Line disconnection probability prediction and future dispatch effect analysis method, system and product based on graph neural network

Abstract

The present application discloses a method, system, and product for predicting the probability of line disconnection and analyzing future state dispatching effects based on a graph neural network, belonging to the technical field of power line systems, including: obtaining historical regional climate data and the first tower body feature data; constructing a tower local climate data prediction model; using the tower local climate data prediction model to predict the tower local climate data; obtaining the second tower body feature data and tower operation and maintenance information, constructing a tower collapse probability prediction model, and obtaining the probability of each tower collapse based on the tower collapse probability prediction model; determining the disconnection probability of the entire line; constructing a digital effect analysis model for future state dispatching of the power grid under extreme climate conditions, and analyzing the loss of future state dispatching to achieve analysis of the future state dispatching effects of the power grid under extreme weather conditions. The method of the present application can improve the fault prediction capability of the power system under extreme climate conditions, filling the gap in the analysis of future state dispatching effects of the power grid.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a method, system, and product for predicting line disconnection probability and analyzing future dispatch effects based on graph neural networks. Background Art

[0002] With the increasing frequency of extreme weather events, catastrophic weather events such as typhoons and heavy rains pose a serious threat to the operational safety of power systems. Existing power grid operation and dispatching systems typically rely on regional meteorological data provided by meteorological authorities for risk assessment, coupled with a small number of meteorological monitoring devices deployed at key towers for local environmental sensing and response. However, due to the widespread distribution of towers and insufficient sensor coverage, it is often difficult to obtain real-time microclimate data around each tower, resulting in significant bias in the dispatching model's risk assessment. Existing technologies estimate the probability of transmission line failure based on regional meteorological data and historical fault statistics. These technologies use regional meteorological data as input and output the outage probability of the entire line. These technologies fail to account for the spatial variability of meteorological conditions between towers, resulting in limited prediction accuracy and an inability to precisely control the source of risk. After predicting fault risks, the power grid can implement proactive dispatching to achieve "future-state dispatching." However, there is currently a lack of systematic methods to evaluate the effectiveness of digital measures for future-state dispatching, making it impossible to quantitatively analyze the improvements in dispatching strategies and loss reductions before and after the implementation of future-state dispatching. This hinders the project's widespread adoption and optimization. Summary of the Invention

[0003] The purpose of this application is to overcome the defects of the existing technology and provide a method, system and product for predicting the probability of line disconnection and analyzing the future state scheduling effect based on graph neural network, so as to solve the problems of insufficient prediction accuracy of the probability of transmission line disconnection under extreme climatic conditions and the inability to quantitatively evaluate the future state scheduling effect, and enhance the ability of the power grid to cope with extreme climate.

[0004] In a first aspect, the present application provides a method for predicting line disconnection probability and analyzing future scheduling effects based on a graph neural network, comprising the following steps:

[0005] Obtain historical regional climate data and the first tower's characteristic data;

[0006] Constructing a tower local climate data prediction model based on the historical regional climate data and the first tower body characteristic data;

[0007] Using the tower local climate data prediction model to predict key meteorological indicators of the tower to obtain tower local climate data;

[0008] Acquiring characteristic data of a second tower and tower operation and maintenance information, constructing a tower collapse probability prediction model based on the characteristic data of the second tower, the tower operation and maintenance information, and local climate data of the tower, and obtaining a probability of tower collapse for each tower based on the tower collapse probability prediction model;

[0009] Based on the probability of each tower collapsing, the probability of disconnection of the entire transmission line is determined by a probability combination model;

[0010] A digital effect analysis model for future-state scheduling of power grids under extreme climate conditions is constructed, and based on the digital effect analysis model for future-state scheduling of power grids under extreme climate conditions, losses under future-state scheduling are analyzed to achieve analysis of the effects of future-state scheduling of power grids under extreme weather conditions.

[0011] Optionally, constructing a tower local climate data prediction model based on the historical regional climate data and the first tower body characteristic data includes:

[0012] Acquire actual local climate data of the tower, set the actual local climate data of the tower as first target output data, and combine historical regional climate data and first tower body characteristic data to obtain first input data;

[0013] Preprocessing the first input data and the first target output data;

[0014] Converting the preprocessed first input data into a graph structure;

[0015] Build the first graph neural network;

[0016] The first input data of the graph structure and the first target output data after preprocessing are used to train the first graph neural network, and the first loss function is used to optimize and update the parameters to obtain a tower local climate data prediction model.

[0017] Optionally, the tower local climate data prediction model is used to predict the key meteorological indicators of the tower, and the predicted value of the tower local climate data at time t is The expression is:

[0018]

[0019] in, is the adjacency matrix, is the first input feature matrix, Represents the tower local climate data prediction model.

[0020] Optionally, obtaining characteristic data of a second tower and tower operation and maintenance information, constructing a tower collapse probability prediction model based on the characteristic data of the second tower, the tower operation and maintenance information, and the local climate data of the tower, and obtaining the probability of tower collapse of each tower based on the tower collapse probability prediction model, including:

[0021] Obtaining the second tower body characteristic data and tower operation and maintenance information;

[0022] Combining the second tower body characteristic data, the tower operation and maintenance information, and the tower local climate data to obtain a second input data set;

[0023] Preprocessing the second input data set;

[0024] Converting the preprocessed second input data set into a graph structure;

[0025] Build the second graph neural network;

[0026] The second graph neural network is trained using a second input data set with a graph structure, and the parameters are optimized and updated using a second loss function to obtain a tower collapse probability prediction model.

[0027] Optionally, based on the probability of each tower falling, the probability of the entire transmission line being disconnected is determined by a probability combination model, and the total probability of the transmission line being disconnected is It can be expressed as:

[0028]

[0029] in, express t Moment i The predicted value of the tower collapse probability of a tower.

[0030] Optionally, a digital effect analysis model for future-state scheduling of the power grid under extreme weather conditions is constructed, and losses under future-state scheduling are analyzed based on the digital effect analysis model for future-state scheduling of the power grid under extreme weather conditions, thereby analyzing the effect of future-state scheduling of the power grid under extreme weather conditions, including:

[0031] Constructing a digital effect analysis model for future power grid dispatch under extreme climate conditions, the digital effect analysis model for future power grid dispatch under extreme climate conditions comprising: a line disconnection scenario generation module, an input module, a dispatch analysis module, and a dispatch effect analysis module;

[0032] Inputting the disconnection probability of each transmission line in the power grid into the disconnection scenario generation module to obtain the disconnection scenario corresponding to each transmission line in the power grid;

[0033] The disconnection scenarios corresponding to all transmission lines in the power grid and the data in the input module are fed into the dispatch analysis module, which includes: a generator unit combination model before extreme weather occurs, a traditional dispatch model under extreme weather, and a future dispatch model under extreme weather;

[0034] Obtaining a unit scheduling plan before the occurrence of extreme weather through the generator unit combination model before the occurrence of extreme weather in the scheduling analysis module;

[0035] Obtaining a traditional scheduling solution under extreme climate conditions by using the traditional scheduling model under extreme climate conditions in the scheduling analysis module;

[0036] Obtaining a future-state scheduling plan under extreme climate conditions through the future-state scheduling model under extreme climate conditions in the scheduling analysis module;

[0037] The unit dispatching plan before the extreme climate occurs, the traditional dispatching plan under the extreme climate, and the future dispatching plan under the extreme climate are input into the dispatching effect analysis module for effect analysis. The power grid loss and social loss under the future dispatching are determined based on the probability of transmission line disconnection and the number of disasters, and the effect of the future dispatching of the power grid under extreme climate is obtained.

[0038] Optionally, obtaining a unit scheduling plan before the occurrence of extreme weather by using the generator unit combination model before the occurrence of extreme weather in the scheduling analysis module includes: analyzing the generator unit combination model before the occurrence of extreme weather in the scheduling analysis module and optimizing using a first objective function to obtain a unit scheduling plan before the occurrence of extreme weather, wherein the first objective function The expression is as follows:

[0039]

[0040] in, For the i Output quotation function of each unit; For the i Start-up cost of each unit; For the i No-load operating cost of each unit; For the ne The power generation cost of each new energy unit; For the i units t Whether to start during the period, if t 1 if the period is started, 0 otherwise; For the i units t Whether to shut down during the period, if t If the period is down, it is 1, otherwise it is 0; For the i Units in t Whether the time period is put into operation, if it is put into operation, it is 1, if not put into operation, it is 0; For the i Units in t Output during the time period; For new energy units t Actual dispatch output during the time period; for t Time period d load value; For the s The cross section t The amount of electricity required to be sent or received during the time period; The first and last nodes of the line are nm The maximum transmission capacity of the line; For the i The vector form of the minimum output of each unit; For the i Units in t The vector form of the output during the time period; For the i Units in t-1 The vector form of the output during the time period; For the i The vector form of the maximum output of each unit; For the ne New energy units in t The vector form of the output during the time period; For the ne New energy units in t The vector form of the maximum output value predicted for the time period; For conventional generator sets; For the i The downward ramp rate limit of each unit; For the i The upward climbing rate limit of each unit; For the n The transmission line where the node is located; is the set of all transmission lines in the power grid; For the n Nodes in t The node voltage phase angle of the time period; For the m Nodes in t The node voltage phase angle of the time period; for t The beginning and end nodes of the time period line are nm lines; T The total time period of the unit dispatch plan before the extreme weather occurs; For the i The start-up period of each unit; For the i downtime of each unit; NE Assemble for new energy generator sets; For the n Nodes of new energy generators; D is the node load collection in the power grid.

