Circuit breaker anti-tripping function test method and system

Abstract

The invention discloses a circuit breaker anti-tripping function test method and system, and belongs to the field of power system protection and control, and the test method specifically comprises the following steps: I, extracting circuit breaker data, and constructing an anti-tripping function map through a relation analysis technology; according to the method, the model optimization efficiency is improved, the local optimization problem is avoided, the model can make more accurate response to the state change of the circuit breaker under the actual operation condition, the flexibility and instantaneity of a test system are improved, the sample representativeness is optimized, the anti-tripping test effect is more reliable, the data processing cost is reduced, and the data processing time is shortened. According to the method, errors caused by human intervention are reduced, the intelligent level of a test process is improved, hysteresis of static assessment can be avoided, misjudgment caused by a single factor is reduced, the accuracy of health assessment is improved, the risk management and control capability of the system is enhanced, the possibility of fault occurrence is reduced, and the overall stability of the power system is improved.

Description

Technical Field

[0001] The present invention relates to the field of power system protection and control, and in particular to a circuit breaker anti-tripping function testing method and system. Background Art

[0002] In modern power systems, the reliable operation of power systems depends on the performance and stability of core equipment, among which circuit breakers are key protection and control equipment, and their functions are directly related to the safe and stable operation of power systems. The anti-tripping function of circuit breakers is to prevent repeated tripping caused by misoperation or faults, avoid unnecessary impact on equipment and systems, and play an important role in improving system stability and avoiding fault damage. However, traditional anti-tripping function test methods often rely on a single static test or empirical judgment, which makes it difficult to comprehensively and effectively evaluate the actual working status and health level of circuit breakers. In addition, the existing anti-tripping function test methods have problems such as much manual intervention, low intelligence, and insufficient test accuracy, which can no longer meet the needs of modern power systems for equipment performance evaluation. With the rapid development of artificial intelligence, deep learning, and quantum computing technologies, intelligent circuit breaker anti-tripping test methods have gradually become a research hotspot.

[0003] After searching, Chinese patent number CN113589156A discloses a circuit breaker anti-tripping function test method. Although the invention obtains the relevant parameters of the anti-tripping test of the tested circuit breaker through self-learning, and completes the open-position anti-tripping and closed-position anti-tripping tests of the circuit breaker according to the relevant parameters, effectively improving the efficiency and accuracy of the circuit breaker anti-tripping test, it is easy to fall into the local optimal problem and cannot make a more accurate response to the state change of the circuit breaker under actual operating conditions, reducing the flexibility and immediacy of the test system; and the existing circuit breaker anti-tripping function test method and system sample representativeness is poor, and the reliability of the anti-tripping test effect is low, which increases the cost and time of data processing, and the error caused by human intervention is high; in addition, the existing circuit breaker anti-tripping function test method and system static evaluation have high hysteresis, and there are many misjudgments caused by a single factor, which reduces the accuracy of health assessment, is not conducive to the risk control of the system, and the possibility of failure is high; for this reason, we propose a circuit breaker anti-tripping function test method and system. Summary of the invention

[0004] The purpose of the present invention is to solve the defects in the prior art and to propose a circuit breaker anti-tripping function testing method and system.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions: A circuit breaker anti-tripping function testing method, the specific steps of the testing method are as follows: Ⅰ. Extract circuit breaker data and construct anti-trip function map through relational analysis technology; II. Establish an anti-jump test model and evolve and select model parameters; III. Iterate the simulated test sample and then perform the circuit breaker anti-tripping function test; IV. Collect and record the operation data and status information of each test scenario; V. Compare test results with model predictions and mark problematic operating states; VI. Analyze test results and dynamically adjust the health assessment of circuit breakers; VII. Update the anti-jump function map and anti-jump test model based on the test and analysis results.

[0006] As a further solution of the present invention, the specific steps of constructing the anti-jump function map in step I are as follows: S1.1: Extract the historical operation data, fault data and maintenance data of the circuit breaker from the power system operation database, and automatically filter each group of extracted data according to the set rules, or manually filter the information irrelevant to the circuit breaker anti-trip function by the staff, and only retain the timestamp, fault type, equipment health status and operation number data; S1.2: Construct an operating data set based on the screened data sets, remove duplicate data and missing values ​​in the operating data set, standardize the record format, and then calculate the correlation coefficient between the data sets in the preprocessed operating data set using the Pearson correlation coefficient method; S1.3: Perform a preliminary analysis on the data in the operation data set, identify the attributes of each set of data, and classify the data into categories such as equipment status, operation parameters, environmental factors, fault records, and maintenance records according to the data content and attributes. Then define basic attributes for each type of node, and create specific nodes based on the extracted data; S1.4: Generate an ID for each node using a unique ID generation rule, and correspond each node's identifier to its specific attribute information. Select data that matches the node category and attribute definition from the running data set, fill the matching data into the node's attribute field, check whether each node contains complete attribute information, and supplement or recreate nodes that are missing attributes. S1.5: Add the node to the overall structure of the anti-trip function map. According to the calculation result of the correlation coefficient, determine whether there is a significant correlation between the two nodes, establish the edge between the nodes in the anti-trip function map, and assign the weight and direction of the relationship according to the strength of the correlation coefficient. Based on the operating status and historical maintenance records, calculate the health score of each circuit breaker as a status indicator, and mark the equipment health status of each circuit breaker in the anti-trip function map.

