A topology error prevention risk management method and system

Abstract

The present invention discloses a topology error prevention risk management method and system, which relates to the field of risk management technology, and comprises the following steps: obtaining a power grid topology structure diagram, and using a graph algorithm to encode the power grid nodes in the topology structure diagram to obtain a node feature vector; constructing a correction parameter group based on the measurement missing node identification result and the node level, and randomly perturbing the correction parameter group based on a preset first constraint condition to obtain a number of correction schemes; respectively using the correction scheme to repair the node feature vectors of the measurement missing nodes to obtain a node repair feature vector set, and performing similarity analysis with preset risk scenarios respectively; calculating the optimal similarity threshold corresponding to the correction scheme based on the similarity analysis result, constructing an optimal pairing group and applying it to actual error prevention risk identification, thereby solving the problem that when faced with node measurement missing, the repaired node features lack a dynamic adaptation mechanism to the actual risk scenario, resulting in reduced accuracy of power grid risk management.

Description

Technical Field

[0001] The present invention relates to the technical field of power system topology error prevention control, and more specifically, to a topology error prevention risk management method and system. Background Art

[0002] As power systems continue to expand in size and become increasingly complex, the safety and reliability requirements for grid operations are becoming increasingly stringent. To prevent operational errors during grid operations, topology risk management and control technologies are becoming a crucial component of grid dispatching and operational support systems. Existing risk management and control methods primarily rely on the integrity of grid measurement data. These methods utilize real-time data collection, such as current, voltage, and switch status, combined with the grid topology to assess operational risk and conduct risk verification.

[0003] For example, the invention patent announcement with publication number CN114256976B discloses a system and method for preventing errors during field operations in a distribution network. The system synchronizes field maintenance work orders for distribution network operations, objectifies the maintenance work orders and corresponding operation tickets, and forms an anti-error operation sequence list. The system then performs anti-error locking on the operation tickets based on the error operation sequence list, selects the corresponding anti-error operation sequence, and performs anti-error control on the operation execution. A network topology-based anti-error verification method is used to determine whether the anti-error verification result is consistent with the operation target of the anti-error operation sequence. The system also interacts with field personnel to send instructions for starting field operations, performing field operations, and ending operations. The system also detects the status of automated and non-automated switches to determine the equipment status. Compared to existing technologies, the present invention achieves tracking and anti-error control of the execution process of work orders and switching operation tickets through real-time feedback on the equipment status during the operation process.

[0004] The above disclosed technical solutions have at least the following technical problems:

[0005] However, due to factors such as communication anomalies, equipment failures, and on-site environmental interference, the power grid measurement data is missing, delayed, or abnormal to a certain extent, resulting in the inability to obtain measurement information of some nodes in a timely and accurate manner. When faced with missing node measurements, traditional anti-misidentification methods usually use default value filling, simple interpolation, or ignoring processing, which cannot fully restore the true operating status of the node, and can easily lead to misjudgment and missed judgment, reducing the accuracy and reliability of anti-misidentification and increasing the potential risks of power grid operation. In addition, the repaired node characteristics lack a dynamic adaptation mechanism to the actual risk scenario, and the truth filling and risk identification are separated, making it difficult to achieve the optimal coordination effect, resulting in a reduction in the accuracy of power grid risk management. In response to the above problems, the present invention proposes a solution. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a topology error prevention risk management method, which evaluates the joint optimal matching of the compensation scheme and the risk identification sensitivity to solve the problem that when faced with missing node measurements, the repaired node characteristics lack a dynamic adaptation mechanism to the actual risk scenario, making it difficult to achieve the optimal coordination effect, resulting in a reduction in the accuracy of power grid risk management.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A topology error prevention risk management and control method comprises the following steps: obtaining a power grid topology diagram, and encoding power grid nodes using a graph algorithm to obtain node feature vectors; constructing a correction parameter group based on measurement missing node identification results and node levels, and randomly perturbing the correction parameter group based on a preset first constraint condition to obtain several correction schemes; respectively repairing the node feature vectors of the measurement missing nodes using the correction schemes to obtain node repair feature vector sets, and performing similarity analysis with preset risk scenarios respectively; and calculating the optimal similarity threshold corresponding to the correction scheme based on the similarity analysis results, constructing an optimal pairing group, and applying it to actual error prevention risk identification.

