Self-adaptive intelligent current limiting and managing method

Abstract

The invention discloses a self-adaptive intelligent current limiting and managing method, and relates to the technical field of current limiting. Comprising the steps of collecting real-time current data, historical current data and a system load state of a node, performing feature extraction on the real-time current data and the historical current data of the node, performing analysis and prediction on the collected current data by using an intelligent algorithm, and predicting current according to a current state analysis result; performing risk assessment on the current state in combination with historical current data of the node and a system load condition, analyzing predicted current data, and judging whether the current state has a short circuit risk or not; according to a short-circuit risk judgment result, determining a short-circuit risk node, selecting a current limit strategy, and adjusting current limit by setting a current limit protection value; and continuously monitoring the current state of the short-circuit risk node, adding the node which frequently causes an abnormal state into a blacklist, and carrying out key monitoring or regular maintenance.

Description

Technical Field

[0001] The invention belongs to the technical field of current limitation, and in particular relates to an adaptive intelligent current limitation and management method. Background Art

[0002] With the continuous expansion of the scale of the power grid and the gradual strengthening of the power grid structure, the single-machine capacity, power plant capacity, substation capacity, load and load density in the system have continued to grow, and the short-circuit current level of each hub site has increased sharply. The problem of excessive short-circuit current at the main site in the load-intensive area has become a bottleneck for the safe operation of the power system. Therefore, power equipment faces higher thermal stability and dynamic stability requirements, and limiting the system short-circuit current has become one of the problems that the current power system urgently needs to solve.

[0003] At present, Jiangsu power grid system analysts still rely on manual experience when taking current limiting measures. The main process is as follows: system analysts collect funds, manually enter data, and adjust the typical operating mode of short-circuit current calculation. After the method is established, they rely on the experience of system analysts to take current limiting measures on nodes.

[0004] The above process relies entirely on manual experience, with heavy workload, low efficiency, strong limitations and poor flexibility, which makes it difficult to meet the future development needs of my country's power grid. The main problems are reflected in the following four aspects:

[0005] At present, when conducting short-circuit current analysis and calculation, it is necessary to export the historical operation data of the power grid and related equipment ledgers in batches each time. The data export, collation and processing are time-consuming. On the other hand, the historical data currently exported still has many problems such as missing data, measurement errors and ledger errors, which greatly restricts the development of power grid operation analysis.

[0006] The generation of operation modes must be technically supported by a large amount of analytical calculations. It is necessary to screen out feasible solutions that meet various constraints from a large number of possible operation plans, and then form a multi-objective optimization plan to make the operation plan operational. As the power grid structure becomes increasingly complex, the amount of calculation of operation modes increases dramatically.

[0007] The uncertainty issues in power grid operation are increasing, and the limitations of traditional operation methods are becoming more and more obvious; moreover, the safety and stability margin of the power grid is gradually decreasing, and the accuracy of adjusting the operation mode is facing challenges.

[0008] The workflow is repetitive. The same process is repeated, but the input conditions vary to varying degrees. The scope and basic requirements involved each year are also the same, but the emphasis is different according to the changes in the development of the power grid. The chapter structure of the annual operation mode report and the special analysis and calculation report is basically fixed.

[0009] Therefore, under the circumstances of heavy computing workload and tight staffing, in order to free system analysts from heavy repetitive and mechanical work, tap their own potential, further improve work efficiency and quality, and better serve and support the dispatching and operation of Jiangsu power grid, an adaptive intelligent current limitation and management method is urgently needed. Summary of the invention

[0010] The object of the present invention is to provide an adaptive intelligent current limiting and management method, which is used to solve the technical problem in the prior art that the current limiting has many variations and is difficult to calculate.

[0011] In order to achieve the above object, the present invention adopts the following technical solutions:

[0012] Step 1: Collect the real-time current data, historical current data and system load status of the nodes, and extract features from the real-time current data and historical current data of the nodes to clarify the power system area, key nodes and current data. Current data refers to the current size, direction, and change rate. Select the appropriate current sensor type according to the monitoring needs, deploy current sensors at key nodes in the power system, monitor the current data in real time, and extract feature types that can reflect the current status from the current data. Feature types include: maximum current value, average current value, degree of discreteness of current values, and current change rate at adjacent time points within a certain period of time. Collect the feature values ​​corresponding to each feature type as the feature value set of the feature type. The node system load status includes: node system equivalent impedance, line impedance, and node system equivalent phase potential.

