Power grid dispatching training method and system based on panoramic digital twinborn simulation
By constructing a digital twin model of the grid and conducting panoramic simulation, the problem that the existing power grid scheduling training system cannot comprehensively simulate the operating status of the power grid, and achieves more real and efficient training effects.
Patent Information
- Application Number
- CN202411398875.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-09
- Publication Date
- 2025-01-28
Smart Images
Figure CN119380596A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power grid dispatching, and more specifically to a power grid dispatching training method and system based on panoramic digital twin simulation. Background Art
[0002] With the rapid development of the power system, the complexity of power grid dispatching is increasing, and the skill requirements for dispatchers are getting higher and higher. Traditional training methods are inefficient, costly, and difficult to simulate the complex conditions of the real power grid.
[0003] With the expansion of the scale of power grid construction and the promotion of the digital economy, the digitization and intelligence of power grids are becoming more and more urgent needs for the development of the power industry. As an emerging technical means, digital twin technology can create a virtual model of physical entities and realize real-time monitoring and simulation training of power grids. Based on information technology systems such as digital identification, automated perception, networked connection, inclusive computing, intelligent control and platform services, digital twin technology provides new ideas and ways to promote the safe and stable operation of power grids and build energy Internet enterprises.
[0004] The existing power grid dispatching training system cannot fully simulate the operation status of the power grid and lacks simulation of the overall power grid, which results in the dispatcher not being able to obtain sufficient practical operation experience during the training process. At the same time, the existing power grid dispatching training system often only focuses on the operation status of the power grid, but does not involve the associated simulation of the power grid hardware settings, making the power grid dispatching training results incomplete.
[0005] Therefore, how to propose a power grid dispatching training method and system based on panoramic digital twin simulation, perform panoramic simulation of the power grid, provide a more realistic training environment, and improve the training quality of dispatchers is an urgent problem that technical personnel in this field need to solve. Summary of the invention
[0006] In view of this, the present invention provides a power grid dispatching training method and system based on panoramic digital twin simulation, which performs panoramic simulation of the power grid, provides a more realistic training environment, and improves the training quality of dispatchers. In order to achieve the above purpose, the present invention adopts the following technical solutions:
[0007] A power grid dispatching training method based on panoramic digital twin simulation, comprising:
[0008] Obtain the operation data of the power grid and build a digital twin model of the power grid;
[0009] Conduct grid operation scenarios and fault simulations based on grid operation data;
[0010] Make grid operation and dispatch decisions based on the selected grid operation scenarios and fault types using the grid digital twin model;
[0011] Provide feedback and evaluation on grid operation and dispatch decisions.
[0012] Optionally, the obtaining of the operating data of the power grid includes: collecting operating data of power stations, substations and transmission lines.
[0013] Optionally, the building of a power grid digital twin model includes:
[0014] Determine the power grid absorption coefficient based on power load data;
[0015] Conduct power grid consumption evaluation on the target area according to the power grid consumption coefficient;
[0016] Data twinning is performed on the power absorption data and power transportation data according to the consumption assessment results, and multiple data twin layers are constructed according to the data twin space;
[0017] Couple multiple data twin layers to build a power grid digital twin model.
[0018] Optionally, the performing power grid operation scenario and fault simulation according to the power grid operation data includes:
[0019] Collect alarm information when an alarm signal appears in the power grid;
[0020] Through knowledge extraction, knowledge fusion and knowledge processing of alarm information, a knowledge graph of power grid operation scenarios is constructed;
[0021] A graph neural network with an attention mechanism is used to read the graph structure features of the knowledge graph of power grid operation scenarios, calculate the attention scores between nodes to obtain the node feature vector, integrate the entity features and relationship features of the nodes, obtain the bidirectional hidden state, and calculate the entity hidden feature vector;
[0022] Reasoning event fault types in knowledge graphs based on graph neural networks with attention mechanisms;
[0023] The grid operation scenarios and fault event data are mapped according to the event fault types.