[0009] In a preferred embodiment, the method of obtaining a power grid topology diagram and encoding the power grid nodes using a graph algorithm to obtain a node feature vector is as follows: obtaining a power grid topology diagram, wherein the power grid topology diagram includes nodes, connection lines and electrical parameters; constructing an adjacency matrix according to the power grid topology diagram, and randomly traversing each node using a graph algorithm based on preset encoding hyperparameters to obtain a node context sequence; inputting the node context sequence into a preset node embedding model, extracting the mapping vector corresponding to each node, and obtaining a node feature vector.

[0010] In a preferred embodiment, the method for constructing a correction parameter group based on the measurement missing node identification result and the node level is specifically as follows: obtaining the measurement parameters of the power grid nodes in real time and constructing a node measurement matrix, and using a time series prediction algorithm to predict the measurement parameters of the nodes to obtain a first prediction matrix; calculating the element difference between the node measurement matrix and the first prediction matrix to obtain a residual matrix of the measurement parameters of each node; performing data analysis on the residual matrix to determine whether there is measurement missing at the node, and obtaining the missing node and missing parameter type; according to the node level division rule, using a clustering algorithm to perform clustering level division on the missing nodes to obtain the level label of the missing node; combining the level label of the missing node and the missing parameter type to generate a correction parameter group.

[0011] In a preferred embodiment, the random perturbation of the compensation parameter group based on the preset first constraint condition is performed to obtain several compensation schemes, specifically: the grid stability constraint is used as the first constraint condition, and the first constraint condition includes the voltage fluctuation range, the line capacity upper limit and the power balance condition; based on the level label of the missing node and the missing parameter type, a first perturbation rule is constructed and the compensation parameter group is randomly perturbed to obtain several initial compensation schemes; a grid state simulation calculation is performed on each initial compensation scheme to obtain a disturbance simulation evaluation value; whether the initial compensation scheme is a valid disturbance is determined according to the disturbance simulation evaluation value to obtain several valid disturbance schemes; and the Pareto front analysis method is used to screen the valid disturbance schemes to obtain several compensation schemes.

[0012] In a preferred embodiment, the node feature vectors of the measurement missing nodes are repaired by using the correction scheme to obtain a node repair feature vector set, specifically: the feature vectors of the adjacent nodes of the measurement missing node are used as the basic value and the preset correction weight is used as the coefficient to construct a linear regression model; based on the linear regression model, the weighted least squares method is used to repair the feature vectors of the measurement missing nodes to obtain a node repair feature vector set.

[0013] In a preferred embodiment, the similarity analysis is performed with preset risk scenarios respectively, specifically as follows: risk scenario vectors are extracted from a historical fault database, and a risk propagation path tree is constructed; based on the risk propagation path tree, a dynamic time warping algorithm is used to align the node repair feature vector with the risk scenario vector to obtain several first alignment paths; second constraints are constructed based on preset electrical constraints and topological path constraints, and the first alignment paths are screened to obtain a filtered valid path set; the minimum path distance is calculated based on the filtered valid path set and normalized according to the preset path weight to obtain the similarity between each residual compensation scheme and the risk scenario, and a similarity distribution table for each residual compensation scheme is constructed.

[0014] In a preferred embodiment, the method of calculating the optimal similarity threshold corresponding to the correction scheme based on the similarity analysis result is specifically as follows: based on the similarity distribution table of each correction scheme, a clustering algorithm is used to initially divide the similarity to obtain a number of similarity intervals; the ratio of the similarity mean and standard deviation of each similarity interval is calculated and a confidence interval is constructed; based on the confidence interval, a preset first threshold of each similarity interval is screened to obtain a second threshold of each similarity interval; based on the second threshold, a particle swarm optimization algorithm is used to iteratively optimize the similarity threshold of each similarity interval to obtain a first similarity threshold of each similarity interval; and a weighted average of the first similarity threshold of each similarity interval is performed to obtain the optimal similarity threshold corresponding to the correction scheme.

[0015] In a preferred embodiment, based on the second threshold, the similarity threshold of each similarity interval is iteratively optimized using a particle swarm optimization algorithm to obtain a first similarity threshold of each similarity interval, specifically: with the second threshold of each similarity interval as the center, a number of threshold combinations to be optimized are randomly initialized according to a preset range; the threshold combinations to be optimized are used as particles, and an initial particle swarm is constructed; a fitness function is constructed, the initial particle swarm is searched iteratively, and the fitness of each particle is calculated; when the iteration reaches a preset termination condition, the iteration is stopped and the first similarity threshold of each similarity interval is output.