[0013] Step 2: Use intelligent algorithms to analyze and predict the collected current data, and predict the current based on the current state analysis results; use the random forest algorithm model to analyze and predict the collected current data, use the feature value set as the input sample set of the model, and use the bootstrap method to increase the difference of the decision tree in the random forest, thereby reducing the generalization error of the model, classify and analyze each feature type, and optimize the model output by minimizing the target square error. The specific method is as follows:

[0014] The eigenvalue set containing n elements corresponding to the eigenvalue type j is D j ={x1, x2, ..., x n}, the set is arranged according to the numerical size, increasing from left to right, set the set node v, and divide the set into two subsets through v, defined as d j，1 ={x i |x i ≤v}(i=1,2,…,n1),d j，2 ={x m |x m>v}(m=n1,…,n2), and n1+n2=n, the specific formula for minimizing the target square error is as follows:

[0015]

[0016] Among them, k1 represents the subset d j,1 The average value of the elements contained, k2 represents the subset d j,2 The average value of the elements contained in it is obtained by repeating the above steps to continuously divide and optimize the samples according to the features, integrating multiple trained sub-decision trees into a random forest, and using this model to predict the current data to obtain the predicted value f of feature type j j (x i ), evaluate whether the model training is complete by calculating the average relative error index. The calculation formula of the average relative error index O is as follows:

[0017]

[0018] Set the average relative error threshold. When O is less than or equal to the threshold, the model training is considered complete, and the trained model is used to predict the current.

[0019] Step 3: Combine the historical current data of the node and the system load to conduct a risk assessment on the current state, analyze the predicted current data, and determine whether there is a short-circuit risk in the current state; based on the historical current data, screen the abnormal state data that causes the current short circuit. The abnormal state refers to the abnormal current state that can cause the current short circuit, such as: a sudden increase in current or a continuous high current. Perform correlation mining on the historical current data to find the correlation between the abnormal state and the current data that causes the abnormal state. Perform statistical analysis on the current data before each abnormal state occurs based on the historical current data to obtain the current characteristic value range interval set that causes the abnormal state at time t. The obtained interval is as follows:

[0020]

[0021] where a m represents the maximum current value at time t, a v represents the average current value at time t, a d Indicates the discreteness of the current value at time t, a t The current change rate at adjacent time points, a1, a2, a3, a4 are the corresponding eigenvalue change parameters obtained by mining the association law. The number of times each eigenvalue in the set interval causes a current short circuit is counted to quantify the short circuit correlation corresponding to the current data. The specific formula is as follows:

[0022]

[0023] The time range is set to t, and n time intervals are divided within the time t. The characteristic value is extracted from each time interval, r j represents the short-circuit correlation of feature type j at time t, x i represents the i-th eigenvalue belonging to this feature type, y i represents the number of times the characteristic value xi causes a current short circuit, represents the average value of n eigenvalues ​​at time t, represents the average number of current short circuits caused by n characteristic values ​​at time t, r j The smaller the value, the smaller the correlation between feature type j and short circuit risk.

[0024] According to the short-circuit correlation of each feature type, the corresponding weight is set, and the predicted current data is analyzed to determine whether there is a short-circuit risk. r1 represents the short-circuit risk value of the maximum current, and the corresponding weight is k1. r2 represents the short-circuit risk value of the average current, and the corresponding weight is k2. r3 represents the short-circuit risk value of the discrete degree of the current value, and the corresponding weight is k3. r4 represents the short-circuit risk value of the current change rate at adjacent time points, and the corresponding weight is k4. The corresponding weight is set according to the short-circuit risk value of each feature type. For example: if r1>r2>r3>r4, then k1>k2>k3>k4 is guaranteed, and k1+k2+k3+k4=1; according to the weight, the short-circuit risk value is set to determine the short-circuit risk of the predicted current data. The specific formula for setting the short-circuit risk value L is as follows:

[0025]

[0026] Where j represents the jth feature type, N j Indicates the total number of predicted feature values ​​of the jth feature type, M j Indicates the number of current characteristic value ranges that predict characteristic values ​​and lead to abnormal conditions; sets short-circuit risk value thresholds y1 and y2, and y2>y1. When L is greater than or equal to y2, it is judged that the current current state has a short-circuit risk that needs to be handled first. When L is greater than or equal to y1 and less than y2, it is judged that the current current state has a general short-circuit risk. When L is less than y1, it is judged that no short-circuit risk has been found in the current current state.