[0007] As a further solution of the present invention, the specific calculation formula of the correlation coefficient described in S1.2 is as follows: ; In the formula, represents the correlation coefficient between the two groups of circuit breaker variables x and y; represents the sample size; Represents the variable x in Observations on samples; represents the mean value of variable x; Represents the variable y in Observations on samples; represents the mean value of variable y; The specific calculation formula for the health score described in S1.5 is as follows: ; In the formula, represents the health score of the hth circuit breaker; Representative Impact factors; Representative The weight of each factor; Represents the number of factors.

[0008] As a further solution of the present invention, the specific steps of establishing the anti-jump test model in step II are as follows: S2.1: Filter out historical anti-jump function test data from the anti-jump function map, and convert each group of extracted data into a time series format. Then, normalize the converted time series data, and divide each group of processed data into a training set, a validation set, and a test set. Based on the Bi-GRU architecture of the input layer, the bidirectional GRU layer, and the output layer, an anti-jump test model is constructed. S2.2: The training set data is input into the anti-jump test model. The input layer transmits each set of data to the bidirectional GRU layer. Then, the forward GRU unit in the bidirectional GRU layer gradually transfers the hidden state from the starting point to the end point of the time series, and the backward GRU unit transfers the hidden state from the end point to the starting point of the time series. S2.3: Calculate candidate hidden states through the reset gates of the forward GRU unit and the backward GRU unit. Based on the calculated candidate hidden states, calculate the hidden state at the current moment through the update gates of the two groups of GRU units, and concatenate the forward and backward hidden states to form a hidden state vector containing bidirectional information. S2.4: Stack multiple bidirectional GRU layers, use the output hidden state of each bidirectional GRU layer as the input of the next layer, and when input to the last layer, concatenate the forward and backward hidden states of the last bidirectional GRU layer, and then post-process the output result through the output layer to generate the final prediction result, and then calculate the error between the prediction result and the true result through the MSE loss function; S2.5: The error value is passed back to the anti-jump test model, and the gradient value of the error value for each layer of the anti-jump test model is calculated. Based on the gradient value of each layer, the parameters of the model are adjusted through the Adam optimization algorithm. The trained anti-jump test model is evaluated using the validation set, and the loss on the validation set is calculated. According to the performance of the validation set, the learning rate of the model and the number of GRU units are adjusted. S2.6: Repeat the training and verification of the anti-trip test model until the preset number of iterations is reached, then evaluate the prediction performance of the final model on the test set, calculate the loss value on the test set, and evaluate the prediction accuracy. Adjust the model parameters again based on the model accuracy, apply the final anti-trip test model to real-time data, and predict the performance of the circuit breaker's anti-trip function.

[0009] As a further solution of the present invention, the specific steps of the evolutionary screening model parameters in step II are as follows: S3.1: Collect the accuracy of the anti-jump test model on the test set. If the accuracy of the model does not reach the preset accuracy, collect the parameter information of the anti-jump test model, and represent each parameter of the model with one or more quantum bits and use quantum states. Represents each parameter; S3.2: After Hadamard transformation of each group of quantum bits, a uniform superposition state of each quantum state is generated to obtain candidate solutions with different parameter combinations. , where N represents the total number of candidate parameter combinations, represents the probability amplitude, Representative A quantum fitness function is defined based on the prediction accuracy of the anti-jump function model, and the fitness of each candidate parameter value is evaluated by the quantum fitness function; S3.3: Construct the objective function. When the fitness of the candidate parameter value meets the preset conditions, the sign of the solution is flipped through the Oracle construction, and then the average value of the probability amplitude of the current quantum state is calculated, and the Grover diffusion operation is performed. Reflect the probability amplitudes of all solutions relative to the mean, where D represents the Grover diffusion operator, represents the projection operator of the current quantum state, and I represents the identity matrix; S3.4: Repeat the application of Oracle construction and Grover diffusion operations, and after each iteration, amplify the probability amplitude of the target solution and compress the probability amplitude of the solution that does not meet the conditions. Then, according to the distance between the target solution and the non-target solution, select Door or The rotation angle of the gate amplifies the probability amplitude of the solution that meets the preset fitness threshold, measures the quantum state, repeats the measurement steps, and retains the measurement results whose fitness meets the preset threshold after multiple measurements; S3.5: Sort the measured parameter combinations by fitness, select the set of parameter values ​​with the highest fitness as the final solution, and determine the optimal solution as the quantum state represented by this parameter combination as the optimal parameter combination of the model, and then fine-tune the parameters obtained by quantum evolution through the gradient descent algorithm.

[0010] A circuit breaker anti-trip function test system, comprising an acquisition preprocessing module, a data screening module, a graph construction module, a learning training module, an evolution screening module, a sample selection module, a real-time monitoring module, a dynamic evaluation module, an alarm prompt module, a management log module and a visual display module; The acquisition preprocessing module is used to extract the data of each circuit breaker from the power system operation database, and clean, convert the format and fill in missing values ​​for the collected data; The data screening module is used to screen out key data corresponding to the circuit breaker tripping event from the pre-processed circuit breaker data according to the anti-tripping function requirements of the circuit breaker; The graph construction module is used to analyze the correlation between various operating data of the circuit breaker, and to construct an anti-trip function graph based on the results of data screening and correlation analysis; The learning and training module is used to build an anti-tripping test model based on the historical anti-tripping function test data of the circuit breaker, and to train and adjust the model; The evolutionary screening module is used to perform evolutionary screening on the parameters of the anti-jump test model, select the best parameter combination, and thus optimize the initial parameter settings of the model; The sample selection module is used to iteratively simulate test samples and select the most representative and influential samples; The real-time monitoring module is used to receive the real-time operation data of the circuit breaker and determine whether the circuit breaker has a potential fault or abnormal tripping risk through an anti-tripping test model; The dynamic evaluation module is used to dynamically estimate the health status of the circuit breaker and the stability of the anti-trip function by combining historical data and current test results; The alarm prompt module is used to detect abnormal status and potential risks of system output, and automatically trigger an alarm according to the threshold set by the system; The management log module is responsible for the daily management and status recording of system operation; The visualization module is used to summarize, analyze and visualize the test and analysis results.