[0016] In a preferred embodiment, the construction of the optimal pairing group and its application in actual error prevention risk identification are specifically as follows: based on the optimal similarity threshold corresponding to the correction scheme, the optimal pairing group is constructed, and the optimal pairing group includes correction parameters, node repair feature vectors and landscape scenes; the correction parameters in the optimal pairing group are pushed to the power grid monitoring system in real time to trigger automatic repair instructions; based on the node repair feature vectors, a preset risk prediction model is used to dynamically evaluate the power grid status; for each risk scenario, an emergency plan is dynamically activated and a risk disposal report is generated for archiving.

[0017] The technical effects and advantages of the topology error prevention risk management method and system of the present invention are as follows:

[0018] 1. The present invention obtains node feature vectors by acquiring a grid topology diagram and encoding nodes using a graph algorithm, thereby establishing a unified quantitative foundation for subsequent repair and identification. By identifying missing node measurements and constructing a correction parameter group based on node levels, accurate compensation for potential node feature loss is achieved, improving the pertinence and adaptability of the correction processing. Furthermore, the correction parameter group is randomly perturbed based on preset constraints to generate multiple sets of residual correction schemes, effectively expanding the diversity of correction strategies and enhancing the robustness of the system in dynamic operating environments.

[0019] 2. The present invention repairs node feature vectors by adopting different residual correction schemes to form multiple node repair feature vector sets, and performs similarity analysis on the repaired feature vectors with preset risk scenario vectors, dynamically searches for the optimal similarity threshold, thereby achieving a joint optimal match between the correction effect and the risk identification sensitivity; finally, it can accurately identify potential high-risk equipment combinations in the context of measurement loss, improve the accuracy and reliability of anti-mistake risk identification, and has good engineering applicability and promotion prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 The figure is a flow chart of a topology error prevention risk management method of the present invention.

[0021] Figure 2This is a structural diagram of a topology error prevention risk management and control system of the present invention. DETAILED DESCRIPTION

[0022] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0023] Example 1, Figure 1 The present invention provides a topology error prevention risk management method, comprising the following steps:

[0024] S1, obtain the power grid topology diagram, and use the graph algorithm to encode the power grid nodes to obtain the node feature vector;

[0025] In this example, the power grid topology diagram is obtained, and the graph algorithm is used to encode the power grid nodes to obtain the node feature vector, which is specifically:

[0026] Obtaining a power grid topology diagram, wherein the power grid topology diagram includes nodes, connection lines, and electrical parameters;

[0027] An adjacency matrix is constructed based on the power grid topology graph, and a graph algorithm is used to randomly traverse each node based on preset encoding hyperparameters to obtain a node context sequence.

[0028] The node context sequence is input into the preset node embedding model, and the mapping vector corresponding to each node is extracted to obtain the node feature vector.

[0029] It's important to note that encoding hyperparameters refer to the set of control parameters set for the algorithm when using a graph algorithm to encode nodes in a power grid topology (i.e., extract feature vectors). These parameters are not automatically learned through training but must be manually set before the algorithm runs. They directly affect the effectiveness of node encoding and the expressiveness of the results. Hyperparameters include walk length, number of sampled walks per node, window size, return parameters, and exploration parameters.

[0030] For example, the walk length in the hyperparameters is 80, the number of sampling walks for each node is 10, the window size is 5, the return parameter p=1 and the exploration parameter q=0.5, and each node is randomly traversed based on the hyperparameter settings to obtain the node context sequence.

[0031] S2, constructing a correction parameter group based on the missing identification results of the node measurement and the node level, and randomly perturbing the correction parameter group based on the preset first constraint condition to obtain several correction schemes;

[0032] In this example, a correction parameter group is constructed based on the measurement missing node identification results and node levels, specifically:

[0033] Acquire the measurement parameters of the power grid nodes in real time and construct a node measurement matrix, and use a time series prediction algorithm to predict the measurement parameters of the nodes to obtain a first prediction matrix;

[0034] Calculate the element difference between the node measurement matrix and the first prediction matrix to obtain the residual matrix of the measurement parameters of each node;

[0035] Perform data analysis on the residual matrix to determine whether there are missing measurements on the nodes, and obtain the missing nodes and missing parameter types;

[0036] According to the node level division rules, a clustering algorithm is used to divide the missing nodes into cluster levels and obtain the level labels of the missing nodes;

[0037] Generate a complement parameter group by combining the level label of the missing node and the type of the missing parameter.