[0041] Optionally, obtaining a traditional scheduling scheme under extreme climate conditions through the traditional scheduling model under extreme climate conditions in the scheduling analysis module includes: analyzing through the traditional scheduling model under extreme climate conditions in the scheduling analysis module and optimizing using a second objective function to obtain a traditional scheduling scheme under extreme climate conditions, wherein the second objective function , the expression is as follows:

[0042]

[0043] in, For the l The second constraint condition when a line fails; For the i Output quotation function of each unit; For the j A unit that can quickly supply capacity in a short time t Output quotation function for a certain period of time; For the ne The power generation cost of each new energy unit; For the j A unit that can quickly supply capacity in a short time t Output during the time period; For the i Units in t Output during the time period; For new energy units t Actual dispatch output during the time period; for t Time period d load value; For the s The cross section t The amount of electricity required to be sent or received during the time period; For the i The vector form of the minimum output of each unit; For the i Units in t The vector form of the output during the time period; For the i Units in t-1 The vector form of the output during the time period; For thei The vector form of the maximum output of each unit; For the ne New energy units in t The vector form of the output during the time period; For the ne New energy units in t The vector form of the maximum output value predicted for the time period; The first and last nodes of the line are nm The maximum transmission capacity of the line; For conventional generator sets; A collection of spare capacity units with short call times; For the n A collection of conventional generator sets of nodes; For the i The downward ramp rate limit of each unit; For the i The upward climbing rate limit of each unit; For the n Nodes in t The node voltage phase angle of the time period; For the m Nodes in t The node voltage phase angle of the time period; for t The beginning and end nodes of the time period line are nm lines; T is the total time period of the traditional scheduling scheme under extreme climate conditions; For the l After the first line fails, remove the l The collection of all other lines.

[0044] Optionally, obtaining a future state scheduling scheme under extreme climate by using the future state scheduling model under extreme climate in the scheduling analysis module includes: analyzing the future state scheduling model under extreme climate in the scheduling analysis module and optimizing using a third objective function to obtain a future state scheduling scheme under extreme climate, wherein the third objective function The expression is as follows:

[0045]

[0046] in, For the i Output quotation function of each unit; For the j A unit that can quickly supply capacity in a short time t Output quotation function for a certain period of time; For the ne The power generation cost of each new energy unit; For the k A slow standby unit is t Output quotation function for a certain period of time; For the i Units in t Output during the time period; For the j A unit that can quickly supply capacity in a short time t Output during the time period; For the k A slow standby unit is t Output during the time period; For new energy units t Actual dispatch output during the time period; for t Time period d load value; For the s The cross section t The amount of electricity required to be sent or received during the time period; For the i The vector form of the minimum output of each unit; For the i Units in t The vector form of the output during the time period; For the i Units in t-1 The vector form of the output during the time period; For the i The vector form of the maximum output of each unit; For the ne New energy units in t The vector form of the output during the time period; For the ne New energy units in t The vector form of the maximum output value predicted for the time period; For the ne The minimum value between the actual historical output value and the predicted value of each new energy unit; For the ne The actual historical output value of each new energy unit; For the ne Historical output forecast value of each new energy generating unit; The first and last nodes of the line are nm The maximum transmission capacity of the line; For the i The downward ramp rate limit of each unit; For the i The upward climbing rate limit of each unit; For the n Nodes in tThe node voltage phase angle of the time period; For the m Nodes in t The node voltage phase angle of the time period; for t The beginning and end nodes of the time period line are nm lines; T is the total time period of the future scheduling plan under extreme climate conditions; NE Assemble for new energy generator sets; For the n Nodes of new energy generators; For the l After the first line fails, remove the l The collection of all other lines of the line; For conventional generator sets; A collection of spare capacity units with short call times; A collection of slow standby capacity units with a longer call time; For the n A collection of conventional generator sets of nodes; G for G 1 、 G 2 、 G 3 The intersection of three sets; D is the node load set in the power grid; D n For the n The load of a node.

[0047] Optionally, the power grid loss under the future state scheduling The expression is:

[0048]

[0049] in, F is the annual occurrence of extreme climate; N is the total number of failure types under a certain extreme climate; T is the total time period of the traditional scheduling scheme under extreme climate conditions; a The type of failure under certain extreme climate conditions; For the a Failure probability of each fault type; is the cost function of the standby unit; For the a Under the fault type t The required standby unit output for a period of time is given by the scheduling analysis module; For the a Under the fault type tThe power supply shortage of the power system during the period; The reserve capacity of the power grid; is the power outage compensation coefficient per unit power;

[0050] The social loss under the future scheduling The expression is:

[0051]

[0052] in, F is the annual occurrence of extreme climate; N is the total number of failure types under a certain extreme climate; T is the total time period of the traditional scheduling scheme under extreme climate conditions; a The type of failure under certain extreme climate conditions; For the a Failure probability of each fault type; For the a Under the fault type t The required standby unit output for a period of time is given by the scheduling analysis module; For the a Under the fault type t The power supply shortage of the power system during the period; The reserve capacity of the power grid.

[0053] In the second aspect, the present application also provides a line disconnection probability prediction and future scheduling effect analysis system based on graph neural network, which is used to execute the line disconnection probability prediction and future scheduling effect analysis method based on graph neural network as described in any one of the first aspects, including: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the line disconnection probability prediction and future scheduling effect analysis method based on graph neural network as described in any one of the first aspects.

[0054] In a third aspect, the present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the line disconnection probability prediction and future scheduling effect analysis method based on graph neural network as described in any one of the first aspects.

[0055] The present application provides a method for predicting the probability of line disconnection and analyzing the effect of future state scheduling based on graph neural network. By constructing a local climate data prediction model for towers, key meteorological indicators are accurately estimated, meteorological influencing factors can be quantified, and the microclimate environment of a single tower can be accurately described, providing high-precision input data for the subsequent calculation of the probability of tower collapse. By constructing a tower collapse probability prediction model, the limitation of a single factor is broken, making the calculation of the probability of tower collapse more in line with reality. Through the probability combination model, the overall line disconnection probability is derived from the individual probability of the tower, which can capture the individual differences of the single tower in structure and maintenance status, and realize scientific evaluation from micro to macro. The prediction model can output the probability of line disconnection of the entire line with high precision by inputting only regional meteorological data, taking into account both practicality and prediction accuracy, and significantly improving the fault prediction ability of the transmission system under extreme climate conditions; by constructing a digital effect analysis model of future-state scheduling of power grids under extreme climate conditions, the advantages of future-state scheduling in emergency response efficiency and economic cost are quantified, filling the gap in the existing technology in the lack of analysis of the effect of future-state scheduling of power grids under extreme climate conditions, quantifying the impact of future-state scheduling on power grid losses and social losses, and providing a scientific decision-making basis for whether the power grid should invest in future-state scheduling, promoting the power grid scheduling under extreme climate conditions from passive response to active optimization. The method of this application systematically improves the practicality and accuracy of power grid risk prediction and scheduling decision-making under extreme climate conditions through hierarchical modeling and digital evaluation. Through data-driven and model-supported, it realizes a closed loop from line disconnection risk prediction to scheduling effect evaluation, providing a strong basis for power grid operation and maintenance personnel to formulate protection strategies and optimize scheduling plans in advance, improving the power grid's ability to cope with extreme climates, and effectively reducing the harm of extreme weather to power grid operations.

[0056] In order to make the above features and advantages of the present invention more clearly understood, embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0058] Picture 1 This is a flowchart of a method for predicting line disconnection probability and analyzing future scheduling effects based on a graph neural network provided in one embodiment of the present application.

[0059] Picture 2This is a flowchart of step S2 in the method for predicting line disconnection probability and analyzing future scheduling effects based on graph neural networks provided in one embodiment of the present application.

[0060] Picture 3 This is a flowchart of step S4 in the method for predicting line disconnection probability and analyzing future scheduling effects based on graph neural networks provided in one embodiment of the present application.

[0061] Picture 4 This is a flowchart of step S6 in the method for predicting line disconnection probability and analyzing future scheduling effects based on graph neural networks provided in one embodiment of the present application.

[0062] Picture 5 This is a structural diagram of a digital effect analysis model for future-state scheduling of power grids under extreme climate conditions in a method for predicting line disconnection probability and analyzing future-state scheduling effects based on a graph neural network provided in one embodiment of the present application. DETAILED DESCRIPTION

[0063] To make the purpose and technical solutions of the embodiments of the present application clearer, the technical solutions of the embodiments of the present application will be clearly and completely described below in conjunction with the drawings of the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the described embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0064] In one embodiment, see Picture 1 The present application provides a method for predicting the probability of line disconnection and analyzing the effect of future-state scheduling based on a graph neural network. The method for predicting the probability of line disconnection and analyzing the effect of future-state scheduling based on a graph neural network may include the following steps: step S1 to step S6.