[0027] Step 4. According to the short-circuit risk judgment results, determine the short-circuit risk nodes, select the current limiting strategy, and adjust the current limit by setting the current limiting protection value. According to the short-circuit risk judgment results, screen out the grid nodes with short-circuit risks, clarify which equipment, which areas, and which time periods require current limiting measures, and clarify specific limiting measures based on the parameters of the limiting current.

[0028] Select the restrictive measures that meet the actual needs according to their advantages and disadvantages. For example, the measures of string-out and line-out have the advantages of simple engineering implementation, low cost and significant current limiting effect, but will weaken the connection between the system grid, resulting in the risk of over-limit of some equipment and reducing the safety margin of the surrounding power grid; the measures of pulling and stopping the line have the advantages of being simple and easy to implement and not requiring supporting projects, but pulling and stopping the line will weaken the connection between the grid and the lines in production will be pulled and stopped at the same time, which is poor in economy; the measures of installing series impedance have the advantages of increasing the electrical distance of the line and increasing the impedance, which are mostly used for ultra-high voltage and long-distance transmission lines, but normal operation will produce large voltage loss, increase network loss and reduce steady-state performance; the measures of accompanying the suspension of phase regulators have the advantages of large compensation capacity, phase-advancing operation, fast excitation, fast voltage recovery speed and long service life, but they cannot play a beneficial regulating role of phase regulators on the system, and the investment cost of phase regulators is large, so the economic efficiency of accompanying the suspension of phase regulators is poor; the busbar segmentation measure is the most economical, simple and effective means of limiting short-circuit current, but it will weaken the electrical connection of the system. At the segmentation point, the electrical connection of the system becomes relatively weak and easily affected by faults.

[0029] According to the node historical current data and system load conditions, the current limit protection value is set. The calculation formula of the current limit protection value I is as follows:

[0030]

[0031] Within the time t, the degree of current restriction must be greater than the current limit protection value I. H represents the protection coefficient value. The protection coefficient range is obtained by analyzing the historical normal current data. The value of the protection coefficient is adjusted within the protection coefficient range to ensure that the power system is kept running stably while limiting the current. Under the premise of ensuring safety, the load demand of the power system is met as much as possible to avoid unnecessary power outages or power restrictions. Z S is the system equivalent impedance, Z L is the line impedance, E S is the equivalent phase potential of the system.

[0032] The protection coefficient range is obtained by analyzing the historical normal current data, and the time range t when the current value is low and the node works normally is selected. The current average value of the node in n time periods is extracted within this time range to determine the lower limit H of the protection coefficient range. - , the specific calculation formula is as follows:

[0033]

[0034] in, represents the average current value in the i-th time period, Indicates the historical current average value when the node is working normally.

[0035] Similarly, according to the historical normal current data, select the time range t when the current value is high and the node works normally, extract the current average value of the node in n time periods within this time range, and determine the upper limit H of the protection coefficient range + , the specific calculation formula is as follows:

[0036]

[0037] in, Represents the average current value in the i-th time period.

[0038] Clarify the range of the protection factor and optimize its setting value H. Based on the pre-selected current limiting measures and historical current data records, conduct a comprehensive analysis of the actual degree of current suppression of this current limiting measure and its execution cost and effect, find the optimal H value setting point and its corresponding current limiting protection value I, and ensure that under the condition of a given current limiting protection value I, the current limiting measure can achieve the best current control effect at the most economical cost.

[0039] Step 5. Continue to monitor the current status of short-circuit risk nodes. For nodes that frequently cause abnormal conditions, add them to the blacklist for key monitoring or regular maintenance. Use power monitoring software or system to monitor the current status of short-circuit risk nodes in real time, and compare the historical normal current data to determine whether the current limiting strategy is effective. Adjust the invalid current limiting strategy, set a node abnormal state number threshold g, establish a blacklist system, and put nodes that cause abnormal conditions greater than or equal to g within a time T into blacklist management. Include these nodes in the key monitoring scope, strengthen the monitoring and analysis of their current status, increase the density and frequency of monitoring points, regularly evaluate the effectiveness of blacklist management and key monitoring, and adjust the monitoring strategy and maintenance plan based on the evaluation results to ensure the effectiveness of the restriction measures.