[0024] Optionally, performing grid operation and dispatch decision-making on the grid digital twin model according to the selected grid operation scenario and fault type includes:
[0025] Set event fault types according to the power grid dispatch training plan;
[0026] Apply the grid operation scenarios and fault event data mapped to the event fault types to the grid digital twin model;
[0027] The digital twin model of the power grid is used to make grid operation and dispatch decisions to handle fault events.
[0028] Optionally, it also includes building a twin model of power grid operation and dispatch decision, and outputting decisions through the twin model of power grid operation and dispatch decision:
[0029] Select the distribution network area to be digitally twinned, extract the network topology model of the area to be operated, and build a digital twin model of the distribution automation operation area based on the mapping of the actual physical area of the distribution network based on real-time physical data and environmental data.
[0030] Optionally, the physical data is real-time physical data of the power distribution area mapped by the power consumption collection system and the SCADA system, including:
[0031] Collect system data deviations, voltage, current, power physical quantity deviations and switch status deviations, and feed them back to the power collection system and SCADA system;
[0032] Based on the environmental data obtained by the meteorological system, the correlation analysis algorithm is used to identify the correlation between abnormal operations, physical alarms and weather, temperature, humidity, and altitude. Based on the linear fitting or logistic regression algorithm, the nonlinear fitting relationship between physical data, switch state quantities and temperature, humidity, and altitude environment is constructed to correct data deviations in the simulation environment.
[0033] Optionally, it also includes optimization of the twin models of power grid operation and dispatch decision-making:
[0034] Acquiring position feature data and structural feature data;
[0035] A node regression matrix of a pre-trained twin model, wherein the node regression matrix is an interpolation regression matrix, which is used to convert the frame coordinates of the twin model into transformed reconstructed frame coordinates;
[0036] Generate a set of rotation matrices and translation vectors as global rigid motion perturbations, and randomly extract corresponding position features and structural features from the position feature data and structural feature data;
[0037] According to the position features, structural features, rotation matrix and translation vector, the twin model features and frame coordinates that are mapped to each other are obtained, and the node regression matrix is used to convert the paired matching model features and frame coordinates into the transformed output frame;
[0038] Training the feedback recurrent neural network based on the mutually mapped model features and frame coordinates;
[0039] The transformed framework is input into the trained feedback recurrent neural network, and the corresponding twin model features are extracted after iteration of the feedback recurrent neural network to complete the reconstruction of the twin model of power grid operation and dispatch decision.
[0040] Optionally, the feedback and evaluation of power grid operation and dispatching decisions include:
[0041] Perform data processing on the fault information mapped to the event fault type to obtain a training set;
[0042] Based on the K-fold cross-validation method, the training set is input into multiple classifiers for training to obtain different complementary subsets. The K classification accuracy rates obtained by performing K cross-validations on the complementary subsets are averaged to obtain the recall rates of different fault types of multiple classifiers.
[0043] According to the recall rate, multiple classifiers are weighted and fused to obtain the target fault classification model;
[0044] The target fault classification model is used to perform fault detection on the information data after the grid operation and dispatch decision, and the detection results are output.
[0045] Optionally, a power grid dispatching training system based on panoramic digital twin simulation includes:
[0046] Power grid digital twin model construction module: used to obtain power grid operation data and build a power grid digital twin model;
[0047] Power grid operation scenario and fault simulation module: used to simulate power grid operation scenarios and faults based on power grid operation data;
[0048] Grid operation and dispatch decision module: used to make grid operation and dispatch decisions for the grid digital twin model according to the selected grid operation scenario and fault type;
[0049] Feedback and evaluation module: used for feedback and evaluation of power grid operation and dispatch decisions.