[0065] Step S1: Obtain historical regional climate data and first tower body characteristic data.

[0066] Step S2: constructing a tower local climate data prediction model based on the historical regional climate data and the first tower body characteristic data.

[0067] Step S3: using the tower local climate data prediction model to predict key meteorological indicators of the tower to obtain tower local climate data.

[0068] Step S4: Obtain the second tower body characteristic data and tower operation and maintenance information, construct a tower collapse probability prediction model based on the second tower body characteristic data, the tower operation and maintenance information, and the tower local climate data, and obtain the probability of tower collapse for each tower based on the tower collapse probability prediction model.

[0069] Step S5: Based on the probability of each tower collapsing, the probability of disconnection of the entire transmission line is determined by a probability combination model.

[0070] Step S6: Construct a digital effect analysis model for future state scheduling of the power grid under extreme climate conditions, analyze the losses under future state scheduling based on the digital effect analysis model for future state scheduling of the power grid under extreme climate conditions, and realize the analysis of the future state scheduling effect of the power grid under extreme weather conditions.

[0071] In the graph neural network-based line disconnection probability prediction and future dispatch effect analysis method of the present application, by constructing a local climate data prediction model for the tower, key meteorological indicators are accurately estimated, meteorological influencing factors can be quantified, and the microclimate environment of a single tower is accurately portrayed, providing high-precision input data for the subsequent tower collapse probability calculation; by constructing a tower collapse probability prediction model, the limitation of a single factor is broken, making the tower collapse probability calculation more in line with reality; through the probability combination model, the overall line disconnection probability is derived from the individual probability of the tower, which can capture the individual differences of the single tower in structure and maintenance status, and realize scientific evaluation from micro to macro; by constructing a tower collapse probability prediction model The prediction model can output the probability of line disconnection of the entire line with high precision by inputting only regional meteorological data, taking into account both practicality and prediction accuracy, and significantly improving the fault prediction ability of the transmission system under extreme climate conditions; by constructing a digital effect analysis model of future-state scheduling of power grids under extreme climate conditions, the advantages of future-state scheduling in emergency response efficiency and economic cost are quantified, filling the gap in the existing technology for the analysis of the effect of future-state scheduling of power grids under extreme climate conditions, quantifying the impact of future-state scheduling on power grid losses and social losses, and providing a scientific decision-making basis for whether the power grid should invest in future-state scheduling, promoting the power grid scheduling under extreme climate conditions from passive response to active optimization. The method of this application systematically improves the practicality and accuracy of power grid risk prediction and scheduling decision-making under extreme climate conditions through hierarchical modeling and digital evaluation. Through data-driven and model-supported, it realizes a closed loop from line disconnection risk prediction to scheduling effect evaluation, providing a strong basis for power grid operation and maintenance personnel to formulate protection strategies and optimize scheduling plans in advance, and effectively reducing the harm of extreme weather to power grid operation.

[0072] In step S1, see Picture 1 In step S1, historical regional climate data and first tower body characteristic data are obtained.

[0073] As an example, historical regional climate data for the area where the power grid is located can be obtained from the meteorological department's database or data interface. ,in, , t For time, d is the regional climate characteristic dimension, gIndicates the g Nearby areas.

[0074] As an example, historical regional climate data It may include: regional wind speed, regional wind direction, regional rainfall, regional temperature, regional humidity, date and time, etc.

[0075] As an example, the first tower body feature data can be obtained by combining the design documents and construction records of the tower construction with the Geographic Information System (GIS) technology. ,in, , m is the characteristic dimension of the tower body characteristic data, i Indicates the i Root tower.

[0076] As an example, the first tower body feature data It indicates the characteristics of the area where the tower is located and the tower structure, and may include information such as tower location, tower height, and tower altitude.

[0077] In step S2, see Picture 1 In step S2, a tower local climate data prediction model is constructed based on the historical regional climate data and the first tower body characteristic data.

[0078] As an example, see Picture 2 , step S2 may include the following steps: step S21 to step S25.

[0079] Step S21: obtaining actual local climate data of the tower, setting the actual local climate data of the tower as first target output data, combining historical regional climate data and first tower body characteristic data to obtain first input data.

[0080] Step S22: Preprocess the first input data and the first target output data.

[0081] Step S23: Convert the preprocessed first input data into a graph structure.

[0082] Step S24: Construct a first graph neural network.

[0083] Step S25: Use the first input data of the graph structure and the preprocessed first target output data to train the first graph neural network, use the first loss function to optimize and update the parameters, and obtain a tower local climate data prediction model.

[0084] As an example, in step S21, the actual local climate data of the tower is obtained through the sensor. ,in, express t Momenti Actual local climate condition data of the tower, n is the local climate characteristic dimension.

[0085] As an example, the actual local climate data of the tower It may include: actual local wind speed of the tower, actual local wind direction of the tower, actual local rainfall of the tower, actual local temperature of the tower, actual local humidity of the tower, etc.

[0086] Furthermore, the actual local climate data of the tower Set it into matrix form and get the first target output data matrix , the expression is:

[0087]

[0088] in, express t The actual local climate condition data of the first tower at the moment, express t The actual local climate condition data of the second tower at the moment, express t The actual local climate condition data of the Nth tower at the moment, N Indicates the number of towers in a certain area. n is the local climate characteristic dimension.

[0089] Furthermore, historical regional climate data And the first tower body characteristic data Combine historical regional climate data Copy to each tower, then i The input eigenvector of the tower is ,in, m is the characteristic dimension of the tower body characteristic data, d is the regional climate characteristic dimension.

[0090] Furthermore, the input feature vectors of all towers Combined into matrix form, the first input data is obtained ,in, N Indicates the number of towers in a certain area. m is the characteristic dimension of the tower body characteristic data, d is the regional climate characteristic dimension. Constitute the first input data set X.

[0091] As an example, in step S22, the first input data and the first target output data are normalized, aiming to eliminate the impact of differences in feature scales on model training by standardizing or normalizing the input data, so that each feature is processed at the same scale, thereby improving the efficiency and stability of model training.

[0092] As an example, you can use min-max normalization to scale the data to a fixed interval. You can also set it to compress the data to the [0, 1] interval.

[0093] As an example, the minimum-maximum normalization operation is performed on the first input data set X, and the expression is as follows:

[0094]

[0095] in, After normalization, i Input feature data, The original input data set i data points, For the dataset X The minimum value in For the dataset X The maximum value in .

[0096] Furthermore, each normalized input feature data is combined to obtain the first input matrix ,in, N Indicates the number of towers in a certain area. m is the characteristic dimension of the tower body characteristic data, d is the regional climate characteristic dimension.

[0097] As an example, output data to the first target The specific method of normalization can refer to the specific method of normalizing the first input data in step S22, which will not be repeated here.

[0098] As an example, in step S23, due to the spatial correlation between the geographical locations of the towers, the towers can be used as nodes, and the connection distance is less than the threshold value. With two towers as edges, we get an undirected graph G=(V,E) ,in, is a node set, representing the set of towers; E is an edge set, indicating that the distance is less than the threshold The set of lines connecting two towers.

[0099] As an example, the edge set E The expression is as follows:

[0100]

[0101] in, i Indicates the i Root tower, j Indicates the j Root tower, Indicates the i The tower and the j The distance between towers.

[0102] As an example, a tower i The coordinates are p i =(x i ,y i ) , tower j The coordinates are p j =(x j ,y j ) , then the tower i With tower j The Euclidean distance between them is: .

[0103] Furthermore, according to the edge set E Calculate the adjacency matrix , the expression is:

[0104]

[0105] Furthermore, for the adjacency matrix A Add the identity matrix I , and get the new adjacency matrix , to ensure that each node can use its own information to add a new adjacency matrix after the loop , the expression is:

[0106]

[0107] Furthermore, the new adjacency matrix after the self-loop is added Normalized to a symmetric Laplacian matrix , the expression is:

[0108]

[0109] in, is the node degree matrix of the graph, which is a diagonal matrix whose elements Indicates the i The degree of the node (i.e. iThe number of edges connecting the nodes). Normalized adjacency matrix Ability to adjust the intensity of node information dissemination to make the dissemination more balanced.

[0110] As an example, in step S24, a first graph neural network is constructed, and feature information of each node and its neighboring nodes is aggregated using graph convolution. The first graph neural network may include an input layer, a graph convolution layer, and an output layer.

[0111] As an example, we can set the first graph neural network to have a total of L layers, and the input layer can be expressed as ,in, represents the node feature matrix of the input layer, is the input feature matrix.

[0112] As an example, l The node feature matrix of the layer is , then l +1 layer node feature matrix It can be obtained by the following formula:

[0113]

[0114] in, For the l The node feature matrix of the layer; For the l The weight matrix of the layer represents the parameters of the linear transformation and needs to be learned through training; is the normalized adjacency matrix, which is used to aggregate the information of neighboring nodes to the current node; σ is a non-linear activation function.

[0115] As an example, the nonlinear activation function σ You can use ReLU Activation function to enhance nonlinear expression capabilities.