[0011] As a further solution of the present invention, the sample selection module iteratively simulates the test samples in the following specific steps: S4.1: Define the objective function of sample selection based on the relationship between nodes in the anti-jump function map and the characteristics of the anti-jump test model, and based on the effect of samples on improving the accuracy or prediction ability of the model. ,in Representative samples The fitness value of Representative samples Contribution to model training accuracy, Representative samples The distance to other samples in the anti-jump function map, and Represent adjustment parameters respectively; S4.2: Collect and process the circuit breaker operation data under different working conditions, and use the processed circuit breaker operation data under different conditions as a set of test samples to construct a set of test sample sets, calculate the fitness of all test samples using the objective function, and based on a preset fitness threshold, use the test samples above the threshold as the "main river" and use the test samples below the threshold as the "tributary"; S4.3: Based on the fitness of the main river samples and tributary samples, the distance between the main river samples and the tributary samples is calculated, and the characteristic values ​​of the tributary samples are continuously adjusted by simulating the linear update of the sample characteristics, so that the tributary samples are constantly close to the main river samples to form new tributary samples; S4.4: Set the fitness threshold based on the average or median fitness of the current tributary samples , calculate the fitness value of each tributary sample through the objective function , and compare the calculated fitness values ​​respectively With threshold The size of , the corresponding samples will be eliminated and will no longer participate in the subsequent optimization process. Represents the fitness value of the ath sample ; S4.5: Select a coefficient greater than 1 as the precipitation factor by random selection. , and then for The test samples were , and the precipitation factor is calculated for each group of high-adaptability test samples that meet the conditions. The amplification operation is performed to increase the weight of high fitness samples in each test sample, where Representative test samples after precipitation The new eigenvalue of ; S4.6: After each round of iteration, the fitness of all samples is recalculated according to the objective function, and the main river and tributary samples are updated. Then, steps S4.3 to S4.5 are repeated until the maximum number of iterations is reached, or the increment of the fitness of the overall test samples is less than the set convergence threshold, or the screened test samples meet the preset quality requirements. Then, the iterative update is stopped, and the test sample set with the highest fitness and diversity is screened out as the optimal training sample set, and it is input into the anti-trip test model to test the circuit breaker anti-trip function.

[0012] As a further solution of the present invention, the specific steps of the dynamic evaluation module dynamically estimating the health status of the circuit breaker and the stability of the anti-trip function are as follows: S5.1: According to the anti-trip function map and the operation history data of the circuit breaker, a Bayesian network structure is constructed, and the key state of the circuit breaker is identified based on the anti-trip function map. At the same time, each key state of the identified circuit breaker is used as a node of the Bayesian network, and the dependency relationship between the states is marked in the form of directed edges to connect the corresponding two groups of nodes; S5.2: And use the historical data in the anti-jump function map to calculate the probability distribution of each node under different state combinations ,in Represents the probability of node A taking a certain state when its parent node B is given. Representing the initial condition probability table values ​​generated by the historical data recorded in the anti-jump function map, and setting the initial condition probability table for each node of the Bayesian network based on the calculated probability distribution; S5.3: Input the real-time data obtained in the anti-trip function test into the corresponding node of the Bayesian network as the current observation value, and calculate the posterior probability of the circuit breaker being healthy in the current state through the Bayesian inference algorithm, update the probability distribution of the target node according to the calculated health probability, and update the conditional probability table according to the new test data; S5.4: The calculated health status probability distribution is used as the final output result. If the posterior probability of the existing state exceeds the preset risk threshold, the alarm prompt module triggers an alarm. At the same time, the risk type is displayed through the visualization module, and maintenance personnel are advised to conduct further inspections or maintenance to ensure equipment safety. New data is continuously collected and input into the Bayesian network, and the health status of each circuit breaker and the stability of the anti-trip function are continuously monitored.

[0013] As a further solution of the present invention, the key states of the circuit breaker described in S5.1 specifically include states of current, voltage, temperature and number of tripping times.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. The circuit breaker anti-jump function test method collects the accuracy of the anti-jump test model on the test set. If the accuracy of the model does not reach the preset accuracy rate, the parameter information of the anti-jump test model is collected, and each parameter of the model is represented by one or more quantum bits, and each parameter is represented by a quantum state. After each group of quantum bits is transformed by Hadamard, a uniform superposition state of each quantum state is generated to obtain candidate solutions of different parameter combinations. A quantum fitness function is defined based on the prediction accuracy of the anti-jump function model, and the fitness of each candidate parameter value is evaluated by the quantum fitness function. The objective function is constructed. When the fitness of the candidate parameter value meets the preset condition, the sign of the solution is flipped through the Oracle construction, and then the average value of each probability amplitude of the current quantum state is calculated. The probability amplitude of all solutions is reflected relative to the mean through the Grover diffusion operation, and the objective function is repeatedly applied. Racle construction and Grover diffusion operation are performed, and after each iteration, the probability amplitude of the target solution is amplified, and the probability amplitude of the solution that does not meet the conditions is compressed. Then, according to the distance between the target solution and the non-target solution, the required rotation angle is selected to amplify the probability amplitude of the solution that meets the preset fitness threshold, the quantum state is measured, the measurement steps are repeated, and the measurement results whose fitness meets the preset threshold are retained after multiple measurements, the measured parameter combinations are sorted according to fitness, and a group of parameter values ​​with the highest fitness are selected as the final solution. The optimal solution is determined to be the quantum state represented by the parameter combination as the optimal parameter combination of the model, and the parameters obtained by quantum evolution are fine-tuned through the gradient descent algorithm to improve the model optimization efficiency and avoid local optimal problems, so that the model can respond more accurately to the state changes of the circuit breaker under actual operating conditions, thereby increasing the flexibility and immediacy of the test system.