[0038] It should be noted that the measured parameters of nodes include node voltage amplitude, node voltage phase angle, active power injection, reactive power injection, current amplitude, and node frequency. Furthermore, by predicting node measurement parameters using a time series prediction algorithm, historical data can be used to accurately estimate the current node status, effectively reducing errors caused by data loss or anomalies. Secondly, the calculation of the residual matrix provides a quantitative basis for missing measurement identification. By comparing the difference between actual measured and predicted values, it accurately determines which nodes have missing measurements, avoiding errors caused by human intervention. Furthermore, based on the node classification rules, a clustering algorithm is used to classify missing nodes. Nodes can be divided into core nodes, key nodes, and ordinary nodes. Core nodes correspond to power grid trunk lines, key nodes correspond to regional hubs, and ordinary nodes correspond to terminal loads. Combining the clustering algorithm to classify missing nodes, an automated clustering method is used to rationally allocate compensation resources based on node importance (such as centrality and load). In this way, the system can dynamically adjust the verification strategy according to the importance of the node, rather than treating all nodes in a one-size-fits-all manner, ensuring that key nodes are corrected first and optimizing resource allocation.

[0039] In this example, the compensation parameter group is randomly perturbed based on the preset first constraint condition to obtain several compensation schemes, specifically:

[0040] Using grid stability constraints as the first constraint condition, the first constraint condition including voltage fluctuation range, line capacity upper limit and power balance condition;

[0041] Based on the level labels of the missing nodes and the types of missing parameters, a first perturbation rule is constructed and the correction parameter group is randomly perturbed to obtain several initial correction schemes;

[0042] Perform grid state simulation calculation on each initial compensation scheme to obtain disturbance simulation evaluation value;

[0043] According to the disturbance simulation evaluation value, it is judged whether the initial compensation scheme is an effective disturbance, and several effective disturbance schemes are obtained;

[0044] The Pareto front analysis method is used to screen the effective perturbation schemes and obtain several compensation schemes.

[0045] The calculation formula of the disturbance simulation evaluation value is as follows:

[0046]

[0047] in, is the disturbance simulation evaluation value, is the simulated voltage value of the missing node i after filling in the current initial filling scheme, is the rated voltage value of the i-th missing node, is the simulated current value of the jth line under the current initial compensation scheme, is the maximum current allowed by the jth line, is the number of missing nodes, is the number of lines in the grid.

[0048] It is important to note that, first, grid stability constraints are used as the first constraint, comprehensively considering the voltage fluctuation range, line capacity limits, and power balance conditions to ensure that the perturbation scheme is feasible within the laws of electrical physics and safe operating boundaries. By combining the level labels of missing nodes and the types of missing parameters, differentiated perturbation rules are designed to randomly perturb the original compensation parameter set to generate multiple initial compensation schemes. Subsequently, grid state simulations are performed for each perturbation scheme, and perturbation simulation evaluation values are calculated to quantify voltage deviation, line overload ratio, and power imbalance, thereby screening effective perturbation schemes that meet stability requirements. Finally, a Pareto front analysis method is introduced to select the optimal compensation scheme from multiple effective perturbation schemes based on a multi-metric trade-off, achieving optimal decision-making for the scheme set. This method has strong adaptability and scalability, fully exploiting the diversity of the compensation space while ensuring grid security for the perturbation schemes. It improves the accuracy, stability, and practical usability of parameter compensation schemes in measurement-missing scenarios, making it suitable for real-time, robust state estimation and control support in large-scale power grid systems.

[0049] S3, respectively use the residual filling scheme to repair the node feature vectors, obtain several node repair feature vector sets, and perform similarity analysis with the preset risk scenario vectors;

[0050] In this example, the node feature vectors of the measurement missing nodes are repaired using the true filling scheme to obtain the node repair feature vector set, which is specifically:

[0051] The linear regression model is constructed by taking the feature vectors of the adjacent nodes of the missing node as the basic value and the preset filling weight as the coefficient.