[0050] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a power grid dispatching training method and system based on panoramic digital twin simulation, which has the following beneficial effects:
[0051] The present invention proposes a power grid dispatching training method based on panoramic digital twin simulation, including: obtaining the operation data of the power grid and constructing a power grid digital twin model; simulating the power grid operation scenario and fault according to the operation data of the power grid; making power grid operation and dispatching decisions on the power grid digital twin model according to the selected power grid operation scenario and fault type; and providing feedback and evaluation on the power grid operation and dispatching decisions. The present invention constructs a power grid digital twin model and a power grid operation and dispatching decision twin model, further reconstructs the power grid operation and dispatching decision twin model, and simulates power grid dispatching training through multiple twin models, thereby providing a more realistic training environment, realizing power grid dispatching training based on panoramic digital twin simulation, improving the training efficiency and quality of dispatchers, reducing training costs, and enhancing the safety and reliability of power grid dispatching. Provide a comprehensive, real, and interactive training platform for power grid dispatchers to effectively improve their professional skills and emergency handling capabilities. BRIEF DESCRIPTION OF THE DRAWINGS
[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0053] Figure 1 A schematic flow chart of a power grid dispatching training method based on panoramic digital twin simulation provided by the present invention.
[0054] Figure 2 A structural framework diagram of a power grid dispatching training system based on panoramic digital twin simulation provided by the present invention. DETAILED DESCRIPTION
[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0056] The embodiment of the present invention discloses a power grid dispatching training method based on panoramic digital twin simulation, such as Figure 1 As shown, including:
[0057] Obtain the operation data of the power grid and build a digital twin model of the power grid;
[0058] Conduct grid operation scenarios and fault simulations based on grid operation data;
[0059] Make grid operation and dispatch decisions based on the selected grid operation scenarios and fault types using the grid digital twin model;
[0060] Provide feedback and evaluation on grid operation and dispatch decisions.
[0061] Furthermore, the obtaining of the operation data of the power grid includes: collecting the operation data of power stations, substations and transmission lines.
[0062] Furthermore, the construction of the power grid digital twin model includes:
[0063] Determine the power grid absorption coefficient based on power load data;
[0064] Conduct power grid consumption evaluation on the target area according to the power grid consumption coefficient;
[0065] Data twinning is performed on the power absorption data and power transportation data according to the consumption assessment results, and multiple data twin layers are constructed according to the data twin space;
[0066] Couple multiple data twin layers to build a power grid digital twin model.
[0067] In a specific implementation, the power grid absorption coefficient is determined by extracting power load data, wherein the power grid absorption coefficient includes power absorption data and power transportation data; the power consumption of the power grid in the target area is recorded to obtain power data; the power consumption period in the target area is mapped according to the power data, the power consumption period is used as the first coordinate axis, and the power data is used as the second coordinate axis to construct a power coordinate system; the power data of all power consumption periods in the power coordinate system are added to obtain power load data; the power load data is decomposed to obtain power absorption load data and power transportation load data of the power grid in the target area; the power absorption load data and the power transportation load data are respectively trend predicted to obtain power absorption data and power transportation data; the power grid absorption coefficient is determined based on the power absorption data and the power transportation data;
[0068] Obtaining the consumption evaluation result, which is obtained by evaluating the power grid in the target area based on the power grid consumption coefficient;
[0069] According to the results of the consumption assessment, data twinning is performed on the power absorption data and the power transportation data, M data twin layers are constructed according to the data twin space, weighted calculation is performed on the power absorption data and the power transportation data based on the results of the consumption assessment, and a power grid operation spectrum is drawn according to the calculation results; multiple smart sensors are deployed based on the power grid operation spectrum; historical power grid operation data is extracted using multiple smart sensors to clone the power absorption data and the power transportation data, and a data twin space is constructed based on the cloned data set; the data twin space is classified to obtain M power data classes; M data twin layers are constructed based on the M power data classes; wherein M is a positive integer greater than 1;
[0070] Couple M data twin layers to obtain a power grid digital twin model; extract the call mapping relationship in the M data twin layers and determine the call mapping factor; superimpose the M data twin layers according to the call mapping factor to determine whether the data dependency index between the superimposed layers is in the preset data coupling interval; if so, the M data twin layers in the superimposed layers are deemed to be independent, and the call mapping factor is the minimum value; construct a power grid digital twin model based on the call mapping factor, superimposed layers and data coupling relationship; if not, determine whether the data dependency index between the superimposed layers is greater than the preset data coupling interval; if greater, the M data twin layers in the superimposed layers are deemed to be not independent, and the call mapping factor is the maximum value; decouple the M data twin layers until the data dependency index between the superimposed layers is in the preset data coupling interval.