[0015] 2. The circuit breaker anti-tripping function test system defines the objective function of sample selection based on the correlation between nodes in the anti-tripping function map and the characteristics of the anti-tripping test model, and based on the effect of samples on improving the model accuracy or prediction ability, collects and processes the circuit breaker operation data under different working conditions, and uses the processed circuit breaker operation data under different conditions as a group of test samples to construct a set of test sample sets. The objective function is used to calculate the fitness of all test samples, and based on the preset fitness threshold, the test samples are divided into main river samples and tributary samples, and the linear update of sample characteristics is simulated. The characteristic values ​​of tributary samples are continuously adjusted to make the tributary samples closer to the main river samples to form new tributary samples. The fitness threshold is set based on the average or median of the current tributary sample fitness. The fitness threshold is calculated, and the calculated fitness values ​​of each test sample are compared with the set fitness threshold, the test samples below the fitness threshold are eliminated, and a precipitation factor is randomly generated to perform weighted adjustment on the test samples above the fitness threshold, and the iteration update is repeated until the maximum number of iterations is reached, or the increment of the fitness of the overall test samples is less than the set convergence threshold, or the screened test samples meet the preset quality requirements, then the iteration update is stopped, and the test sample set with the highest fitness and diversity is screened out as the optimal training sample set, and at the same time, it is input into the anti-trip test model to test the circuit breaker anti-trip function, optimize the sample representativeness, make the anti-trip test effect more reliable, reduce the cost and time of data processing, reduce the error caused by human intervention, and improve the intelligence level of the test process.

[0016] 3. The circuit breaker anti-trip function test system constructs a Bayesian network structure according to the anti-trip function map and the operation history data of the circuit breaker, and identifies the key state of the circuit breaker based on the anti-trip function map. At the same time, the key states of the identified circuit breaker are used as nodes of the Bayesian network, and the dependencies between the states are marked in the form of directed edges to connect the corresponding two groups of nodes. The historical data in the anti-trip function map is used to calculate the probability distribution of each node under different state combinations, and the initial condition probability table is set for each node of the Bayesian network based on the calculated probability distribution. The real-time data obtained in the anti-trip function test is input into the corresponding node of the Bayesian network as the current observation value, and the posterior probability of the circuit breaker being healthy in the current state is calculated through the Bayesian inference algorithm. The calculated health probability updates the probability distribution of the target node, and at the same time updates the conditional probability table based on the new test data, and uses the calculated health status probability distribution as the final output result. If the posterior probability of the existence state exceeds the preset risk threshold, the alarm prompt module triggers an alarm, and displays the risk type through the visualization module, and recommends maintenance personnel to conduct further inspections or maintenance to ensure equipment safety. New data is continuously collected and input into the Bayesian network, and the health status of each circuit breaker and the stability of the anti-trip function are continuously evaluated dynamically. This can avoid the lag of static evaluation, reduce misjudgment caused by a single factor, improve the accuracy of health assessment, and help enhance the system's risk management capabilities, reduce the possibility of failures, and improve the overall stability of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0018] Figure 1 A flowchart of a circuit breaker anti-tripping function testing method proposed by the present invention; Figure 2 The present invention provides a system block diagram of a circuit breaker anti-tripping function testing system. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0020] Example 1, reference Figure 1 This embodiment discloses a circuit breaker anti-tripping function testing method, and the specific steps of the testing method are as follows: Extract circuit breaker data and build an anti-trip function map through relational analysis technology.

[0021] Specifically, the historical operation data, fault data and maintenance data of the circuit breaker are extracted from the power system operation database, and each group of extracted data is automatically filtered according to the set rules, or the staff manually filters the information irrelevant to the circuit breaker anti-tripping function, and only retains the timestamp, fault type, equipment health status and operation number data. Based on the screened data groups, an operation data set is constructed, duplicate data and missing values ​​in the operation data set are removed, and the record format is standardized. Then, the correlation coefficient between each group of data in the preprocessed operation data set is calculated by the Pearson correlation coefficient method, and the data in the operation data set is preliminarily analyzed to identify the attributes of each group of data. According to the data content and attributes, the data is divided into categories such as equipment status, operating parameters, environmental factors, fault records and maintenance records, and basic attributes are defined for each type of node. Create specific nodes based on the extracted data, use a unique ID generation rule to generate an ID for each node, and correspond the identifier of each node to its specific attribute information, select data that matches the node category and attribute definition from the operating data set, fill the matching data into the attribute field of the node, check whether each node contains complete attribute information, and supplement or recreate the nodes that lack attributes, add the nodes to the overall structure of the anti-trip function map, and determine whether there is a significant association between the two nodes based on the calculation results of the correlation coefficient, establish edges between the nodes in the anti-trip function map, and assign weights and directionality to the relationships based on the strength of the correlation coefficient. Based on the operating status and historical maintenance records, calculate the health score of each circuit breaker as a status indicator, and mark the equipment health status of each circuit breaker in the anti-trip function map.

