous-to-different matching relationship model are used to analyze the results and warn of abnormal situations, including: The result analysis is to form a heat map and case report based on the prediction results by using the channel-to-region mapping relationship model and the heterogeneous-to-heterogeneous matching relationship model. The heat map shows the predicted distribution of abnormal conditions in transmission channels in different regions. The case report prepares a detailed case report for the predicted event, recording the event process and subsequent actions. Abnormal situation warning is a threshold warning for the impact degree of different abnormal events. When the impact degree reaches the preset threshold, the corresponding emergency procedure is activated and pushed in real time through SMS and enterprise-level message push platform to notify the on-duty personnel to take corresponding measures. The push content includes time, weather, transmission channel, affected area and impact degree.

8. A power transmission channel abnormality monitoring and early warning system, used to implement the power transmission channel abnormality monitoring and early warning method according to any one of claims 1 to 7, characterized in that: include: The data acquisition module acquires channel data based on the power grid resource business center, acquires regional data based on the power grid power consumption information acquisition system, and acquires historical meteorological grid data based on the meteorological center; A relationship model building module builds a channel-to-region mapping relationship model and a heterogeneous-to-heterogeneous matching relationship model based on the acquired channel data and region data; The model training module performs temporal and spatial alignment on the abnormal event data recorded in the historical meteorological raster data and the channel data, performs data preprocessing and feature engineering operations on the abnormal event data in the historical meteorological raster data and the corresponding channel data to form an input sequence, and trains and tunes the GD-Informer model based on the input sequence to obtain a trained GD-Informer model; The result judgment and output module performs real-time prediction of new meteorological data based on the trained GD-Informer model to obtain the prediction results, and then analyzes the results based on the prediction results, the channel-to-region mapping relationship model, and the heterogeneous-to-different matching relationship model, and warns of abnormal conditions in the transmission channel and the region through SMS and enterprise-level message push platforms.

Images