[0071] Furthermore, the power grid operation scenario and fault simulation according to the power grid operation data includes:
[0072] Collect alarm information when an alarm signal appears in the power grid;
[0073] Through knowledge extraction, knowledge fusion and knowledge processing of alarm information, a knowledge graph of power grid operation scenarios is constructed;
[0074] A graph neural network with an attention mechanism is used to read the graph structure features of the knowledge graph of power grid operation scenarios, calculate the attention scores between nodes to obtain the node feature vector, integrate the entity features and relationship features of the nodes, obtain the bidirectional hidden state, and calculate the entity hidden feature vector;
[0075]
[0076] Among them, p′ i represents the forward hidden feature vector, p′ i ′ represents the backward hidden feature vector, Representation Matrix and d′ (l+1) Multiplication operation, is the relation sparse matrix of the relation set of node i, d′ (l+1) is the hidden state of the set of neighbor nodes of node i.
[0077] Reasoning event fault types in knowledge graphs based on graph neural networks with attention mechanisms;
[0078] The grid operation scenarios and fault event data are mapped according to the event fault types.
[0079] Furthermore, the grid operation and dispatch decision making of the grid digital twin model according to the selected grid operation scenario and fault type includes:
[0080] Set event fault types according to the power grid dispatch training plan;
[0081] Apply the grid operation scenarios and fault event data mapped to the event fault types to the grid digital twin model;
[0082] The digital twin model of the power grid is used to make grid operation and dispatch decisions to handle fault events.
[0083] Furthermore, it also includes building a twin model of power grid operation and dispatch decision-making, and outputting decisions through the twin model of power grid operation and dispatch decision-making:
[0084] Select the distribution network area to be digitally twinned, extract the network topology model of the area to be operated, and build a digital twin model of the distribution automation operation area based on the mapping of the actual physical area of the distribution network based on real-time physical data and environmental data.
[0085] Furthermore, the physical data is real-time physical data of the power distribution area mapped by the power consumption collection system and the SCADA system, including:
[0086] Collect system data deviations, voltage, current, power physical quantity deviations and switch status deviations, and feed them back to the power collection system and SCADA system;
[0087] Based on the environmental data obtained by the meteorological system, the association analysis algorithm is used to identify abnormal operations, physical
[0088] The alarm is associated with weather, temperature, humidity, and altitude. Based on linear fitting or logistic regression algorithms, a nonlinear fitting relationship between physical data, switch state quantities and temperature, humidity, and altitude environments is constructed to correct data deviations in the simulation environment.
[0089] Furthermore, it also includes the optimization of the twin models of power grid operation and dispatch decision-making:
[0090] Acquiring position feature data and structural feature data;
[0091] Collect the change sequence of the twin model, and sample the change sequence every four frames at most to obtain the motion information of the node; expand the position feature by adding random normal distribution perturbations to generate position feature training data; collect the structural data of different devices, and expand it by adding random normal distribution perturbations to generate structural feature training data;
[0092] A node regression matrix of a pre-trained twin model, wherein the node regression matrix is an interpolation regression matrix, which is used to convert the frame coordinates of the twin model into transformed reconstructed frame coordinates;
[0093] The frame coordinates of the twin model are The transformed reconstruction frame coordinates are Let the interpolation regression matrix be O up , minimize the following energy function to train O up :
[0094] E up =|O up Gi|2+E s ;
[0095] Among them, E s refers to the regularization term that constrains the topological symmetry of the human body;
[0096] Generate a set of rotation matrices and translation vectors as global rigid motion perturbations, and randomly extract corresponding position features and structural features from the position feature data and structural feature data;
[0097] According to the position features, structural features, rotation matrix and translation vector, the twin model features and frame coordinates that are mapped to each other are obtained, and the node regression matrix is used to convert the paired matching model features and frame coordinates into the transformed output frame;
[0098] The framework is: G tar =G(R, T, α, β) and I tar =O up G tar ;
[0099] Among them, the position feature is a, the structural feature is β, the rotation matrix is R, and the translation vector is T.