[0022] It should be further explained that the specific calculation formula of the correlation coefficient is as follows: ; In the formula, represents the correlation coefficient between the two groups of circuit breaker variables x and y; represents the sample size; Represents the variable x in Observations on samples; represents the mean value of variable x; Represents the variable y in Observations on samples; represents the mean value of variable y; The specific calculation formula for the health score is as follows: ; In the formula, represents the health score of the hth circuit breaker; Representative Impact factors; Representative The weight of each factor; Represents the number of factors.

[0023] Establish an anti-jump test model and evolutionarily screen model parameters.

[0024] Specifically, historical anti-jump function test data is screened out from the anti-jump function map, and each group of extracted data is converted into a time series format, and then the converted time series data is normalized, and each group of processed data is divided into a training set, a validation set and a test set, and an anti-jump test model is constructed based on the Bi-GRU architecture of the input layer, the bidirectional GRU layer and the output layer, and the training set data is input into the anti-jump test model, and the input layer transmits each group of data to the bidirectional GRU layer, and then the forward GRU unit in the bidirectional GRU layer gradually transfers the hidden state from the starting point to the end point of the time series in sequence, and at the same time, the backward GRU unit transfers the hidden state from the end point to the starting point of the time series in sequence, and the candidate hidden state is calculated by the reset gate of the forward GRU unit and the backward GRU unit, and the hidden state at the current moment is calculated according to the calculated candidate hidden state through the update gate of the two groups of GRU units, and the forward and backward hidden states are spliced ​​to form a hidden state vector containing bidirectional information, and multiple layers of bidirectional GRU layers are stacked to pass each layer of bidirectional information. The output hidden state of the GRU layer is used as the input of the next layer. When it is input to the last layer, the forward and backward hidden states of the last bidirectional GRU layer are concatenated, and the output result is post-processed through the output layer to generate the final prediction result. The error between the prediction result and the true result is calculated through the MSE loss function, and the error value is passed back to the anti-jump test model, and the error value is calculated for the gradient value of each layer of the anti-jump test model. Based on the gradient value of each layer, the parameters of the model are adjusted through the Adam optimization algorithm, and the trained anti-jump test model is evaluated using the validation set, the loss on the validation set is calculated, and according to the performance of the validation set, the learning rate of the model and the number of GRU units are adjusted. The anti-jump test model is repeatedly trained and verified until the preset number of iterations is reached, and then the prediction performance of the final model is evaluated on the test set, the loss value on the test set is calculated, and the prediction accuracy is evaluated. The model parameters are adjusted again based on the accuracy of the model, and the final anti-jump test model is applied to real-time data to predict the anti-jump function performance of the circuit breaker.

[0025] Specifically, the accuracy of the anti-jump test model on the test set is collected. If the accuracy of the model does not reach the preset accuracy rate, the parameter information of the anti-jump test model is collected, and each parameter of the model is represented by one or more quantum bits and represented by quantum states. Represents each parameter; After Hadamard transformation of each group of quantum bits, a uniform superposition state of each quantum state is generated to obtain candidate solutions with different parameter combinations. , where N represents the total number of candidate parameter combinations, represents the probability amplitude, Representative A quantum fitness function is defined based on the prediction accuracy of the anti-jump function model. The fitness of each candidate parameter value is evaluated by the quantum fitness function, and the objective function is constructed. When the fitness of the candidate parameter value meets the preset conditions, the sign of the solution is flipped through the Oracle construction, and then the average value of each probability amplitude of the current quantum state is calculated. The Grover diffusion operation is used to calculate the average value of the probability amplitude of the current quantum state. Reflect the probability amplitudes of all solutions relative to the mean, where D represents the Grover diffusion operator, Represents the projection operator of the current quantum state, I represents the identity matrix, and repeatedly applies Oracle construction and Grover diffusion operations. After each iteration, the probability amplitude of the target solution is amplified, while the probability amplitude of the solution that does not meet the conditions is compressed. Then, according to the distance between the target solution and the non-target solution, the target solution is selected. Door or The rotation angle of the gate amplifies the probability amplitude of the solution that meets the preset fitness threshold, measures the quantum state, repeats the measurement steps, and retains the measurement results whose fitness meets the preset threshold after multiple measurements. The measured parameter combinations are sorted by fitness, and the set of parameter values ​​with the highest fitness is selected as the final solution. The optimal solution is determined to be the quantum state represented by the parameter combination as the optimal parameter combination of the model, and then the parameters obtained by quantum evolution are fine-tuned through the gradient descent algorithm.

[0026] Iterate the simulated test sample and then perform the circuit breaker anti-trip function test.

[0027] Collect and record operational data and status information for each test scenario.

[0028] Compare test results to model predictions and flag problematic operating states.

[0029] Analyze test results and dynamically adjust circuit breaker health assessments.

[0030] Update the anti-jump function map and anti-jump test model based on the test and analysis results.

[0031] Example 2, reference Figure 2 The present embodiment discloses a circuit breaker anti-trip function test system, including an acquisition preprocessing module, a data screening module, a graph construction module, a learning training module, an evolutionary screening module, a sample selection module, a real-time monitoring module, a dynamic evaluation module, an alarm prompt module, a management log module and a visual display module.

[0032] The acquisition preprocessing module is used to extract the data of each circuit breaker from the power system operation database, and to clean, convert the format and fill in missing values ​​for the collected data; the data screening module is used to filter out the key data corresponding to the circuit breaker tripping event from the preprocessed circuit breaker data according to the anti-tripping function requirements of the circuit breaker.