[0052] Based on the linear regression model, the weighted least squares method is used to repair the feature vectors of the measurement missing nodes and obtain the node repair feature vector set.

[0053] It should be noted that, firstly, this method introduces an integrated mechanism for multiple disturbance compensation schemes. This comprehensively considers multiple disturbance scenarios that meet grid stability constraints when constructing the node feature repair model, thereby improving the adaptability and robustness of the repair feature vector to future trends. Traditional interpolation methods, on the other hand, typically only perform local inferences based on static adjacent node data and fail to consider multiple operational possibilities.

[0054] Secondly, by introducing the infill weights to construct a linear regression model, this method can reflect the actual influence of different adjacent nodes in the feature space, thereby retaining more structural dependencies and contextual features in the repair results; traditional methods cannot impose differentiated contributions on neighboring nodes, and their accuracy is limited.

[0055] Thirdly, this method uses weighted least squares in the regression process, which enhances the robustness to errors in the incomplete perturbation data and avoids the damage of abnormal perturbation values to the overall repair accuracy. Traditional interpolation methods are more sensitive to extreme values and missing data, and are prone to distortion or oscillation.

[0056] Finally, because this method combines node level labels with missing parameter types, it not only achieves adaptive perturbation constraint setting, but also enables the repair process to have differentiated processing capabilities, enhancing conservatism for key nodes and increasing flexibility for low-priority nodes. This hierarchical repair capability is difficult to achieve with conventional methods.

[0057] In this example, similarity analysis is performed with the preset risk scenarios, specifically:

[0058] Extract risk scenario vectors from the historical failure database and construct a risk propagation path tree;

[0059] Based on the risk propagation path tree, the dynamic time warping algorithm is used to align the node repair feature vector with the risk scenario vector to obtain several first alignment paths;

[0060] Constructing a second constraint condition based on the preset electrical constraint condition and topological path constraint condition, and filtering the first alignment path to obtain a filtered valid path set;

[0061] The minimum path distance is calculated based on the screened valid path set and normalized according to the preset path weight to obtain the similarity between each residual compensation scheme and the risk scenario, and a similarity distribution table of each residual compensation scheme is constructed.

[0062] The preset electrical constraints include node voltage amplitude range and line power flow constraints, specifically:

[0063] The node voltage amplitude range is specifically:

[0064]

[0065] in, is the voltage amplitude of the ith node, and are the minimum and maximum voltages allowed at the i-th node, A collection of nodes.

[0066] The specific line power flow constraints are:

[0067]

[0068] in, is the complex power from node i to node j, is the maximum permissible transmission capacity of the branch from node i to node j, is the collection of all branches in the power grid.

[0069] It's important to note that the risk propagation path tree is a graph structure built based on grid topology and fault propagation mechanisms. It describes how risks (such as faults or power fluctuations caused by faults) propagate from a specific grid node (e.g., a faulty node) along paths within the grid to other nodes. It helps identify the risk propagation path and the relationship between each node and the fault source.

[0070] Furthermore, a dynamic time warping (DTW) algorithm is used to time-align the feature vectors of the repaired nodes with the risk scenario vectors, generating several first-aligned paths. This process not only accounts for the temporal asynchrony of node changes but also preserves the dynamic trends of feature changes over time. To ensure the physical feasibility and electrical rationality of the paths, second constraints based on electrical constraints such as voltage, current, and power, as well as network topology connection logic, are introduced to filter the first-aligned paths, resulting in a set of filtered valid paths. Based on this set, the minimum distance of each path is further calculated and normalized using a preset path propagation weight factor. Ultimately, a similarity score is obtained between each correction solution and the historical risk scenario, and a corresponding similarity distribution table is constructed. This process not only improves the risk perception capabilities of the correction solution but can also be used for early warning and decision support.

[0071] S4, based on the similarity analysis results, calculates the optimal similarity threshold corresponding to the residual correction scheme, constructs the optimal pairing group and applies it to the actual error prevention risk identification process.

[0072] In this example, the optimal similarity threshold corresponding to the truth-filling solution is calculated based on the similarity analysis results, specifically:

[0073] Based on the similarity distribution table of each truth-filling scheme, a clustering algorithm is used to initially divide the similarity and obtain several similarity intervals;

[0074] Calculate the ratio of the similarity mean to the standard deviation for each similarity interval and construct a confidence interval;

[0075] Screening a preset first threshold value of each similarity interval based on the confidence interval to obtain a second threshold value of each similarity interval;

[0076] Based on the second threshold, the similarity threshold of each similarity interval is iteratively optimized using a particle swarm optimization algorithm to obtain a first similarity threshold of each similarity interval;

[0077] The first similarity threshold of each similarity interval is weighted averaged to obtain the optimal similarity threshold corresponding to the truth filling solution.