[0116] For example, stacking multiple graph convolutional layers forms the backbone of the first graph neural network. Each layer expands the receptive field of information, meaning that the final representation of each node can incorporate information from the adjacency matrix over a wider range. For example, the first layer can integrate information from direct neighbors, while the second layer can integrate information from second-order neighbors. By stacking multiple layers, the first graph neural network can capture a wider range of node contexts.

[0117] As an example, the output layer can be expressed as:

[0118]

[0119] in, express tThe predicted value of the local climate of the tower at the moment, For the L The node feature matrix of the layer, that is, the node feature matrix of the last layer; is the weight matrix of the output layer.

[0120] As an example, in step S25, the first graph neural network is trained using the input data of the graph structure, and the weight matrix of each layer is continuously updated to make the predicted value of the local climate of the tower as close as possible to the true value in the first target output, and determine the optimal weight matrix.

[0121] Specifically, the first input data of the graph structure and the first target output data are constructed into a training sample, wherein the first input data of the graph structure may include: the normalized adjacency matrix of the graph structure , input feature matrix The target output data may include: actual local climate data matrix of the tower .

[0122] Furthermore, the first graph neural network is trained using the training samples. First, the parameters of the first graph neural network are initialized, and the first input data of the graph structure is fed into the constructed first graph neural network. The first input data of the graph structure is passed layer by layer according to the structure of the first graph neural network, and relevant features are extracted to obtain a predicted value of the local climate of the tower. Then, the predicted value of the local climate of the tower is compared with the first target output data. The loss value is calculated using the first loss function to evaluate the degree of deviation between the predicted value of the local climate of the tower obtained by the current first graph neural network and the actual local climate of the tower in the first target output data. Then, check whether the current loss value reaches the training target. If so, it means that the prediction accuracy of the first graph neural network has reached the optimal value, and the training is terminated. If not, continue to determine whether the preset maximum number of training times has been reached. If so, terminate the training. If not, perform backpropagation. Based on the obtained loss value, use the backpropagation algorithm to reversely calculate the gradient from the output layer to the input layer to determine the parameters that need to be adjusted in the first graph neural network. Based on the gradient calculated by backpropagation, use a suitable optimization algorithm to update the parameters of the first graph neural network, and then continue forward propagation. Repeat the above training process to optimize the parameters of the first graph neural network in the direction of reducing the loss value until the training target is reached or the preset maximum number of training times is reached. End the training and obtain the final tower local climate data prediction model.

[0123] As an example, the first loss function may adopt the mean square error loss function (MSE), which is expressed as:

[0124]

[0125] in, N Indicates the number of towers in a certain area. For the i Actual local climate data of the tower, For the i The predicted value of the local climate data of the tower, l Indicates the l layer.

[0126] As an example, the training goal is to continuously reduce the loss value of the first loss function until it approaches zero or becomes stable.

[0127] As an example, the update rule for gradient descent is:

[0128]

[0129] in, is the learning rate, For the updated l The weight matrix of the layer, For the l The weight matrix of the layer, is the loss value of the first loss function.

[0130] As an example, the optimization algorithm may adopt a gradient descent method, or adaptive moment estimation (Adam), stochastic gradient descent (SGD), etc.

[0131] As an example, the maximum number of training times may be set based on the performance of the hardware device.

[0132] It is important to note that each iteration uses the current parameters to calculate the forward propagation (i.e., the local climate forecast value of the tower is obtained based on the historical regional climate data) and the loss is calculated. , use the gradient to update the weight parameters of each layer, and iterate repeatedly until the loss value converges (that is, tends to be stable), completing the model training.

[0133] In step S3, see Picture 1 In step S3, the tower local climate data prediction model is used to predict key meteorological indicators of the tower to obtain tower local climate data.

[0134] As an example, the tower local climate data prediction model is used to predict the key meteorological indicators of the tower, obtain the tower local climate data, and realize the prediction of the local key meteorological indicators of each tower, providing a highly reliable data basis for subsequent tower risk assessment.

[0135] As an example, the local climate data prediction value of the tower at time t output by the tower local climate data prediction model is It can be expressed as:

[0136]

[0137] in, is the adjacency matrix, is the first input feature matrix, Represents the tower local climate data prediction model.

[0138] In step S4, see Picture 1 In step S4, the characteristic data of the second tower and the tower operation and maintenance information are obtained, and a tower collapse probability prediction model is constructed based on the second tower characteristic data, the tower operation and maintenance information, and the tower local climate data. The probability of tower collapse of each tower is obtained based on the tower collapse probability prediction model.

[0139] As an example, see Picture 3 , step S4 may include the following steps: step S41 to step S46.

[0140] Step S41: Acquire the second tower body characteristic data and tower operation and maintenance information.

[0141] Step S42: combining the second tower body characteristic data, the tower operation and maintenance information, and the tower local climate data to obtain a second input data set.

[0142] Step S43: pre-processing the second input data set.

[0143] Step S44: converting the preprocessed second input data set into a graph structure.

[0144] Step S45: Construct a second graph neural network.

[0145] Step S46: Use the second input data set of the graph structure to train the second graph neural network, use the second loss function to optimize and update the parameters, and obtain a tower collapse probability prediction model.

[0146] As an example, in step S41, the second tower body characteristic data can be obtained by combining the design documents and construction records of the tower construction with the geographic information system (GIS) technology. ,in, , For the i The root tower's ontological characteristic data, It represents the characteristic dimension of the tower body characteristic data.

[0147] As an example, the second tower body feature data It may include: tower height, maximum operating life of the tower, type of tower structure material, etc.

[0148] As an example, tower operation and maintenance information can be obtained through the power company's operation and maintenance management system. ,in, , For the i The operation and maintenance status data of the tower, Represents the characteristic dimension of tower operation and maintenance status data.

[0149] As an example, tower operation and maintenance information It may include: the corrosion level of the tower, the time since the last maintenance, the years of use of the tower, etc.

[0150] As an example, in step S42, the tower local climate data is obtained by the tower local climate data prediction model. ,in, express t Moment i Local climate data of the tower, Represents the characteristic dimension of the local climate data of the tower.

[0151] As an example, tower local climate data It may include: local wind speed of the tower, local wind direction of the tower, local rainfall of the tower, local temperature of the tower, local humidity of the tower, etc.

[0152] Furthermore, the local climate data of the tower , Second tower body characteristic data , tower operation and maintenance information Combine them to get the input characteristics of each tower, which can be expressed as:

[0153]

[0154] in, for t Moment i Input characteristics of the root tower; The feature dimension representing the input features of each tower is d 1 、 d 2 、 d 3 The union of .

[0155] Furthermore, the input features of all towers are combined to obtain the second input data set , the expression is:

[0156]

[0157] in, express tThe input characteristics of the first tower at time t, express t The input characteristics of the second tower at time t, express t The input characteristics of the Nth tower at time, N Indicates the number of towers in a certain area. is the feature dimension of the input features of each tower.

[0158] As an example, in step S43, the second input data set Normalization is performed to eliminate the impact of different feature scale differences on model training, so that all features are processed at the same scale, improving the efficiency and stability of model training.

[0159] As an example, you can use min-max normalization to scale the data to a fixed interval. You can also set it to compress the data to the [0, 1] interval.

[0160] As an example, for the second input dataset Perform the minimum-maximum normalization operation, the expression is as follows:

[0161]

[0162] in, After normalization, i Input data, is the first i data points, For the second input dataset The minimum value in For the second input dataset The maximum value in .

[0163] As an example, the data of each feature dimension in the second input data set is normalized by minimum-maximum normalization to obtain normalized input data, and the normalized input data is combined into the second input matrix ,in, N Indicates the number of towers in a certain area. is the feature dimension of the input features of each tower.

[0164] As an example, in step S44, the specific method of converting the preprocessed second input data set into a graph structure is consistent with the specific method of converting the preprocessed first input data into a graph structure in step S23, and will not be repeated here.

[0165] As an example, in step S45, a second graph neural network model is constructed, and graph convolution is used to aggregate feature information of each node and its neighboring nodes. The second graph neural network may include an input layer, a graph convolution layer, and an output layer.

[0166] As an example, the second graph neural network model can be set to have a total of L2 layers, and the input layer can be expressed as ,in, Represents the node feature matrix of the input layer of the second graph neural network model, is the second input matrix.

[0167] As an example, in the second graph neural network model l The node feature matrix of the layer is , then in the second graph neural network model l +1 layer node feature matrix It can be obtained by the following formula:

[0168]

[0169] in, is the first l The node feature matrix of the layer; is the first l The weight matrix of the layer represents the parameters of the linear transformation and needs to be learned through training; is the normalized adjacency matrix, which is used to aggregate the information of neighboring nodes to the current node; σ is a non-linear activation function.

[0170] As an example, the nonlinear activation function σ You can use ReLU Activation function to enhance nonlinear expression capabilities.

[0171] For example, stacking multiple graph convolutional layers forms the backbone of the second graph neural network model. Each layer expands the receptive field of information, meaning that the final representation of each node can incorporate information from the adjacency matrix over a wider range. For example, the first layer can integrate information from direct neighbors, while the second layer can integrate information from second-order neighbors. By stacking multiple layers, the second graph neural network model can capture a wider range of node contexts.