[0040] Compared with the prior art, the beneficial effects are:

[0041] 1. The present invention uses a random forest algorithm to analyze and predict the collected current data. Each decision tree in the random forest optimizes its prediction performance by minimizing the target square error. In the current prediction scenario, by continuously splitting nodes, the model learns the complex patterns in the current data, so that the future current state can be predicted more accurately.

[0042] 2. The present invention quantifies the short-circuit correlation corresponding to the current data by determining the range interval set of current characteristic values ​​in which abnormal conditions occur. By quantifying the correlation, we can further understand the influence of different current characteristics on short-circuit events, and provide data support for subsequent weight setting. According to the short-circuit correlation of each characteristic type, the corresponding weight is set, which helps to accurately reflect the contribution of different current characteristics to short-circuit risks and improve the accuracy of short-circuit risk judgment.

[0043] 3. The present invention obtains a protection coefficient range by analyzing historical normal state current data, sets a current limiting protection value based on the protection coefficient value, and ensures stable operation of the power system while limiting the current by adjusting the protection coefficient value within the protection coefficient range. Under the premise of ensuring safety, the load demand of the power system is met as much as possible to avoid unnecessary power outages or power restrictions. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A flow chart of an adaptive intelligent current limiting and management method of the present invention;

[0045] Figure 2 A diagram showing the steps of the method for setting a short circuit risk value according to the present invention. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] Example 1

[0048] like Figure 1 As shown, the adaptive intelligent current limiting and management method specifically includes the following steps:

[0049] Step 1: Collect the real-time current data, historical current data and system load status of the nodes, extract features from the real-time current data and historical current data of the nodes, clarify the power system areas, key nodes and current data that need to be monitored, current data refers to the current size, direction, and change rate, select the appropriate current sensor type according to the monitoring needs, deploy current sensors at key nodes in the power system, monitor the current data in real time, and extract feature types that can reflect the current status from the current data. Feature types include: maximum current value within a certain period of time, average current value, degree of discreteness of current values, and current change rate at adjacent time points, collect the feature values ​​corresponding to each feature type as the feature value set of the feature type, and the node system load status includes: node system equivalent impedance, line impedance, node system equivalent phase potential.

[0050] Step 2: Use intelligent algorithms to analyze and predict the collected current data, and predict the current based on the current state analysis results; use the random forest algorithm model to analyze and predict the collected current data, use the feature value set as the input sample set of the model, and use the bootstrap method to increase the difference of the decision tree in the random forest, thereby reducing the generalization error of the model, classify and analyze each feature type, and optimize the model output by minimizing the target square error. The specific method is as follows:

[0051] The eigenvalue set containing n elements corresponding to the eigenvalue type j is D j ={x1, x2, ..., x n}, the set is arranged according to the numerical size, increasing from left to right, set the set node v, and divide the set into two subsets through v, defined as d j，1 ={x i |x i ≤v}(i=1,2,…,n1),d j，2 ={x m |x m >v}(m=n1,…,n2), and n1+n2=n, the specific formula for minimizing the target square error is as follows:

[0052]

[0053] Among them, k1 represents the subset d j,1 The average value of the elements contained, k2 represents the subset d j,2 The average value of the elements contained in it is obtained by repeating the above steps to continuously divide and optimize the samples according to the features, integrating multiple trained sub-decision trees into a random forest, and using this model to predict the current data to obtain the predicted value f of feature type j j (x i), evaluate whether the model training is complete by calculating the average relative error index. The calculation formula of the average relative error index O is as follows:

[0054]

[0055] Set the average relative error threshold. When O is less than or equal to the threshold, the model training is considered complete, and the trained model is used to predict the current.

[0056] Step 3: Combine the historical current data of the node and the system load to conduct a risk assessment on the current state, analyze the predicted current data, and determine whether there is a short circuit risk in the current current state; Figure 2 As shown, based on the historical current data, the abnormal state data that causes the current short circuit is screened. The abnormal state refers to the abnormal current state that can cause the current short circuit, such as: a sudden increase in current or a continuous high current. The historical current data is mined for correlation, and the correlation law between the abnormal state and the current data that causes the abnormal state is found. According to the historical current data, the current data before each abnormal state occurs is statistically analyzed to obtain the current characteristic value range interval set that causes the abnormal state within time t. The obtained interval is shown as follows:

[0057]