[0033] The graph construction module is used to analyze the correlation between the various operating data of the circuit breaker, and to construct an anti-trip function graph based on the results of data screening and correlation analysis; the learning and training module is used to construct an anti-trip test model based on the historical anti-trip function test data of the circuit breaker, and to train and adjust the model; the evolutionary screening module is used to perform evolutionary screening on the parameters of the anti-trip test model, select the best parameter combination, and thus optimize the initial parameter settings of the model; the sample selection module is used to iteratively simulate test samples and select the most representative and influential samples.

[0034] Specifically, according to the correlation between nodes in the anti-jump function map and the anti-jump test model characteristics, and based on the effect of samples on improving the model accuracy or prediction ability, the objective function of sample selection is defined. ,in Representative samples The fitness value of Representative samples Contribution to model training accuracy, Representative samples The distance to other samples in the anti-jump function map, and They represent adjustment parameters respectively, collect and process circuit breaker operation data under different working conditions, and use the processed circuit breaker operation data under different conditions as a group of test samples to construct a group of test sample sets. The objective function is used to calculate the fitness of all test samples, and based on the preset fitness threshold, the test samples above the threshold are regarded as "main rivers", and the test samples below the threshold are regarded as "tributaries". Based on the fitness of the main river samples and tributary samples, the distance between the main river samples and the tributary samples is calculated. By simulating the linear update of sample characteristics, the characteristic values ​​of the tributary samples are continuously adjusted to make the tributary samples and the main river samples closer and closer to form new tributary samples. The fitness threshold is set based on the average or median of the current tributary sample fitness. , calculate the fitness value of each tributary sample through the objective function , and compare the calculated fitness values ​​respectively With threshold The size of , the corresponding samples will be eliminated and will no longer participate in the subsequent optimization process. Represents the fitness value of the ath sample , select a coefficient greater than 1 as the precipitation factor by random selection , and then for The test samples were , and the precipitation factor is calculated for each group of high-adaptability test samples that meet the conditions. The amplification operation is performed to increase the weight of high fitness samples in each test sample, where Representative test samples after precipitation After each round of iteration, the fitness of all samples is recalculated according to the objective function, and the main river and tributary samples are updated. Then, steps S4.3 to S4.5 are repeated until the maximum number of iterations is reached, or the increment of the fitness of the overall test samples is less than the set convergence threshold, or the screened test samples meet the preset quality requirements. Then, the iterative update is stopped, and the test sample set with the highest fitness and diversity is screened out as the optimal training sample set, and it is input into the anti-trip test model for circuit breaker anti-trip function test.

[0035] The real-time monitoring module is used to receive the real-time operation data of the circuit breaker and determine whether the circuit breaker has potential faults or abnormal tripping risks through the anti-tripping test model; the dynamic evaluation module is used to dynamically estimate the health status of the circuit breaker and the stability of the anti-tripping function by combining historical data and current test results.

[0036] Specifically, according to the anti-trip function map and the operation history data of the circuit breaker, a Bayesian network structure is constructed, and the key state of the circuit breaker is identified based on the anti-trip function map. At the same time, each key state of the identified circuit breaker is used as a node of the Bayesian network, and the dependency relationship between the states is marked in the form of directed edges to connect the corresponding two groups of nodes. The historical data in the anti-trip function map is used to calculate the probability distribution of each node under different state combinations. ,in The initial condition probability table value is generated by the historical data recorded in the anti-trip function map, and the initial condition probability table is set for each node of the Bayesian network based on the calculated probability distribution. The real-time data obtained in the anti-trip function test is input into the corresponding node of the Bayesian network as the current observation value, and the posterior probability of the circuit breaker being healthy in the current state is calculated through the Bayesian inference algorithm. The probability distribution of the target node is updated according to the calculated health probability, and the conditional probability table is updated according to the new test data. The calculated health state probability distribution is used as the final output result. If the posterior probability of the existence state exceeds the preset risk threshold, the alarm prompt module triggers an alarm, and the risk type is displayed through the visualization module, and the maintenance personnel are advised to conduct further inspection or maintenance to ensure the safety of the equipment, continuously collect new data and input it into the Bayesian network, and continuously perform dynamic circuit breaker health status and anti-trip function stability on each circuit breaker.

[0037] The alarm prompt module is used to detect abnormal status and potential risks of system output, and automatically trigger alarms according to the thresholds set by the system; the management log module is responsible for the daily management and status recording of system operation; the visualization display module is used to summarize, analyze and visualize the results of testing and analysis.

Claims

1. A circuit breaker anti-tripping function testing method, characterized in that: The specific steps of this test method are as follows: Ⅰ. Extract circuit breaker data and construct anti-trip function map through relational analysis technology; II. Establish an anti-jump test model and evolve and select model parameters; III. Iterate the simulated test sample and then perform the circuit breaker anti-tripping function test; IV. Collect and record the operation data and status information of each test scenario; V. Compare test results with model predictions and mark problematic operating states; VI. Analyze test results and dynamically adjust the health assessment of circuit breakers; VII. Update the anti-jump function map and anti-jump test model based on the test and analysis results.