[0078] It should be noted that, first, based on the similarity distribution table for each truth-filling solution, a clustering algorithm is used to perform a preliminary division of similarity into several similarity intervals. Next, the ratio of the mean similarity to the standard deviation for each similarity interval is calculated, and a confidence interval is constructed. This allows for preliminary screening of the threshold for each interval, ensuring the high reliability of the threshold setting. The confidence interval provides a range for each similarity interval, indicating the range of similarity variation within that interval, within which there is a certain probability of containing the true similarity value. Based on the confidence interval, eligible similarity intervals and thresholds can be effectively screened, providing a stable foundation for subsequent particle swarm optimization. Based on the second threshold further screened based on the confidence interval, iterative optimization is performed in conjunction with the particle swarm optimization algorithm to adjust the threshold for each similarity interval. Particle swarm optimization can find the optimal threshold across multiple similarity intervals, thereby improving the adaptability and robustness of the solution. Finally, the weighted average of the first similarity thresholds for each similarity interval is taken to determine the optimal similarity threshold for the truth-filling solution.

[0079] In this example, based on the second threshold, the particle swarm optimization algorithm is used to iteratively optimize the similarity threshold of each similarity interval to obtain the first similarity threshold of each similarity interval, specifically:

[0080] Taking the second threshold of each similarity interval as the center, randomly initialize several threshold combinations to be optimized according to the preset range;

[0081] The threshold combination to be optimized is used as a particle, and an initial particle swarm is constructed;

[0082] Construct a fitness function, perform search iterations on the initial particle swarm and calculate the fitness of each particle;

[0083] When the iteration reaches a preset termination condition, the iteration is stopped and the first similarity threshold of each similarity interval is output.

[0084] It should be noted that, first, several threshold combinations to be optimized are randomly initialized within a preset range, centered around the second threshold of each similarity interval. These threshold combinations serve as particles in a particle swarm, with each particle representing a possible similarity threshold solution. Through the particle swarm optimization algorithm, the particle swarm continuously searches based on the initial random combinations and evaluates the performance of each particle using a fitness function. The fitness function reflects the performance of the threshold combination under certain predetermined criteria, helping the algorithm identify optimal solutions. As iterations proceed, each particle continuously adjusts its position within the search space, guided by historical and current optimal solutions, gradually approaching the optimal solution. The iterative process continues until a preset termination condition (such as a maximum number of iterations or convergence accuracy) is met, at which point the first similarity threshold for each similarity interval is output.

[0085] Furthermore, the particle swarm optimization algorithm effectively avoids the local optimal solution problem common in traditional methods. Leveraging the collective intelligence of particles and their local search capabilities, the PSO algorithm can quickly find the global optimal solution or a near-optimal solution within a large search space. Furthermore, particle swarm optimization is independent of the specific mathematical form of the problem and is applicable to complex, nonlinear optimization problems. It can adaptively adjust the search path, improving the efficiency and accuracy of the optimization process. This method is particularly well-suited for similarity optimization problems with uncertainty and multi-dimensional constraints, and can enhance system stability and reliability.

[0086] In this example, the optimal pairing group is constructed and applied to actual error-proof risk identification, specifically:

[0087] Constructing an optimal pairing group based on an optimal similarity threshold corresponding to a correction scheme, wherein the optimal pairing group includes correction parameters, node repair feature vectors, and landscape scenes;

[0088] Push the compensation parameters in the optimal pairing group to the power grid monitoring system in real time to trigger automatic repair instructions;

[0089] Based on the node repair feature vector, the preset risk prediction model is used to dynamically evaluate the power grid status;

[0090] For each risk scenario, the emergency plan is dynamically activated and a risk disposal report is generated for archiving.