[0058] where a m represents the maximum current value at time t, a v represents the average current value at time t, a d Indicates the discreteness of the current value at time t, a t The current change rate at adjacent time points, a1, a2, a3, a4 are the corresponding eigenvalue change parameters obtained by mining the association law. The number of times each eigenvalue in the set interval causes a current short circuit is counted to quantify the short circuit correlation corresponding to the current data. The specific formula is as follows:

[0059]

[0060] The time range is set to t, and n time intervals are divided within the time t. The characteristic value is extracted from each time interval, r j represents the short-circuit correlation of feature type j at time t, x i represents the i-th eigenvalue belonging to this feature type, y i represents the eigenvalue x i The number of short circuits that cause current, represents the average value of n eigenvalues ​​at time t, represents the average number of current short circuits caused by n characteristic values ​​at time t, r jThe smaller the value, the smaller the correlation between feature type j and short circuit risk.

[0061] According to the short-circuit correlation of each feature type, the corresponding weight is set, and the predicted current data is analyzed to determine whether there is a short-circuit risk. r1 represents the short-circuit risk value of the maximum current, and the corresponding weight is k1. r2 represents the short-circuit risk value of the average current, and the corresponding weight is k2. r3 represents the short-circuit risk value of the discrete degree of the current value, and the corresponding weight is k3. r4 represents the short-circuit risk value of the current change rate at adjacent time points, and the corresponding weight is k4. The corresponding weight is set according to the short-circuit risk value of each feature type. For example: if r1>r2>r3>r4, then k1>k2>k3>k4 is guaranteed, and k1+k2+k3+k4=1; according to the weight, the short-circuit risk value is set to determine the short-circuit risk of the predicted current data. The specific formula for setting the short-circuit risk value L is as follows:

[0062]

[0063] Where j represents the jth feature type, N j Indicates the total number of predicted feature values ​​of the jth feature type, M j Indicates the number of current characteristic value ranges that predict characteristic values ​​and lead to abnormal conditions; sets short-circuit risk value thresholds y1 and y2, and y2>y1. When L is greater than or equal to y2, it is judged that the current current state has a short-circuit risk that needs to be handled first. When L is greater than or equal to y1 and less than y2, it is judged that the current current state has a general short-circuit risk. When L is less than y1, it is judged that no short-circuit risk has been found in the current current state.

[0064] Step 4. According to the short-circuit risk judgment results, determine the short-circuit risk nodes, select the current limiting strategy, and adjust the current limit by setting the current limiting protection value to ensure that the stable operation of the power system is maintained while limiting the current; through the short-circuit risk judgment results, screen out the grid nodes with short-circuit risks, clarify which equipment, which areas, and which time periods require current limiting measures, and clarify specific limiting measures based on the parameters of the limiting current.

[0065] According to the advantages and disadvantages of the restrictive measures, select the restrictive measures that meet the actual needs of the short-circuit risk grid nodes. For example, the string-out and line-out measures have the advantages of simple engineering implementation, low cost, and significant current limiting effect, but they will weaken the system grid connection, resulting in the risk of over-limit of some equipment and reducing the safety margin of the surrounding power grid; the line-pulling and line-stopping measures have the advantages of being simple and easy to implement and not requiring supporting projects, but pulling and stopping the lines will weaken the grid connection, and the lines in production will be pulled and stopped at the same time, which is poor in economy; the installation of series impedance measures has the advantages of increasing the electrical distance of the line and increasing the impedance, and is mostly used for ultra-high voltage and long-distance transmission lines, but normal operation will produce large voltage losses, increase network losses, and reduce steady-state performance; the measures of accompanying the shutdown of phase regulators have the advantages of large compensation capacity, phase-advancing operation, fast excitation, fast voltage recovery speed and long service life, but they cannot play a beneficial regulatory role of phase regulators on the system, and the investment cost of phase regulators is large, so the economic efficiency of accompanying the shutdown of phase regulators is poor; the busbar segmentation measures are the most economical, simple and effective means of limiting short-circuit current, but they will weaken the electrical connection of the system. At the segmentation point, the electrical connection of the system becomes relatively weak and easily affected by faults.

[0066] According to the node historical current data and system load conditions, the current limit protection value is set. The calculation formula of the current limit protection value I is as follows:

[0067]

[0068] Where H represents the protection coefficient value. The protection coefficient range is obtained by analyzing the historical normal current data. The protection coefficient value is adjusted within the protection coefficient range to ensure the stable operation of the power system while limiting the current. Under the premise of ensuring safety, the load demand of the power system is met as much as possible to avoid unnecessary power outages or power restrictions. S is the system equivalent impedance, Z L is the line impedance, E S is the equivalent phase potential of the system.