[0100] Training the feedback recurrent neural network based on the mutually mapped model features and frame coordinates;
[0101] The transformed framework is input into the trained feedback recurrent neural network, and the corresponding twin model features are extracted after iteration of the feedback recurrent neural network to complete the reconstruction of the twin model of power grid operation and dispatch decision.
[0102] Furthermore, the feedback and evaluation of the grid operation and dispatching decision include:
[0103] Perform data processing on the fault information mapped to the event fault type to obtain a training set;
[0104] Based on the K-fold cross-validation method, the training set is input into multiple classifiers for training to obtain different complementary subsets. The K classification accuracy rates obtained by performing K cross-validations on the complementary subsets are averaged to obtain the recall rates of different fault types of multiple classifiers.
[0105] Specifically, a five-fold cross validation is performed on multiple initial classifiers based on the input training set samples to obtain the recall rate R of different categories of each target classifier, which can be expressed as: R = (r ij ) m×n , r ij is the recall rate of the sample of the i-th category on the j-th classifier; at the same time, after multiple trainings, the parameters in the initial classifier are continuously adjusted, and finally the mapping function of each target classifier is obtained, which is expressed as: j (x), j = 1, 2, 3...n, j is the number of classifiers.
[0106] According to the recall rate, multiple classifiers are weighted and fused to obtain the target fault classification model;
[0107]
[0108] Among them, r ij is the recall rate of the sample with the i-th category label on the j-th target classifier, p ij is the predicted probability value of the jth target classifier classifying the category into the i-th category, n represents the number of target classifiers that classify the category into the i-th category label, L i It represents the final score of the target fault classification model that is finally classified into category label i.
[0109] The target fault classification model is used to perform fault detection on the information data after the grid operation and dispatch decision, and the detection results are output.
[0110] The multiple classifiers are one or more arbitrary combinations of K nearest neighbor classifiers, decision tree classifiers, support vector machine classifiers, Bayesian classifiers or random forest classifiers.
[0111] In a specific implementation manner, the data processing of the fault information mapped to the event fault type to obtain a training set includes:
[0112] Perform data preprocessing on the fault information to obtain an initial training set;
[0113] Identify the fault information to determine missing samples; process the missing samples to obtain an initial training set; or standardize the fault information to obtain an initial training set;
[0114]
[0115] in, is the standardized data of the fault information sample of the i-th category label, x i is the original data in the fault information, x max is the maximum value of the fault information sample of the i-th category label, x min is the minimum value of the fault information sample of the i-th category label;
[0116] Alternatively, a correlation analysis is performed on the fault information to determine the correlation of each sample in the fault information sample set, and samples are screened according to the correlation of each sample in the fault information to obtain an initial training set.
[0117] Adjust the imbalance ratio of the initial training set to obtain the training set;
[0118] Oversampling the minority class samples in the initial training set to obtain an oversampled initial training set;
[0119] Under-sampling is performed on the over-sampled initial training set to remove noise values of the over-sampled initial training set, thereby obtaining the training set with unbalanced data eliminated.
[0120] In a specific implementation, by performing SMOTE oversampling on the minority class samples, the minority class samples are increased to the same number as the majority class. The calculation process of oversampling is as follows:
[0121] x new =x+random(0,1)*(x′-x);
[0122] Among them, x is the data of the minority class sample, randomly select a nearest neighbor sample x′, and then randomly select a sample point x from the line connecting x to x′ new As new minority class sample data; for each fault type with a category label, the SMOTE oversampling algorithm is applied to increase the number of minority class samples to the same number as the majority class samples with the largest number of categories.
[0123] The data set after SMOTE oversampling is globally undersampled to remove the noise of each category label fault information data. Specifically, ENN undersampling can also be used. ENN undersampling is to obtain K nearest neighbor samples of the majority class sample. If all or most of these nearest neighbor samples are inconsistent with the majority class sample type, the sample is considered to be a noise value and is removed. Using the ENN algorithm can reduce the number of majority class samples to a certain extent.