[0091] It should be noted that, first, based on the optimal similarity threshold corresponding to the correction scheme, an optimal pairing group is constructed. This pairing group includes correction parameters, node repair feature vectors, and risk scenarios. This approach integrates correction parameters, node repair characteristics, and historical fault risk scenarios, forming an efficient decision-making support framework. Next, the correction parameters in the optimal pairing group are pushed to the power grid monitoring system in real time, triggering automatic repair instructions. This enables rapid response and automatic implementation of repair measures when potential grid issues arise, avoiding delays or errors caused by manual intervention. Based on the node repair feature vectors and utilizing a pre-set risk prediction model, the system dynamically assesses the grid's operating status, predicts potential fault risks, and takes preventive measures. Finally, for each risk scenario, the system dynamically activates the corresponding emergency plan, generates a risk management report, and archives it to provide a basis for subsequent risk analysis and resolution.

[0092] The advantage of this approach is that it not only quickly and accurately identifies potential grid failure risks but also improves the speed and accuracy of grid fault response by triggering automated repair instructions. Dynamic assessment and real-time feedback ensure the safety and stability of grid operations. Furthermore, the generation and archiving of emergency response plans based on risk scenarios provides data support for subsequent fault analysis and optimization, enhancing the intelligent level of grid management and emergency response capabilities.

[0093] Example 2, Figure 2 The present invention provides a topology error prevention risk management system, which includes a data acquisition module, a parameter generation module, a feature repair module, and a risk management module:

[0094] The data acquisition module is used to obtain the power grid topology diagram and use the graph algorithm to encode the power grid nodes to obtain the node feature vector;

[0095] A parameter generation module is used to construct a correction parameter group based on the measurement missing node identification result and the node level, and randomly perturb the correction parameter group based on a preset first constraint condition to obtain multiple correction schemes;

[0096] The feature repair module is used to repair the node feature vectors of the measurement missing nodes using the filling scheme, obtain the node repair feature vector set, and perform similarity analysis with the preset risk scenarios respectively;

[0097] The risk control module is used to calculate the optimal similarity threshold corresponding to the truth-filling solution based on the similarity analysis results, construct the optimal pairing group and apply it to actual error-proof risk identification.

[0098] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0099] The above embodiments may be implemented in whole or in part through software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0100] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0101] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0102] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0103] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A topology error prevention risk management method, characterized in that: The following steps are involved: Obtain a grid topology diagram, and use a graph algorithm to encode the grid nodes in the topology diagram to obtain node feature vectors; A correction parameter group is constructed based on the measurement missing node identification results and node levels, and the correction parameter group is randomly perturbed based on the preset first constraint condition to obtain several correction schemes; The node feature vectors of the measurement missing nodes are repaired using the filling scheme to obtain the node repair feature vector set, and the similarity analysis is performed with the preset risk scenarios respectively; Based on the similarity analysis results, the optimal similarity threshold corresponding to the truth-filling solution is calculated, and the optimal pairing group is constructed and applied to the actual error-prevention risk identification; Among them, the compensation parameter group is constructed based on the measurement missing node identification results and node levels, specifically: Constructing a node measurement matrix and using a time series prediction algorithm to predict measurement parameters of the node measurement matrix to obtain a first prediction matrix; Calculate the element difference between the node measurement matrix and the first prediction matrix to obtain the residual matrix of the measurement parameters of each node; Perform data analysis on the residual matrix to obtain the missing nodes and missing parameter types; Use clustering algorithm to classify missing nodes into clusters and obtain the level labels of missing nodes; Generate a complement parameter group by combining the level label of the missing node and the type of the missing parameter.

2. The topology error prevention risk management method according to claim 1, characterized in that: The graph algorithm is used to encode the grid nodes in the topology diagram to obtain the node feature vector, which is specifically: An adjacency matrix is constructed based on the power grid topology graph, and a graph algorithm is used to randomly traverse each node based on preset encoding hyperparameters to obtain a node context sequence. The node context sequence is input into the preset node embedding model, and the mapping vector corresponding to each node is extracted to obtain the node feature vector.

3. The topology error prevention risk management method according to claim 2, characterized in that: The random perturbation of the compensation parameter group based on the preset first constraint condition is performed to obtain several compensation schemes, specifically: Based on the level labels of the missing nodes and the types of missing parameters, a first perturbation rule is constructed and the correction parameter group is randomly perturbed to obtain several initial correction schemes; Performing a grid state simulation calculation on each initial compensation scheme based on a first constraint condition to obtain a disturbance simulation evaluation value, wherein the first constraint condition is a grid stability constraint; According to the disturbance simulation evaluation value, it is judged whether the initial compensation scheme is an effective disturbance, and several effective disturbance schemes are obtained; The Pareto front analysis method is used to screen the effective perturbation schemes and obtain several compensation schemes.