[0172] As an example, the output layer of the second graph neural network model can be set to fully connected, expressed as:

[0173]

[0174] in, express tThe predicted value of the tower collapse probability at the moment is the first L 2 The node feature matrix of the layer, that is, the node feature matrix of the last layer in the second graph neural network model; is the weight matrix of the output layer in the second graph neural network model.

[0175] As an example, in step S46, the tower actual collapse label is set according to the actual collapse of the tower, and the tower actual collapse label is output as the second target. ,in, for t The actual collapsed tower label of the first tower at the moment, for t The actual tower collapse label of the second tower at the moment, for t The actual collapsed tower label of the Nth tower at the moment.

[0176] As an example, for i The actual tower collapse tag of the tower The value of can be 0 or 1, that is, .

[0177] Furthermore, the second graph neural network model is trained using the second input data set of the graph structure, and the weight matrix of each layer is continuously updated to make the predicted value of the tower collapse label as close as possible to the true value in the second target output, and determine the optimal weight matrix.

[0178] Specifically, the second input data set and the second target output of the graph structure are constructed into a training sample, wherein the second input data set of the graph structure is the input data in the training sample, and the second target output is the output data in the training sample. The input data in the training sample includes: the normalized adjacency matrix of the graph structure , the second input matrix The second target output may include: the actual tower tower label matrix .

[0179] Furthermore, the second graph neural network model is trained using the training samples. First, the parameters of the second graph neural network model are initialized, and the second input data set of the graph structure is fed into the constructed second graph neural network model. The second input data set of the graph structure is passed layer by layer according to the structure of the second graph neural network model, and relevant features are extracted to obtain the predicted value of the tower collapse label. Then, the predicted value of the tower collapse label is compared with the second target output label, and the loss value is calculated using the second loss function to evaluate the degree of deviation between the predicted value of the tower collapse label obtained by the current second graph neural network model and the actual tower collapse label in the second target output. Then, check whether the current loss value reaches the training target. If so, it means that the prediction accuracy of the second graph neural network model has reached the optimal value, and the training is terminated. If not, continue to determine whether the preset maximum number of training times has been reached. If so, terminate the training. If not, perform back propagation. Based on the obtained loss value, use the back propagation algorithm to reversely calculate the gradient from the output layer to the input layer to determine the parameters that need to be adjusted in the second graph neural network model. Based on the gradient calculated by back propagation, use a suitable optimization algorithm to update the parameters of the second graph neural network model, and then continue forward propagation. Repeat the above training process to optimize the parameters of the second graph neural network model in the direction of reducing the loss value until the training target is reached or the preset maximum number of training times is reached. End the training and obtain the final tower collapse probability prediction model.

[0180] As an example, the second loss function can adopt a binary cross entropy loss function, which is expressed as:

[0181]

[0182] in, N Indicates the number of towers in a certain area. For the i The actual tower tag of the tower, For the i The predicted value of the tower collapse label of the tower.

[0183] As an example, the training goal is to continuously reduce the loss value of the second loss function until it approaches zero or becomes stable.

[0184] As an example, the update rule for gradient descent is:

[0185]

[0186] in, is the learning rate of the second graph neural network model, is the updated first l The weight matrix of the layer, is the firstl The weight matrix of the layer, is the loss value of the second loss function.

[0187] As an example, the optimization algorithm may adopt a gradient descent method, or adaptive moment estimation (Adam), stochastic gradient descent (SGD), etc.

[0188] As an example, the maximum number of training times may be set based on the performance of the hardware device.

[0189] Furthermore, the tower collapse probability prediction model is used to predict the probability of tower collapse of all towers in the area, and the probability of tower collapse under extreme climate scenarios is obtained for each tower. The local key meteorological indicators of each tower are predicted, providing data support for the subsequent calculation of line disconnection probability.

[0190] As an example, the tower collapse probability prediction model outputs the tower collapse probability prediction value at time t It can be expressed as:

[0191]

[0192] in, is the adjacency matrix, is the second input matrix, Represents the tower collapse probability prediction model.

[0193] In step S5, see Picture 1 In step S5, based on the probability of each tower collapsing, the probability of disconnection of the entire transmission line is determined by a probability combination model.

[0194] As an example, each tower collapse event can be regarded as an independent event. Then the probability that the entire transmission line does not break is the joint probability that each tower in the entire transmission line does not collapse.

[0195] As an example, a probability combination model can be constructed using a probability modeling method. Based on the probability combination model, the overall disconnection probability of the entire transmission line is determined, and the local node failure risks are integrated into a system-level transmission line disconnection risk indicator, ultimately obtaining the disconnection probability of each transmission line in the power grid.

[0196] As an example, a transmission line line Total disconnection probability It can be expressed as:

[0197]

[0198] in, express t Moment i The predicted value of the tower collapse probability of a tower.

[0199] As an example, a transmission line line The tower shall include at least one inverted tower.

[0200] In step S6, refer to Picture 1 In step S6, a digital effect analysis model of future-state scheduling of the power grid under extreme climate conditions is constructed, and based on the digital effect analysis model of future-state scheduling of the power grid under extreme climate conditions, the losses under future-state scheduling are analyzed to realize the analysis of the future-state scheduling effect of the power grid under extreme weather conditions.

[0201] As an example, see Picture 4 , step S6 may include the following steps: step S61 to step S67.

[0202] Step S61: constructing a digital effect analysis model for future state scheduling of power grids under extreme climate conditions, wherein the digital effect analysis model for future state scheduling of power grids under extreme climate conditions comprises: a line disconnection scenario generation module, an input module, a scheduling analysis module, and a scheduling effect analysis module.

[0203] Step S62: inputting the disconnection probability of each transmission line in the power grid into a disconnection scenario generation module to obtain a disconnection scenario corresponding to each transmission line in the power grid.

[0204] Step S63: Send the disconnection scenarios corresponding to all transmission lines in the power grid and the data in the input module to the scheduling analysis module, which includes: a generator unit combination model before extreme weather occurs, a traditional scheduling model under extreme weather, and a future state scheduling model under extreme weather.

[0205] Step S64: obtaining a unit scheduling plan before the extreme weather occurs through the generator unit combination model before the extreme weather occurs in the scheduling analysis module.

[0206] Step S65: Obtain a traditional scheduling solution under extreme climate conditions through the traditional scheduling model under extreme climate conditions in the scheduling analysis module.

[0207] Step S66: Obtain a future-state scheduling plan under extreme climate conditions through the future-state scheduling model under extreme climate conditions in the scheduling analysis module.

[0208] Step S67: Input the unit dispatching plan before the extreme climate occurs, the traditional dispatching plan under extreme climate, and the future dispatching plan under extreme climate into the dispatching effect analysis module for effect analysis. Based on the probability of transmission line disconnection and the number of disasters, the power grid loss and social loss under the future dispatching are determined to obtain the effect of the future dispatching of the power grid under extreme climate.

[0209] As an example, in step S61, a digital effect analysis model for future grid dispatch under extreme climate conditions is constructed. Picture 5 The digital effect analysis model for future grid scheduling under extreme climate conditions includes: a line disconnection scenario generation module, an input module, a scheduling analysis module, and a scheduling effect analysis module. The output end of the input module is connected to the first input end of the scheduling analysis module, the output end of the line disconnection scenario generation module is connected to the second input end of the scheduling analysis module, and the output end of the scheduling analysis module is connected to the input end of the scheduling effect analysis module.

[0210] As an example, the input module is used to obtain and organize input data, which may include power grid structure, power grid line parameters, power grid node load power, generator set parameters, unit spare capacity purchase cost, etc.

[0211] As an example, the line disconnection scenario generation module is used to judge the line disconnection probability of each transmission line in the power grid based on the tower collapse probability prediction model, and obtain the line disconnection scenario corresponding to each transmission line in the power grid.

[0212] As an example, the scheduling analysis module includes: a generator set combination model before extreme climate occurs, a traditional scheduling model under extreme climate, and a future-state scheduling model under extreme climate. It is used to input the line disconnection scenarios and input data corresponding to each transmission line in the power grid, and uses the generator set combination model before extreme climate occurs, the traditional scheduling model under extreme climate, and the future-state scheduling model under extreme climate for analysis respectively to obtain the unit scheduling plan before extreme climate occurs, the traditional scheduling plan under extreme climate, and the future-state scheduling plan under extreme climate.

[0213] As an example, the scheduling effect analysis module is used to analyze the decision-making differences between the unit scheduling plan before extreme climate occurs, the traditional scheduling plan under extreme climate, and the future scheduling plan under extreme climate, determine the power grid losses and social losses under future scheduling, and obtain the effect of future scheduling of the power grid under extreme climate.

[0214] As an example, in step S62, based on the probability of disconnection of each transmission line in the power grid obtained by the tower collapse probability prediction model, the disconnection probability of each transmission line is compared with the set value through the scenario generation module. θ Compare and find that if the probability of disconnection of the transmission line is greater than the set value θ , then the disconnection scenario corresponding to the transmission line is generated, indicating that the transmission line will be scheduled in the future state; if the disconnection probability of the transmission line is not greater than the set value θ , no disconnection scenario is generated, indicating that the transmission line will not be scheduled in the future. All transmission lines in the power grid are judged and the disconnection scenarios corresponding to all transmission lines are output.