[0124] The present invention trains multiple classifiers, uses the recall rate for weighted fusion, constructs a target fault classification model, uses the recall rate as the weight for weighted fusion of the target fault classification model, and trains the model by voting based on the weighted accuracy of the category, which can reflect the advantages of different single classifiers, thereby further improving the accuracy and stability of effective classification and prediction of the training model. The information data after the grid operation and dispatch decision made by the target fault classification model is input into the constructed target fault classification model, and the weighted fusion classification model is used to predict the fault type, which can accurately identify the type of power failure information data fault, thereby achieving accurate classification and prediction of the fault.
[0125] In a specific implementation, based on the real-time information set of the power grid digital twin model, a state prediction score of the power grid digital twin model is obtained;
[0126] Obtaining a preset state prediction score threshold;
[0127] Determine whether the state prediction score of the power grid digital twin model meets a preset state prediction score threshold;
[0128] If the state prediction score of the power grid digital twin model does not meet the preset state prediction score threshold, multiple proposed rectification and treatment plans for the power grid digital twin model are obtained;
[0129] The digital twin model of the power grid is evaluated in real time according to the real-time information collection of the digital twin model of the power grid and the real-time evaluation model of the dispatching, so as to obtain the real-time evaluation result of the digital twin model of the power grid;
[0130] Construct a real-time evaluation model of the power grid digital twin model based on neural networks;
[0131] Inputting data information in the real-time information set of the power grid digital twin model into the real-time evaluation model of the power grid digital twin model;
[0132] Obtaining output information of a real-time evaluation model of a power grid digital twin model, the output information including a real-time evaluation result of the power grid digital twin model;
[0133] Grid operation and dispatch decision evaluation is performed based on real-time evaluation results.
[0134] The present invention uses multiple twin models to simulate power grid dispatching training, thereby providing a more realistic training environment, improving the training efficiency and quality of dispatchers, reducing training costs, and enhancing the safety and reliability of power grid dispatching. It provides a comprehensive, realistic, and interactive training platform for power grid dispatchers, effectively improving their professional skills and emergency response capabilities.
[0135] In a specific implementation, a power grid dispatching training system based on panoramic digital twin simulation, such as Figure 2 As shown, including:
[0136] Power grid digital twin model construction module: used to obtain power grid operation data and build a power grid digital twin model;
[0137] Power grid operation scenario and fault simulation module: used to simulate power grid operation scenarios and faults based on power grid operation data;
[0138] Grid operation and dispatch decision module: used to make grid operation and dispatch decisions for the grid digital twin model according to the selected grid operation scenario and fault type;
[0139] Feedback and evaluation module: used for feedback and evaluation of power grid operation and dispatch decisions.
[0140] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.
[0141] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A power grid dispatching training method based on panoramic digital twin simulation, characterized in that: include: Obtain the operation data of the power grid and build a digital twin model of the power grid; Conduct grid operation scenarios and fault simulations based on grid operation data; Make grid operation and dispatch decisions based on the selected grid operation scenarios and fault types using the grid digital twin model; Provide feedback and evaluation on grid operation and dispatch decisions.
2. A power grid dispatching training method based on panoramic digital twin simulation according to claim 1, characterized in that: The obtaining of the operation data of the power grid includes: collecting the operation data of power stations, substations and transmission lines.
3. The power grid dispatching training method based on panoramic digital twin simulation according to claim 1 is characterized in that: The construction of the power grid digital twin model includes: Determine the power grid absorption coefficient based on power load data; Conduct power grid consumption evaluation on the target area according to the power grid consumption coefficient; Data twinning is performed on the power absorption data and power transportation data according to the consumption assessment results, and multiple data twin layers are constructed according to the data twin space; Couple multiple data twin layers to build a power grid digital twin model.
4. The power grid dispatching training method based on panoramic digital twin simulation according to claim 1 is characterized in that: The power grid operation scenario and fault simulation according to the power grid operation data includes: Collect alarm information when an alarm signal appears in the power grid; Extract knowledge, fuse knowledge, and process knowledge of alarm information to build a knowledge graph of power grid operation scenarios; A graph neural network with an attention mechanism is used to read the graph structure features of the knowledge graph of power grid operation scenarios, calculate the attention scores between nodes to obtain the node feature vector, integrate the entity features and relationship features of the nodes, obtain the bidirectional hidden state, and calculate the entity hidden feature vector; Reasoning event fault types in knowledge graphs based on graph neural networks with attention mechanisms; The grid operation scenarios and fault event data are mapped according to the event fault types.