4. The topology error prevention risk management method according to claim 3 is characterized in that: The node feature vectors of the measurement missing nodes are repaired using the true filling scheme to obtain a node repair feature vector set, specifically: The linear regression model is constructed by taking the feature vectors of the adjacent nodes of the missing node as the basic value and the preset filling weight as the coefficient. Based on the linear regression model, the weighted least squares method is used to repair the feature vectors of the measurement missing nodes and obtain the node repair feature vector set.

5. The topology error prevention risk management method according to claim 4 is characterized in that: The similarity analysis is performed with the preset risk scenarios respectively, specifically: Extract risk scenario vectors from the historical failure database and construct a risk propagation path tree; Based on the risk propagation path tree, the dynamic time warping algorithm is used to align the node repair feature vector with the risk scenario vector to obtain several first alignment paths; Constructing a second constraint condition based on the preset electrical constraint condition and topological path constraint condition, and filtering the first alignment path to obtain a filtered valid path set; The minimum path distance is calculated based on the screened valid path set and normalized according to the preset path weight to obtain the similarity between each residual compensation scheme and the risk scenario, and a similarity distribution table of each residual compensation scheme is constructed.

6. The topology error prevention risk management method according to claim 5, characterized in that: The optimal similarity threshold corresponding to the truth-filling solution is calculated based on the similarity analysis result, specifically: Based on the similarity distribution table of each truth-filling scheme, a clustering algorithm is used to initially divide the similarity and obtain several similarity intervals; Calculate the ratio of the similarity mean to the standard deviation for each similarity interval and construct a confidence interval; Based on the confidence interval, the preset first threshold of each similarity interval is screened multiple times, and the particle swarm optimization algorithm is used to iteratively optimize the multiple screening results to obtain the first similarity threshold of each similarity interval; The first similarity threshold of each similarity interval is weighted averaged to obtain the optimal similarity threshold corresponding to the truth filling solution.

7. The topology error prevention risk management method according to claim 6, characterized in that: The particle swarm optimization algorithm is used to iteratively optimize the multiple screening results to obtain the first similarity threshold of each similarity interval, which is specifically: Filtering the preset first threshold of each similarity interval based on the confidence interval to obtain a second threshold of each similarity interval; Taking the second threshold of each similarity interval as the center, randomly initialize several threshold combinations to be optimized according to the preset range; The threshold combination to be optimized is taken as a particle, and the initial particle swarm is constructed. The initial particle swarm is searched and iterated and the fitness of each particle is calculated. When the iteration reaches a preset termination condition, the iteration is stopped and the first similarity threshold of each similarity interval is output.

8. The topology error prevention risk management method according to claim 7, characterized in that: The construction of the optimal pairing group and its application in actual error-proof risk identification are specifically as follows: Constructing an optimal pairing group based on an optimal similarity threshold corresponding to a correction scheme, wherein the optimal pairing group includes correction parameters, node repair feature vectors, and landscape scenes; Push the compensation parameters to the power grid monitoring system in real time to trigger automatic repair instructions; Based on the node repair feature vector, the preset risk prediction model is used to dynamically evaluate the power grid status; For each risk scenario, the emergency plan is dynamically activated and a risk disposal report is generated for archiving.

9. A topology error prevention risk management and control system, applied to a topology error prevention risk management and control method according to any one of claims 1 to 8, characterized in that: It includes data acquisition module, parameter generation module, feature repair module and risk management module. Specifically: The data acquisition module is used to obtain the power grid topology diagram and use the graph algorithm to encode the power grid nodes in the topology diagram to obtain the node feature vector; A parameter generation module is used to construct a correction parameter group based on the measurement missing node identification result and the node level, and randomly perturb the correction parameter group based on a preset first constraint condition to obtain multiple correction schemes; The feature repair module is used to repair the node feature vectors of the measurement missing nodes using the filling scheme, obtain the node repair feature vector set, and perform similarity analysis with the preset risk scenarios respectively; The risk control module is used to calculate the optimal similarity threshold corresponding to the truth-filling solution based on the similarity analysis results, construct the optimal pairing group and apply it to actual error-proof risk identification.

Images