[0215] As an example, in step S63, the disconnection scenarios corresponding to all transmission lines in the power grid and the data in the input module are sent to the scheduling analysis module, and the scheduling analysis module includes: a generator set combination model before extreme climate occurs, a traditional scheduling model under extreme climate, and a future state scheduling model under extreme climate.

[0216] For example, in step S64, before extreme weather occurs, the power grid lacks future state warnings and cannot predict impending transmission line failures or unit output limitations. Consequently, the grid schedules according to the existing grid structure. This pre-extreme weather unit combination model in the scheduling analysis module is analyzed and optimized using the first objective function to determine a pre-extreme weather unit scheduling plan.

[0217] As an example, the first objective function can be based on the power generation cost, startup cost, no-load operation cost and power generation cost of conventional units, aiming to minimize the power generation cost, startup cost, no-load operation cost and power generation cost of new energy units. Specifically, a first constraint con1.1 can be added based on the upper and lower limits of conventional unit output; a second constraint con1.2 can be added based on the output upper limit of new energy units such as wind power generation and photovoltaic power generation; a third constraint con1.3 and a fourth constraint con1.4 can be added based on the output variation range limit of conventional units in adjacent time periods, and the third and fourth constraints can be climbing constraints; a fifth constraint con1.5 can be added based on the unit start-up and shutdown and state transition logic; a sixth constraint con1.6 and a seventh constraint con1.7 can be added based on the minimum continuous operating time or downtime requirement for conventional units; an eighth constraint con1.8 can be added to ensure that the start-up and shutdown and operating states are binary integer variables; a ninth constraint con1.9 can be added based on the supply and demand balance of the power grid; and a tenth constraint con1.10 and an eleventh constraint con1.11 can be added based on the capacity limit for safe operation of transmission lines. Since there are almost no independent power grid areas in actual power systems, and power grid areas are connected with other external power grids based on sections, the section connection power can be added to the ninth constraint con1.9 and the tenth constraint con1.10. Assume that a natural disaster occurs during period t, the first objective function The expression is as follows:

[0218]

[0219] in, For the i Output quotation function of each unit; For the i Start-up cost of each unit; For the i No-load operating cost of each unit; For the neThe power generation cost of each new energy unit; For the i units t Whether to start during the period, if t 1 if the period is started, 0 otherwise; For the i units t Whether to shut down during the period, if t If the period is down, it is 1, otherwise it is 0; For the i Units in t Whether the time period is put into operation, if it is put into operation, it is 1, if not put into operation, it is 0; For the i Units in t Output during the time period; For new energy units t Actual dispatch output during the time period; for t Time period d load value; For the s The cross section t The amount of electricity required to be sent or received during the time period; The first and last nodes of the line are nm The maximum transmission capacity of the line; For the i The vector form of the minimum output of each unit; For the i Units in t The vector form of the output during the time period; For the i Units in t-1 The vector form of the output during the time period; For the i The vector form of the maximum output of each unit; For the ne New energy units in t The vector form of the output during the time period; For the ne New energy units in t The vector form of the maximum output value predicted for the time period; For conventional generator sets; For the i The downward ramp rate limit of each unit; For the i The upward climbing rate limit of each unit; For the n The transmission line where the node is located; is the set of all transmission lines in the power grid; For then Nodes in t The node voltage phase angle of the time period; For the m Nodes in t The node voltage phase angle of the time period; for t The beginning and end nodes of the time period line are nm lines; T The total time period of the unit dispatch plan before the extreme weather occurs; For the i The start-up period of each unit; For the i downtime of each unit; NE Assemble for new energy generator sets; For the n Nodes of new energy generators; D is the node load collection in the power grid.

[0220] As an example, in step S65, after an extreme weather disaster occurs, transmission lines affected by the extreme weather disaster may fail, resulting in power flow congestion on certain transmission lines or limited output of wind and solar generators. In this case, the power grid can be re-dispatched based on the post-fault grid structure or generator information. The traditional extreme weather dispatch model in the dispatch analysis module is analyzed and optimized using the second objective function to obtain a traditional extreme weather dispatch solution.

[0221] As an example, since the current capacity after a fault may not be sufficient to support the grid's power supply capacity and safe operation requirements, the grid needs to urgently activate spare capacity with fast ramping capabilities. Since the grid is dispatched urgently, the power generation resources that the grid can purchase and mobilize in a short period of time only include spare capacity with a short call time. The spare capacity with a short call time can include: rotating spare capacity (call time within 10 minutes) and fast spare capacity (call time within 30 minutes). Therefore, when a transmission line fails, the objective function of the traditional dispatching model adds the cost of rotating spare capacity and the cost of fast spare capacity compared to when there is no fault. The output constraint and climbing constraint of the standby unit can be added to the constraint conditions of the first objective function to obtain the second objective function. , the expression is as follows:

[0222]

[0223] in, For the l The second constraint condition when a line fails; For the i Output quotation function of each unit; For the jA unit that can quickly supply capacity in a short time t Output quotation function for a certain period of time; For the ne The power generation cost of each new energy unit; For the j A unit that can quickly supply capacity in a short time t Output during the time period; For the i Units in t Output during the time period; For new energy units t Actual dispatch output during the time period; for t Time period d load value; For the s The cross section t The amount of electricity required to be sent or received during the time period; For the i The vector form of the minimum output of each unit; For the i Units in t The vector form of the output during the time period; For the i Units in t-1 The vector form of the output during the time period; For the i The vector form of the maximum output of each unit; For the ne New energy units in t The vector form of the output during the time period; For the ne New energy units in t The vector form of the maximum output value predicted for the time period; The first and last nodes of the line are nm The maximum transmission capacity of the line; For conventional generator sets; A collection of spare capacity units with short call times; For the n A collection of conventional generator sets of nodes; For the i The downward ramp rate limit of each unit; For the i The upward climbing rate limit of each unit; For the n Nodes in t The node voltage phase angle of the time period; For the m Nodes in t The node voltage phase angle of the time period; fort The beginning and end nodes of the time period line are nm lines; T is the total time period of the traditional scheduling scheme under extreme climate conditions; For the l After the first line fails, remove the l The collection of all other lines.

[0224] For example, when analyzing a typhoon scenario, transmission line failures and typhoon-induced fluctuations in wind turbine and photovoltaic output are possible. Grid dispatch needs to be based on the post-fault grid structure, while also imposing output limits on wind and photovoltaic units. When analyzing a wildfire scenario, transmission line failures caused by wildfires can be considered, and dispatch can be based on the post-fault grid structure.

[0225] As an example, in step S66, the grid will develop a dispatch plan for typical extreme weather scenarios in the dispatch analysis module, reserving sufficient backup capacity. This plan is analyzed using the dispatch analysis module's future-state dispatch model under extreme weather conditions and optimized using the third objective function to obtain a future-state dispatch plan under extreme weather conditions.

[0226] For example, in extreme climate scenarios, the future-state scheduling model requires more preparation time than the traditional scheduling model, resulting in a wider range of resources available for purchase. The variables of the third objective function can be expanded to include backup capacity resources with longer call times, based on the traditional real-time scheduling model. These backup capacity resources with longer call times can include slow standby units (call times of more than 30 minutes), providing more options.

[0227] As an example, when l The third objective function of the future state scheduling model under extreme climate conditions adds the cost of slow standby capacity when a transmission line fails, and the constraints related to slow standby units are added. Taking into account the output prediction error of wind and solar units, in order to ensure the reliability of the scheduling plan under fault conditions, the output upper limit of the new energy unit can be set to the predicted value multiplied by the historical minimum actual prediction ratio in this embodiment. Since the future state scheduling model under extreme climate conditions adds analysis of the probability of line failures and the output limitations of wind and solar generators under extreme climate conditions, the power grid can consider the impact of extreme climate on the power system and make advance plans, thereby adding more alternative units for scheduling, which reduces the economic cost of scheduling. l The third objective function when a line fails The expression is as follows:

[0228]

[0229] in, For thei Output quotation function of each unit; For the j A unit that can quickly supply capacity in a short time t Output quotation function for a certain period of time; For the ne The power generation cost of each new energy unit; For the k A slow standby unit is t Output quotation function for a certain period of time; For the i Units in t Output during the time period; For the j A unit that can quickly supply capacity in a short time t Output during the time period; For the k A slow standby unit is t Output during the time period; For new energy units t Actual dispatch output during the time period; for t Time period d load value; For the s The cross section t The amount of electricity required to be sent or received during the time period; For the i The vector form of the minimum output of each unit; For the i Units in t The vector form of the output during the time period; For the i Units in t-1 The vector form of the output during the time period; For the i The vector form of the maximum output of each unit; For the ne New energy units in t The vector form of the output during the time period; For the ne New energy units in t The vector form of the maximum output value predicted for the time period; For the ne The minimum value between the actual historical output value and the predicted value of each new energy unit; For the ne The actual historical output value of each new energy unit; For the ne Historical output forecast value of each new energy generating unit; The first and last nodes of the line are nmThe maximum transmission capacity of the line; For the i The downward ramp rate limit of each unit; For the i The upward climbing rate limit of each unit; For the n Nodes in t The node voltage phase angle of the time period; For the m Nodes in t The node voltage phase angle of the time period; for t The beginning and end nodes of the time period line are nm lines; T is the total time period of the future scheduling plan under extreme climate conditions; NE Assemble for new energy generator sets; For the n Nodes of new energy generators; For the l After the first line fails, remove the l The collection of all other lines of the line; For conventional generator sets; A collection of spare capacity units with short call times; A collection of slow standby capacity units with a longer call time; For the n A collection of conventional generator sets of nodes; G for G 1 、 G 2 、 G 3 The intersection of three sets; D is the node load set in the power grid; D n For the n The load of a node.