5. The power grid dispatching training method based on panoramic digital twin simulation according to claim 1 is characterized in that: The grid operation and dispatch decision-making of the grid digital twin model according to the selected grid operation scenario and fault type includes: Set event fault types according to the power grid dispatch training plan; Apply the grid operation scenarios and fault event data mapped to the event fault types to the grid digital twin model; The digital twin model of the power grid is used to make grid operation and dispatch decisions to handle fault events.
6. A power grid dispatching training method based on panoramic digital twin simulation according to claim 5, characterized in that: It also includes building a twin model of power grid operation and dispatch decision-making, and outputting decisions through the twin model of power grid operation and dispatch decision-making: Select the distribution network area to be digitally twinned, extract the network topology model of the area to be operated, and build a digital twin model of the distribution automation operation area based on the mapping of the actual physical area of the distribution network based on real-time physical data and environmental data.
7. A power grid dispatching training method based on panoramic digital twin simulation according to claim 6, characterized in that: The physical data is based on the power consumption collection system and the SCADA system mapping the real-time physical data of the power distribution area, including: Collect system data deviations, voltage, current, power physical quantity deviations and switch status deviations, and feed them back to the power collection system and SCADA system; Based on the environmental data obtained by the meteorological system, the correlation analysis algorithm is used to identify the correlation between abnormal operations, physical alarms and weather, temperature, humidity, and altitude. Based on the linear fitting or logistic regression algorithm, the nonlinear fitting relationship between physical data, switch state quantities and temperature, humidity, and altitude environment is constructed to correct data deviations in the simulation environment.
8. The power grid dispatching training method based on panoramic digital twin simulation according to claim 6 is characterized in that: It also includes optimization of the twin models of power grid operation and dispatch decision-making: Acquiring position feature data and structural feature data; A node regression matrix of a pre-trained twin model, wherein the node regression matrix is an interpolation regression matrix, which is used to convert the frame coordinates of the twin model into transformed reconstructed frame coordinates; Generate a set of rotation matrices and translation vectors as global rigid motion perturbations, and randomly extract corresponding position features and structural features from the position feature data and structural feature data; According to the position features, structural features, rotation matrix and translation vector, the twin model features and frame coordinates that are mapped to each other are obtained, and the node regression matrix is used to convert the paired matching model features and frame coordinates into the transformed output frame; Training the feedback recurrent neural network based on the mutually mapped model features and frame coordinates; The transformed framework is input into the trained feedback recurrent neural network, and the corresponding twin model features are extracted after iteration of the feedback recurrent neural network to complete the reconstruction of the twin model of power grid operation and dispatch decision.
9. The power grid dispatching training method based on panoramic digital twin simulation according to claim 1 is characterized in that: The feedback and evaluation of grid operation and dispatch decisions include: Perform data processing on the fault information mapped to the event fault type to obtain a training set; Based on the K-fold cross-validation method, the training set is input into multiple classifiers for training to obtain different complementary subsets. The K classification accuracy rates obtained by performing K cross-validations on the complementary subsets are averaged to obtain the recall rates of different fault types of multiple classifiers. According to the recall rate, multiple classifiers are weighted and fused to obtain the target fault classification model; The target fault classification model is used to perform fault detection on the information data after the grid operation and dispatch decision, and the detection results are output.
10. A power grid dispatching training system based on panoramic digital twin simulation, characterized in that: include: Power grid digital twin model construction module: used to obtain power grid operation data and build a power grid digital twin model; Power grid operation scenario and fault simulation module: used to simulate power grid operation scenarios and faults based on power grid operation data; Grid operation and dispatch decision module: used to make grid operation and dispatch decisions for the grid digital twin model according to the selected grid operation scenario and fault type; Feedback and evaluation module: used for feedback and evaluation of power grid operation and dispatch decisions.