2. A circuit breaker anti-tripping function testing method according to claim 1, characterized in that: The specific steps of constructing the anti-jump function map described in step I are as follows: S1.1: Extract the historical operation data, fault data and maintenance data of the circuit breaker from the power system operation database, and automatically filter each group of extracted data according to the set rules, or manually filter the information irrelevant to the circuit breaker anti-trip function by the staff, and only retain the timestamp, fault type, equipment health status and operation number data; S1.2: Construct an operating data set based on the screened data sets, remove duplicate data and missing values ​​in the operating data set, standardize the record format, and then calculate the correlation coefficient between the data sets in the preprocessed operating data set using the Pearson correlation coefficient method; S1.3: Perform a preliminary analysis on the data in the operation data set, identify the attributes of each set of data, and classify the data into categories such as equipment status, operation parameters, environmental factors, fault records, and maintenance records according to the data content and attributes. Then define basic attributes for each type of node, and create specific nodes based on the extracted data; S1.4: Generate an ID for each node using a unique ID generation rule, and correspond each node's identifier to its specific attribute information. Select data that matches the node category and attribute definition from the running data set, fill the matching data into the node's attribute field, check whether each node contains complete attribute information, and supplement or recreate nodes that are missing attributes. S1.5: Add the node to the overall structure of the anti-trip function map. According to the calculation result of the correlation coefficient, determine whether there is a significant correlation between the two nodes, establish the edge between the nodes in the anti-trip function map, and assign the weight and direction of the relationship according to the strength of the correlation coefficient. Based on the operating status and historical maintenance records, calculate the health score of each circuit breaker as a status indicator, and mark the equipment health status of each circuit breaker in the anti-trip function map.

3. A circuit breaker anti-tripping function testing method according to claim 2, characterized in that: The specific calculation formula of the correlation coefficient described in S1.2 is as follows: ; In the formula, represents the correlation coefficient between the two groups of circuit breaker variables x and y; represents the sample size; Represents the variable x in Observations on samples; represents the mean value of variable x; Represents the variable y in Observations on samples; represents the mean value of variable y; The specific calculation formula for the health score described in S1.5 is as follows: ; In the formula, represents the health score of the hth circuit breaker; Representative Impact factors; Representative The weight of each factor; Represents the number of factors.

4. A circuit breaker anti-tripping function testing method according to claim 2, characterized in that: The specific steps of establishing the anti-jump test model in step II are as follows: S2.1: Filter out historical anti-jump function test data from the anti-jump function map, and convert each group of extracted data into a time series format. Then, normalize the converted time series data, and divide each group of processed data into a training set, a validation set, and a test set. Based on the Bi-GRU architecture of the input layer, the bidirectional GRU layer, and the output layer, an anti-jump test model is constructed. S2.2: The training set data is input into the anti-jump test model. The input layer transmits each set of data to the bidirectional GRU layer. Then, the forward GRU unit in the bidirectional GRU layer gradually transfers the hidden state from the starting point to the end point of the time series, and the backward GRU unit transfers the hidden state from the end point to the starting point of the time series. S2.3: Calculate candidate hidden states through the reset gates of the forward GRU unit and the backward GRU unit. Based on the calculated candidate hidden states, calculate the hidden state at the current moment through the update gates of the two groups of GRU units, and concatenate the forward and backward hidden states to form a hidden state vector containing bidirectional information. S2.4: Stack multiple bidirectional GRU layers, use the output hidden state of each bidirectional GRU layer as the input of the next layer, and when input to the last layer, concatenate the forward and backward hidden states of the last bidirectional GRU layer, and then post-process the output result through the output layer to generate the final prediction result, and then calculate the error between the prediction result and the true result through the MSE loss function; S2.5: The error value is passed back to the anti-jump test model, and the gradient value of the error value for each layer of the anti-jump test model is calculated. Based on the gradient value of each layer, the parameters of the model are adjusted through the Adam optimization algorithm. The trained anti-jump test model is evaluated using the validation set, and the loss on the validation set is calculated. According to the performance of the validation set, the learning rate of the model and the number of GRU units are adjusted. S2.6: Repeat the training and verification of the anti-trip test model until the preset number of iterations is reached, then evaluate the prediction performance of the final model on the test set, calculate the loss value on the test set, and evaluate the prediction accuracy. Adjust the model parameters again based on the model accuracy, apply the final anti-trip test model to real-time data, and predict the performance of the circuit breaker's anti-trip function.

5. A circuit breaker anti-tripping function testing method according to claim 4, characterized in that: The specific steps of the evolutionary screening model parameters in step II are as follows: S3.1: Collect the accuracy of the anti-jump test model on the test set. If the accuracy of the model does not reach the preset accuracy, collect the parameter information of the anti-jump test model, and represent each parameter of the model with one or more quantum bits and use quantum states. Represents each parameter; S3.2: After Hadamard transformation of each group of quantum bits, a uniform superposition state of each quantum state is generated to obtain candidate solutions with different parameter combinations. , where N represents the total number of candidate parameter combinations, represents the probability amplitude, Representative A quantum fitness function is defined based on the prediction accuracy of the anti-jump function model, and the fitness of each candidate parameter value is evaluated by the quantum fitness function; S3.3: Construct the objective function. When the fitness of the candidate parameter value meets the preset conditions, the sign of the solution is flipped through the Oracle construction, and then the average value of the probability amplitude of the current quantum state is calculated, and the Grover diffusion operation is performed. Reflect the probability amplitudes of all solutions relative to the mean, where D represents the Grover diffusion operator, represents the projection operator of the current quantum state, and I represents the identity matrix; S3.4: Repeat the application of Oracle construction and Grover diffusion operations, and after each iteration, amplify the probability amplitude of the target solution and compress the probability amplitude of the solution that does not meet the conditions. Then, according to the distance between the target solution and the non-target solution, select Door or The rotation angle of the gate amplifies the probability amplitude of the solution that meets the preset fitness threshold, measures the quantum state, repeats the measurement steps, and retains the measurement results whose fitness meets the preset threshold after multiple measurements; S3.5: Sort the measured parameter combinations by fitness, select the set of parameter values ​​with the highest fitness as the final solution, and determine the optimal solution as the quantum state represented by this parameter combination as the optimal parameter combination of the model, and then fine-tune the parameters obtained by quantum evolution through the gradient descent algorithm.