[0069] The protection coefficient range is obtained by analyzing the historical normal current data, and n time periods with current values ​​lower than the historical current average value when the node is working normally are selected. The current average value of the n time periods is extracted to determine the lower limit H of the protection coefficient range. - , the specific calculation formula is as follows:

[0070]

[0071] in, represents the average current value in the i-th time period, Indicates the historical current average value when the node is working normally.

[0072] Similarly, based on the historical normal current data, select m time periods with current values ​​higher than the historical current average value when the node is working normally, extract the current average value of the m time periods, and determine the upper limit H of the protection coefficient range. + , the specific calculation formula is as follows:

[0073]

[0074] in, Represents the average current value in the i-th time period.

[0075] Clarify the range of the protection factor and optimize its setting value H. Based on the pre-selected current limiting measures and historical current data records, conduct a comprehensive analysis of the actual degree of current suppression of this current limiting measure and its execution cost and effect, find the optimal H value setting point and its corresponding current limiting protection value I, and ensure that under the condition of a given current limiting protection value I, the current limiting measure can achieve the best current control effect at the most economical cost.

[0076] Step 5. Continue to monitor the current status of short-circuit risk nodes to ensure the effectiveness of restriction measures. For nodes that frequently cause abnormal conditions, add them to the blacklist for key monitoring or regular maintenance. Use power monitoring software or system to monitor the current status of short-circuit risk nodes in real time, and compare the historical normal current data to determine whether the current limiting strategy is effective. Adjust the invalid current limiting strategy, set a node abnormal state number threshold g, establish a blacklist system, and put nodes that cause abnormal conditions greater than or equal to g within a time T into blacklist management. Include these nodes in the key monitoring scope, strengthen the monitoring and analysis of their current status, increase the density and frequency of monitoring points, regularly evaluate the effectiveness of blacklist management and key monitoring, and adjust the monitoring strategy and maintenance plan based on the evaluation results to ensure the effectiveness of restriction measures.

[0077] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

[0078] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. An adaptive intelligent current limiting and management method, characterized in that: include: Step 1: Collect the real-time current data, historical current data and system load status of the node, and perform feature extraction on the real-time current data and historical current data of the node; Step 2: Analyze and predict the collected current data using intelligent algorithms, and predict the current based on the current state analysis results; Step 3: Combine the historical current data of the node and the system load to conduct a risk assessment on the current state, analyze the predicted current data, and determine whether there is a short circuit risk in the current current state; Step 4: According to the short-circuit risk judgment result, determine the short-circuit risk node, select the current limiting strategy, and adjust the current limit by setting the current limiting protection value; Step 5: Continue to monitor the current status of short-circuit risk nodes. For nodes that frequently cause abnormal conditions, add them to the blacklist for key monitoring or regular maintenance.

2. The adaptive intelligent current limiting and management method according to claim 1, characterized in that: The system load state in step 1 includes: system equivalent impedance, line impedance, and system equivalent phase potential; Feature extraction is performed on the real-time current data and historical current data of the node, specifically: feature types are extracted from the real-time current data and historical current data, and the feature types include: the maximum current value, the average current value, the discrete degree of the current value, and the current change rate at adjacent time points within time t.

3. The adaptive intelligent current limiting and management method according to claim 1, characterized in that: Step 2 uses an intelligent algorithm to analyze and predict the collected current data, and predicts the current based on the current state analysis results. Specifically, a random forest algorithm model is used to analyze and predict the collected current data, and the eigenvalue set is used as the input sample set of the model. Each feature type is classified and analyzed, and the model output is optimized by minimizing the target square error.

4. The adaptive intelligent current limiting and management method according to claim 3, characterized in that: The specific method to minimize the target square error and optimize the model output is as follows: The eigenvalue set containing n elements corresponding to the eigenvalue type j is D j ={x1, x2, ..., x n }, the set is arranged according to the numerical size, increasing from left to right, set the set node v, and divide the set into two subsets through v, defined as d j，1 ={x i |x i ≤v}(i=1,2,…,n1),d j，2 ={x m |x m >v}(m=n1,…,n2), and n1+n2=n, the specific formula for minimizing the target square error is as follows: Among them, k1 represents the subset d j,1 The average value of the elements contained, k2 represents the subset d j,2 The average value of the elements contained in it is obtained by repeating the above steps to continuously divide and optimize the samples according to the features, integrating multiple trained sub-decision trees into a random forest, and using this model to predict the current data to obtain the predicted value f of feature type j j (x i ), and evaluate whether the model training is complete by calculating the average relative error indicator.