[0230] As an example, in step S67, the dispatch analysis module can determine the traditional and future-state dispatch results for the power grid under different extreme climate conditions. Based on the dispatch results under different extreme climate conditions, the increased temporary power demand of the power grid is determined, and the effects of the future-state dispatch measures under each extreme climate condition are determined.

[0231] As an example, by inputting the unit scheduling plan before the extreme climate occurs, the traditional scheduling plan under extreme climate, and the future scheduling plan under extreme climate into the scheduling effect analysis module for analysis, it can be found that the traditional scheduling plan under extreme climate can only start emergency standby capacity such as rotating reserve after a fault occurs. After the fault, it can be rescheduled according to information such as the grid after the fault, and can call on spare capacity resources with a longer time; the future scheduling plan under extreme climate will advance the rescheduling process, and quickly adjust based on the predicted results when a fault occurs, thereby improving the efficiency of the initial emergency response.

[0232] As an example, the results for each period can be averaged and the grid and social losses can be estimated based on the probability of transmission line outages and the number of disasters.

[0233] For example, when an extreme weather emergency occurs, the power grid urgently activates its backup capacity. If the backup capacity is sufficient to meet the system power supply shortage, the grid company's loss is only the purchase cost of the backup capacity; if the backup capacity is insufficient to meet the power supply shortage, some users will face the risk of power outages. At this time, the grid loss can be the purchase cost of the backup capacity and the power outage compensation cost given to users whose power outages are caused by the fault. The expression is as follows:

[0234]

[0235] in, F is the annual occurrence of extreme climate; N is the total number of failure types under a certain extreme climate; T is the total time period of the traditional scheduling scheme under extreme climate conditions; a The type of failure under certain extreme climate conditions; For the a Failure probability of each fault type; is the cost function of the standby unit; For the a Under the fault type t The required standby unit output for a period of time is given by the scheduling analysis module; For the a Under the fault type t The power supply shortage of the power system during the period; The reserve capacity of the power grid; It is the power outage compensation coefficient per unit electricity.

[0236] For example, when a power system fails, if the power system has a power shortage and causes a power outage, it will lead to industrial shutdown, commercial interruption and instability of residents' lives, causing serious social losses and reduced social efficiency. The power shortage can be selected as an indicator to measure social losses based on the amount of power outages caused by users. The social losses caused by emergencies in extreme climates are The expression is as follows:

[0237]

[0238] in, F is the annual occurrence of extreme climate; N is the total number of failure types under a certain extreme climate; T is the total time period of the traditional scheduling scheme under extreme climate conditions; a The type of failure under certain extreme climate conditions; For the a Failure probability of each fault type; For the a Under the fault type t The required standby unit output for a period of time is given by the scheduling analysis module; For the a Under the fault type t The power supply shortage of the power system during the period; The reserve capacity of the power grid.

[0239] As an example, by analyzing the power grid losses and social losses of future-state scheduling plans under extreme climate conditions, we can quantitatively analyze the impact of future-state scheduling on scheduling effects before and after its implementation, and realize the effectiveness analysis of future-state scheduling measures under various extreme climate conditions.

[0240] In the graph neural network-based line disconnection probability prediction and future scheduling effect analysis method of the present application, by constructing a graph neural network model, the mapping from regional-scale climate information to node-scale microclimate characteristics is realized, and the local climate data near each tower can be predicted with high precision in the absence of real-time observation; the accuracy and precision of the line disconnection probability prediction are improved through two-stage modeling. In the first stage, the graph neural network model is trained using regional-level meteorological data to predict the local meteorological characteristics of the location of each tower. In the second stage, the tower collapse probability model is trained based on the predicted local meteorological data and tower structural properties of the tower, and further through By calculating the disconnection probability of the entire transmission line through combinational logic, it is possible to output the disconnection probability of the entire line with high precision by inputting only regional meteorological data, taking into account both practicality and prediction accuracy, and significantly improving the fault prediction capability of the transmission system under extreme climate conditions; through the digital effect analysis model of future-state scheduling of power grids under extreme climate conditions, the lack of analysis of future-state scheduling effects of power grids under extreme climate conditions in existing technologies is filled. By conducting a detailed analysis of power grid scheduling under typical extreme climate conditions, the impact of future-state scheduling plans on power grid losses and social losses is quantified, providing a scientific decision-making basis for whether the power grid should invest in future-state scheduling.

[0241] It should be understood that although the steps in the flowcharts of the accompanying drawings are shown in sequence as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least some of the steps in the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The order of execution of these sub-steps or stages is not necessarily sequential, but can be executed in rotation or alternation with other steps or at least a portion of the sub-steps or stages of other steps.

[0242] In another embodiment, the present application also provides a line disconnection probability prediction and future state scheduling effect analysis system based on graph neural network, and the line disconnection probability prediction and future state scheduling effect analysis system based on graph neural network includes: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement any of the above-mentioned line disconnection probability prediction and future state scheduling effect analysis methods based on graph neural network.

[0243] In another embodiment, the present application also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the various steps of the line disconnection probability prediction and future scheduling effect analysis method based on graph neural network provided in the above embodiment.

[0244] The computer-executable instructions for implementing the methods of the present application may be written in any combination of one or more programming languages. These computer-executable instructions may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the computer-executable instructions are executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer-executable instructions may be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or electronic device.

[0245] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a cathode ray tube (CRT) or a liquid crystal display (LCD) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; feedback provided to the user can be any form of sensory feedback (e.g., visual feedback or tactile feedback); and input from the user can be received in any form, including acoustic input, voice input, or tactile input.

[0246] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as data electronics), or a computing system that includes middleware components (e.g., application electronics), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0247] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0248] Although the present application has been disclosed above with reference to the embodiments, they are not intended to limit the present application. Anyone with ordinary knowledge in the technical field may make slight changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be determined by the scope of the appended patent application.

Claims

1. A method for predicting line disconnection probability and analyzing future scheduling effects based on graph neural networks, characterized in that: The following steps are involved: Obtain historical regional climate data and the first tower's characteristic data; Constructing a tower local climate data prediction model based on the historical regional climate data and the first tower body characteristic data, including: obtaining actual tower local climate data, setting the actual tower local climate data as first target output data, combining the historical regional climate data and the first tower body characteristic data to obtain first input data; preprocessing the first input data and the first target output data; and converting the preprocessed first input data into a graph structure; Build the first graph neural network; The first graph neural network is trained using the first input data of the graph structure and the first target output data after preprocessing, and the parameters are optimized and updated using the first loss function to obtain a tower local climate data prediction model; Using the tower local climate data prediction model to predict key meteorological indicators of the tower to obtain tower local climate data; Acquiring characteristic data of a second tower and tower operation and maintenance information, constructing a tower collapse probability prediction model based on the characteristic data of the second tower, the tower operation and maintenance information, and local climate data of the tower, and obtaining a probability of tower collapse for each tower based on the tower collapse probability prediction model; Based on the probability of each tower collapsing, the probability of disconnection of the entire transmission line is determined by a probability combination model; Construct a digital effect analysis model for the future state scheduling of the power grid under extreme climate, analyze the losses under future state scheduling based on the digital effect analysis model for the future state scheduling of the power grid under extreme climate, and realize the analysis of the future state scheduling effect of the power grid under extreme weather, including: constructing a digital effect analysis model for the future state scheduling of the power grid under extreme climate, the digital effect analysis model for the future state scheduling of the power grid under extreme climate includes: a line break scenario generation module, an input module, a scheduling analysis module, and a scheduling effect analysis module; input the line break probability of each transmission line in the power grid into the line break scenario generation module to obtain the line break scenario corresponding to each transmission line in the power grid; send the line break scenarios corresponding to all transmission lines in the power grid and the data in the input module to the scheduling analysis module, the scheduling analysis module includes: the generator group before the extreme climate occurs A combined model, a traditional scheduling model under extreme climate, and a future-state scheduling model under extreme climate; the unit scheduling plan before the occurrence of extreme climate is obtained through the generator unit combination model before the occurrence of extreme climate in the scheduling analysis module; the traditional scheduling plan under extreme climate is obtained through the traditional scheduling model under extreme climate in the scheduling analysis module; the future-state scheduling plan under extreme climate is obtained through the future-state scheduling model under extreme climate in the scheduling analysis module; the unit scheduling plan before the occurrence of extreme climate, the traditional scheduling plan under extreme climate, and the future-state scheduling plan under extreme climate are input into the scheduling effect analysis module for effect analysis, and the power grid loss and social loss under the future-state scheduling are determined based on the probability of transmission line disconnection and the number of disasters, so as to obtain the effect of the future-state scheduling of the power grid under extreme climate.