6. A circuit breaker anti-tripping function testing system, used to implement the circuit breaker anti-tripping function testing method according to any one of claims 1 to 5, characterized in that: It includes acquisition preprocessing module, data screening module, graph construction module, learning training module, evolution screening module, sample selection module, real-time monitoring module, dynamic evaluation module, alarm prompt module, management log module and visual display module; The acquisition preprocessing module is used to extract the data of each circuit breaker from the power system operation database, and clean, convert the format and fill in missing values ​​for the collected data; The data screening module is used to screen out key data corresponding to the circuit breaker tripping event from the pre-processed circuit breaker data according to the anti-tripping function requirements of the circuit breaker; The graph construction module is used to analyze the correlation between various operating data of the circuit breaker, and to construct an anti-trip function graph based on the results of data screening and correlation analysis; The learning and training module is used to build an anti-tripping test model based on the historical anti-tripping function test data of the circuit breaker, and to train and adjust the model; The evolutionary screening module is used to perform evolutionary screening on the parameters of the anti-jump test model, select the best parameter combination, and thus optimize the initial parameter settings of the model; The sample selection module is used to iteratively simulate test samples and select the most representative and influential samples; The real-time monitoring module is used to receive the real-time operation data of the circuit breaker and determine whether the circuit breaker has a potential fault or abnormal tripping risk through an anti-tripping test model; The dynamic evaluation module is used to dynamically estimate the health status of the circuit breaker and the stability of the anti-trip function by combining historical data and current test results; The alarm prompt module is used to detect abnormal status and potential risks of system output, and automatically trigger an alarm according to the threshold set by the system; The management log module is responsible for the daily management and status recording of system operation; The visualization module is used to summarize, analyze and visualize the test and analysis results.

7. A circuit breaker anti-tripping function test system according to claim 6, characterized in that: The specific steps of iterative simulation test samples of the sample selection module are as follows: S4.1: Define the objective function of sample selection based on the relationship between nodes in the anti-jump function map and the characteristics of the anti-jump test model, and based on the effect of samples on improving the accuracy or prediction ability of the model. ,in Representative samples The fitness value of Representative samples Contribution to model training accuracy, Representative samples The distance to other samples in the anti-jump function map, and Represent adjustment parameters respectively; S4.2: Collect and process the circuit breaker operation data under different working conditions, and use the processed circuit breaker operation data under different conditions as a set of test samples to construct a set of test sample sets, calculate the fitness of all test samples using the objective function, and based on a preset fitness threshold, use the test samples above the threshold as the "main river" and the test samples below the threshold as the "tributary"; S4.3: Based on the fitness of the main river samples and tributary samples, the distance between the main river samples and the tributary samples is calculated, and the characteristic values ​​of the tributary samples are continuously adjusted by simulating the linear update of the sample characteristics, so that the tributary samples are constantly close to the main river samples to form new tributary samples; S4.4: Set the fitness threshold based on the average or median fitness of the current tributary samples , calculate the fitness value of each tributary sample through the objective function , and compare the calculated fitness values ​​respectively With threshold The size of , the corresponding samples will be eliminated and will no longer participate in the subsequent optimization process. Represents the fitness value of the ath sample ; S4.5: Select a coefficient greater than 1 as the precipitation factor by random selection. , and then for The test samples were , and the precipitation factor is calculated for each group of high-adaptability test samples that meet the conditions. The amplification operation is performed to increase the weight of high fitness samples in each test sample, where Representative test samples after precipitation The new eigenvalue of ; S4.6: After each round of iteration, the fitness of all samples is recalculated according to the objective function, and the main river and tributary samples are updated. Then, steps S4.3 to S4.5 are repeated until the maximum number of iterations is reached, or the increment of the fitness of the overall test samples is less than the set convergence threshold, or the screened test samples meet the preset quality requirements. Then, the iterative update is stopped, and the test sample set with the highest fitness and diversity is screened out as the optimal training sample set, and it is input into the anti-trip test model to test the circuit breaker anti-trip function.

8. A circuit breaker anti-tripping function test system according to claim 6, characterized in that: The specific steps of the dynamic evaluation module dynamically estimating the health status of the circuit breaker and the stability of the anti-trip function are as follows: S5.1: According to the anti-trip function map and the operation history data of the circuit breaker, a Bayesian network structure is constructed, and the key state of the circuit breaker is identified based on the anti-trip function map. At the same time, each key state of the identified circuit breaker is used as a node of the Bayesian network, and the dependency relationship between the states is marked in the form of directed edges to connect the corresponding two groups of nodes; S5.2: And use the historical data in the anti-jump function map to calculate the probability distribution of each node under different state combinations ,in Represents the probability of node A taking a certain state when its parent node B is given. Representing the initial condition probability table values ​​generated by the historical data recorded in the anti-jump function map, and setting the initial condition probability table for each node of the Bayesian network based on the calculated probability distribution; S5.3: Input the real-time data obtained in the anti-trip function test into the corresponding node of the Bayesian network as the current observation value, and calculate the posterior probability of the circuit breaker being healthy in the current state through the Bayesian inference algorithm, update the probability distribution of the target node according to the calculated health probability, and update the conditional probability table according to the new test data; S5.4: The calculated health status probability distribution is used as the final output result. If the posterior probability of the existing state exceeds the preset risk threshold, the alarm prompt module triggers an alarm. At the same time, the risk type is displayed through the visualization module, and maintenance personnel are advised to conduct further inspections or maintenance to ensure equipment safety. New data is continuously collected and input into the Bayesian network, and the health status of each circuit breaker and the stability of the anti-trip function are continuously monitored.

Images