5. The adaptive intelligent current limiting and management method according to claim 4, characterized in that: The specific method of evaluating whether the model training is complete by calculating the average relative error index is as follows: The calculation formula of the average relative error index O is as follows: Set the average relative error threshold. When O is less than or equal to the threshold, the model training is considered complete, and the trained model is used to predict the current.

6. The adaptive intelligent current limiting and management method according to claim 1, characterized in that: Step three specifically includes, based on the historical current data, screening abnormal state data that causes current short circuit, the abnormal state refers to the abnormal current state that can cause current short circuit, performing correlation mining on the historical current data, and performing statistical analysis on the current data before each abnormal state occurs based on the historical current data to obtain a set of current characteristic value range intervals that cause the abnormal state within time t. The obtained intervals are as follows: where a m represents the maximum current value at time t, a v represents the average current value at time t, a d Indicates the discreteness of the current value at time t, a t The current change rate at adjacent time points, a1, a2, a3, a4 are the corresponding eigenvalue change parameters obtained by mining the association law. The number of times each eigenvalue in the set interval causes a current short circuit is counted to quantify the short circuit correlation corresponding to the current data. The specific formula is as follows: The time range is set to t, and n time intervals are divided within the time t. The characteristic value is extracted from each time interval, r j represents the short-circuit correlation of feature type j at time t, x i represents the i-th eigenvalue belonging to this feature type, y i Represents the eigenvalue x i The number of short circuits that cause current, represents the average value of n eigenvalues ​​at time t, represents the average number of current short circuits caused by n characteristic values ​​at time t, r j The smaller the value, the smaller the correlation between feature type j and short circuit risk; According to the short-circuit correlation of each feature type, the corresponding weight is set, and the predicted current data is analyzed to determine whether there is a short-circuit risk. r1 represents the short-circuit risk value of the maximum current, and the corresponding weight is k1. r2 represents the short-circuit risk value of the average current, and the corresponding weight is k2. r3 represents the short-circuit risk value of the discrete degree of the current value, and the corresponding weight is k3. r4 represents the short-circuit risk value of the current change rate at adjacent time points, and the corresponding weight is k4. According to the short-circuit risk value of each feature type, the corresponding weight is set. According to the weight, the short-circuit risk value is set to determine the short-circuit risk of the predicted current data. The specific formula for setting the short-circuit risk value L is as follows: Where j represents the jth feature type, N j Indicates the total number of predicted feature values ​​of the jth feature type, M j Indicates the number of current characteristic value ranges that predict characteristic values ​​and lead to abnormal conditions; sets short-circuit risk value thresholds y1 and y2, and y2>y1. When L is greater than or equal to y2, it is judged that the current current state has a short-circuit risk that needs to be handled first. When L is greater than or equal to y1 and less than y2, it is judged that the current current state has a general short-circuit risk. When L is less than y1, it is judged that no short-circuit risk has been found in the current current state.

7. The adaptive intelligent current limiting and management method according to claim 1, characterized in that: Step 4: Set the current limit protection value according to the node historical current data and system load conditions. The calculation formula of the current limit protection value I is as follows: Where H represents the protection factor, which is obtained by analyzing the historical normal current data, and Z S is the system equivalent impedance, Z L is the line impedance, E S is the equivalent phase potential of the system; The method for determining the protection factor range is as follows: Select n time periods in which the current value is lower than the historical current average value when the node is working normally, extract the current average value of the n time periods, and determine the lower limit H of the protection coefficient range. - , the specific calculation formula is as follows: in, represents the average current value in the i-th time period, Indicates the historical current average value when the node is working normally; Similarly, based on the historical normal current data, select m time periods with current values ​​higher than the historical current average value when the node is working normally, extract the current average value of the m time periods, and determine the upper limit H of the protection coefficient range. + , the specific calculation formula is as follows: in, represents the average current value of the i-th time period. - , H + ) and select the H value.

8. The adaptive intelligent current limiting and management method according to claim 1, characterized in that: Step 5: Set the node abnormal state number threshold g, establish a blacklist system, and put nodes that cause abnormal states more than or equal to g times within time T into blacklist management.

Images