2. The method for predicting line disconnection probability and analyzing future scheduling effects based on graph neural network according to claim 1 is characterized in that: The tower local climate data prediction model is used to predict the key meteorological indicators of the tower. The predicted value of the tower local climate data at time t is The expression is: in, is the adjacency matrix, is the first input feature matrix, Represents the tower local climate data prediction model.

3. The method for predicting line disconnection probability and analyzing future scheduling effects based on graph neural network according to claim 1 is characterized in that: Acquiring characteristic data of a second tower and tower operation and maintenance information, constructing a tower collapse probability prediction model based on the characteristic data of the second tower, the tower operation and maintenance information, and the tower local climate data, and obtaining a probability of tower collapse for each tower based on the tower collapse probability prediction model, including: Obtaining the second tower body characteristic data and tower operation and maintenance information; Combining the second tower body characteristic data, the tower operation and maintenance information, and the tower local climate data to obtain a second input data set; Preprocessing the second input data set; Converting the preprocessed second input data set into a graph structure; Build the second graph neural network; The second graph neural network is trained using a second input data set with a graph structure, and the parameters are optimized and updated using a second loss function to obtain a tower collapse probability prediction model.

4. The method for predicting line disconnection probability and analyzing future scheduling effects based on graph neural network according to claim 1 is characterized in that: Based on the probability of each tower falling, the probability of the entire transmission line being disconnected is determined by the probability combination model, and the total probability of the transmission line being disconnected is Expressed as: in, express t Moment i The predicted value of the tower collapse probability of a tower.

5. The method for predicting line disconnection probability and analyzing future scheduling effects based on graph neural network according to claim 1 is characterized in that: Obtaining a unit scheduling plan before extreme weather occurs by using the generator unit combination model before extreme weather occurs in the scheduling analysis module, including: analyzing the generator unit combination model before extreme weather occurs in the scheduling analysis module, and optimizing using a first objective function to obtain a unit scheduling plan before extreme weather occurs, wherein the first objective function The expression is as follows: in, For the i Output quotation function of each unit; For the i Start-up cost of each unit; For the i No-load operating cost of each unit; For the ne The power generation cost of each new energy unit; For the i units t Whether to start during the period, if t 1 if the period is started, 0 otherwise; For the i units t Whether to shut down during the period, if t If the period is down, it is 1, otherwise it is 0; For the i Units in t Whether the time period is put into operation, if it is put into operation, it is 1, if not put into operation, it is 0; For the i Units in t Output during the time period; For new energy units t Actual dispatch output during the time period; for t Time period d load value; For the s The cross section t The amount of electricity required to be sent or received during the time period; The first and last nodes of the line are nm The maximum transmission capacity of the line; For the i The vector form of the minimum output of each unit; For the i Units in t The vector form of the output during the time period; For the i Units in t-1 The vector form of the output during the time period; For the i The vector form of the maximum output of each unit; For the ne New energy units in t The vector form of the output during the time period; For the ne New energy units in t The vector form of the maximum output value predicted for the time period; For conventional generator sets; For the i The downward ramp rate limit of each unit; For the i The upward climbing rate limit of each unit; For the n The transmission line where the node is located; is the set of all transmission lines in the power grid; For the n Nodes in t The node voltage phase angle of the time period; For the m Nodes in t The node voltage phase angle of the time period; for t The beginning and end nodes of the time period line are nm lines; T The total time period of the unit dispatch plan before the extreme weather occurs; For the i The start-up period of each unit; For the i downtime of each unit; NE Assemble for new energy generator sets; For the n Nodes of new energy generators; D is the node load collection in the power grid.

6. The method for predicting line disconnection probability and analyzing future scheduling effects based on graph neural network according to claim 1 is characterized in that: Obtaining a traditional scheduling scheme under extreme climate by using the traditional scheduling model under extreme climate in the scheduling analysis module includes: analyzing the traditional scheduling model under extreme climate in the scheduling analysis module and optimizing using a second objective function to obtain a traditional scheduling scheme under extreme climate, wherein the second objective function , the expression is as follows: in, For the l The second constraint condition when a line fails; For the i Output quotation function of each unit; For the j A unit that can quickly supply capacity in a short time t Output quotation function for a certain period of time; For the ne The power generation cost of each new energy unit; For the j A unit that can quickly supply capacity in a short time t Output during the time period; For the i Units in t Output during the time period; For new energy units t Actual dispatch output during the time period; for t Time period d load value; For the s The cross section t The amount of electricity required to be sent or received during the time period; For the i The vector form of the minimum output of each unit; For the i Units in t The vector form of the output during the time period; For the i Units in t-1 The vector form of the output during the time period; For the i The vector form of the maximum output of each unit; For the ne New energy units in t The vector form of the output during the time period; For the ne New energy units in t The vector form of the maximum output value predicted for the time period; The first and last nodes of the line are nm The maximum transmission capacity of the line; For conventional generator sets; A collection of spare capacity units with short call times; For the n A collection of conventional generator sets of nodes; For the i The downward ramp rate limit of each unit; For the i The upward climbing rate limit of each unit; For the n Nodes in t The node voltage phase angle of the time period; For the m Nodes in t The node voltage phase angle of the time period; for t The beginning and end nodes of the time period line are nm lines; T is the total time period of the traditional scheduling scheme under extreme climate conditions; For the l After the first line fails, remove the l The collection of all other lines.

7. The method for predicting line disconnection probability and analyzing future scheduling effects based on graph neural network according to claim 1 is characterized in that: Obtaining a future state scheduling scheme under extreme climate by using the future state scheduling model under extreme climate in the scheduling analysis module, including: analyzing the future state scheduling model under extreme climate in the scheduling analysis module and optimizing using a third objective function to obtain a future state scheduling scheme under extreme climate, wherein the third objective function The expression is as follows: in, For the i Output quotation function of each unit; For the j A unit that can quickly supply capacity in a short time t Output quotation function for a certain period of time; For the ne The power generation cost of each new energy unit; For the k A slow standby unit is t Output quotation function for a certain period of time; For the i Units in t Output during the time period; For the j A unit that can quickly supply capacity in a short time t Output during the time period; For the k A slow standby unit is t Output during the time period; For new energy units t Actual dispatch output during the time period; for t Time period d load value; For the s The cross section t The amount of electricity required to be sent or received during the time period; For the i The vector form of the minimum output of each unit; For the i Units in t The vector form of the output during the time period; For the i Units in t-1 The vector form of the output during the time period; For the i The vector form of the maximum output of each unit; For the ne New energy units in t The vector form of the output during the time period; For the ne New energy units in t The vector form of the maximum output value predicted for the time period; For the ne The minimum value between the actual historical output value and the predicted value of each new energy unit; For the ne The actual historical output value of each new energy unit; For the ne Historical output forecast value of each new energy generating unit; The first and last nodes of the line are nm The maximum transmission capacity of the line; For the i The downward ramp rate limit of each unit; For the i The upward climbing rate limit of each unit; For the n Nodes in t The node voltage phase angle of the time period; For the m Nodes in t The node voltage phase angle of the time period; for t The beginning and end nodes of the time period line are nm lines; T is the total time period of the future scheduling plan under extreme climate conditions; NE Assemble for new energy generator sets; For the n Nodes of new energy generators; For the l After the first line fails, remove the l The collection of all other lines of the line; For conventional generator sets; A collection of spare capacity units with short call times; A collection of slow standby capacity units with a longer call time; For the n A collection of conventional generator sets of nodes; G for G 1 、 G 2 、 G 3 The intersection of three sets; D is the node load set in the power grid; D n For the n The load of a node.

8. The method for predicting line disconnection probability and analyzing future scheduling effects based on graph neural network according to claim 1 is characterized in that: The power grid loss under the future dispatch The expression is: in, F is the annual occurrence of extreme climate; N is the total number of failure types under a certain extreme climate; T is the total time period of the traditional scheduling scheme under extreme climate conditions; a The type of failure under certain extreme climate conditions; For the a Failure probability of each fault type; is the cost function of the standby unit; For the a Under the fault type t The required standby unit output for a period of time is given by the scheduling analysis module; For the a Under the fault type t The power supply shortage of the power system during the period; The reserve capacity of the power grid; is the power outage compensation coefficient per unit power; The social loss under the future scheduling The expression is: in, F is the annual occurrence of extreme climate; N is the total number of failure types under a certain extreme climate; T is the total time period of the traditional scheduling scheme under extreme climate conditions; a The type of failure under certain extreme climate conditions; For the a Failure probability of each fault type; For the a Under the fault type t The required standby unit output for a period of time is given by the scheduling analysis module; For the a Under the fault type t The power supply shortage of the power system during the period; The reserve capacity of the power grid.

9. A line disconnection probability prediction and future scheduling effect analysis system based on graph neural network, characterized by: The line disconnection probability prediction and future scheduling effect analysis system based on graph neural network includes: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the line disconnection probability prediction and future scheduling effect analysis method based on graph neural network as described in any one of claims 1 to 8.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method for predicting line disconnection probability and analyzing future scheduling effects based on graph neural networks as described in any one of claims 1 to 8 are implemented.

Images