Power grid multi-agent large model safety evaluation index calculation method

Abstract

The invention provides a power grid multi-agent large model safety evaluation index calculation method, and relates to the technical field of intelligent control. The method comprises the following steps: acquiring a voltage change rate and a frequency change rate, dividing four stages according to disturbance starting and extreme points, pairing curve trends, judging delay and synchronization characteristics, extracting a continuous change structure, and generating a disturbance subsection evolution structure; according to the method, through staged analysis of node voltage and frequency change trends and combination of trend consistency judgment and abnormal segment identification, dynamic structure reconstruction and trend mapping of a disturbance process are realized, and coherence and precision of index response state identification are improved through trend boundary collection and frequency classification; cooperative characteristic identification is enhanced through the coupling direction and delay analysis of inter-node multi-parameter change, the node adjustment priority is defined through regulation and control response sorting, an adjustment path under multi-dimensional linkage is constructed, and the refined management and control capacity of disturbance response and the pertinence of control decision are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control, and particularly to a method for calculating safety assessment indicators of a large multi-agent model of a power grid. Background Art

[0002] The technical field of intelligent control includes a technical system that uses methods such as control theory, artificial intelligence, and system modeling to dynamically adjust and control decision-making for state parameters in complex systems. The core content of this technical field is to achieve autonomous adjustment and efficient control of the system in a dynamic environment, mainly including coordinated control of multi-agent systems, distributed decision-making and calculation, state prediction and optimization modeling, etc. It is widely applied in scenarios such as power systems, unmanned driving, industrial automation, energy management, etc., involving multiple links such as modeling methods, control strategies, and evaluation mechanisms. By integrating multi-source data and learning mechanisms, it realizes refined management and control of the operating state of the target system.

[0003] Among them, the method for calculating safety assessment indicators of a large multi-agent model of a power grid refers to, on the basis of introducing the coordinated operation of a multi-agent system and a large-scale model in a power system, by setting multiple safety assessment indicators, using the interaction information between agents, and based on a preset state information sampling rule and index function calculation method, numerically quantifying and evaluating each operating state. Specifically, it covers a method for extracting stability assessment indicators calculated based on node topological attributes, a voltage over-limit judgment logic constructed based on operating state parameters, and a method for generating global risk measurement indicators formed by analyzing multi-agent interaction modes. This method completes the hierarchical calculation and combined generation of various indicators through means such as rule definition, logic function construction, and data mapping.

[0004] In the prior art, during the process of dealing with the evolution of dynamic disturbances, there is a lack of the ability to identify the changing trend of parameters in time stages, resulting in the difficulty of visually presenting the index response path caused by disturbances, and affecting the continuity and structure of state recognition. In trend correlation analysis, there is no unified standard for judging the trend consistency between multiple parameters, and it is easy to ignore short-term non-cooperative behaviors between trends, making it difficult to effectively locate trend breakpoints and abnormal segments. The trend change process does not form a composite judgment mechanism for time distribution and trend frequency, resulting in insufficiently fine judgment of the triggering conditions of regulation behaviors and difficulty in adapting to the response classification requirements of various types of indicators in actual scenarios. The parameter change relationship between nodes lacks the ability to identify delay and direction, making it difficult to form a coupling path for strongly coupled state changes and increasing the limitations of the analysis of the impact of group behaviors. In the adjustment logic, the boundary response and sorting rules are fuzzy, resulting in conflicts in the node response order and a lack of a clear hierarchical priority framework for the adjustment path. Taking the power system as an example, node regulation measures mostly rely on fixed threshold judgments, ignoring the influence of differences between nodes, and easily causing the superposition of regulation errors, affecting the overall stability and regulation efficiency of the system. Summary of the Invention

[0005] To solve the technical problem in the prior art that there is a lack of time-stage identification of the parameter change trend during the process of dealing with dynamic disturbance evolution, an embodiment of the present invention provides a method and device for calculating safety assessment indicators of a power grid multi-agent large model. The technical solution is as follows:

[0006] On the one hand, a method for calculating safety assessment indicators of a power grid multi-agent large model is provided. This method is implemented by a safety assessment indicator calculation device, and the method includes:

[0007] S1: Obtain the voltage change rate and frequency change rate, divide it into four stages according to the disturbance start and extreme points, pair the curve trends, judge the delay and synchronization characteristics, extract the continuous change structure, and generate the disturbance segmented evolution structure;

[0008] S2: According to the disturbance segmented evolution structure, extract three types of trend parameters, judge the trend consistency, locate the abnormal segments and turning points, sort out the trend transfer paths, and generate the trend correlation pattern sequence;

[0009] S3: Call the trend correlation pattern sequence, extract the trend transformation segments in adjacent time periods of the nodes, record the change frequency and judge whether it exceeds the index response range, perform boundary collection on the start and end positions of the drift segments, classify the node states according to the boundary distribution quantity and trend duration, identify the differential response path characteristics, and obtain the index participation change categories;

[0010] S4: Call the index participation change categories, screen the three-parameter synchronous combinations, identify the influence direction and response order, classify and combine according to the cooperation frequency, and obtain the parameter coupling distribution type;

[0011] S5: Based on the parameter coupling distribution type, extract the combined response intervals, judge the boundary adjustment direction, sort the combination priorities according to the influence degree, and generate the multi-index linkage sorting rules.

[0012] On the other hand, a device for calculating safety assessment indicators of a power grid multi-agent large model is provided. This device is applied to the method for calculating safety assessment indicators of a power grid multi-agent large model, and the device includes:

[0013] A disturbance segmented evolution structure generation module, configured to obtain the voltage change rate and frequency change rate, divide it into four stages according to the disturbance start and extreme points, pair the curve trends, judge the delay and synchronization characteristics, extract the continuous change structure, and generate the disturbance segmented evolution structure;

[0014] A trend correlation pattern sequence generation module, configured to extract three types of trend parameters according to the disturbance segmented evolution structure, judge the trend consistency, locate the abnormal segments and turning points, sort out the trend transfer paths, and generate the trend correlation pattern sequence;

[0015] The index response path feature recognition module is used to call the trend correlation pattern sequence, extract the trend transformation segments in adjacent periods of nodes, record the change frequencies and determine whether they exceed the index response range, perform boundary aggregation on the start and end positions of the drift segments, classify the node states based on the boundary distribution quantity and the trend duration, identify the differential response path features, and obtain the index participation change categories;

[0016] The parameter coupling distribution type acquisition module is used to call the index participation change categories, screen the three-parameter synchronous combinations, identify the influence directions and response orders, classify and combine them according to the cooperation frequencies, and obtain the parameter coupling distribution types;

[0017] The multi-index linkage sorting rule generation module is used to extract the combined response intervals based on the parameter coupling distribution types, judge the boundary adjustment directions, sort the combination priorities according to the influence degrees, and generate the multi-index linkage sorting rules.

[0018] On the other hand, a security assessment index calculation device is provided. The security assessment index calculation device includes: a processor; a memory, and a computer-readable instruction is stored on the memory. When the computer-readable instruction is executed by the processor, any one of the methods in the above-mentioned power grid multi-agent large model security assessment index calculation method is implemented.

[0019] On the other hand, a computer-readable storage medium is provided. At least one instruction is stored in the storage medium, and the at least one instruction is loaded and executed by the processor to implement any one of the methods in the above-mentioned power grid multi-agent large model security assessment index calculation method.

[0020] The beneficial effects brought by the technical solutions provided in the embodiments of the present invention at least include:

[0021] The present invention proposes a power grid multi-agent large model security assessment index calculation method. By stagewise analyzing the change trends of node voltages and frequencies, combining trend consistency judgment and abnormal segment identification, dynamic structure reconstruction and trend mapping of the disturbance process are realized. Trend boundary aggregation and frequency classification improve the coherence and accuracy of index response state identification. The coupling direction and delay analysis of multi-parameter changes between nodes enhance the identification of collaborative characteristics. The regulation response sorting clarifies the adjustment priorities of nodes, constructs the adjustment paths under multi-dimensional linkage, and improves the refined management and control ability of disturbance response and the pertinence of control decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for description in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0023] Figure 1 It is a flowchart of a method for calculating safety evaluation indexes of a large model of a multi-agent power grid provided by an embodiment of the present invention;

[0024] Figure 2 It is a block diagram of a device for calculating safety evaluation indexes of a large model of a multi-agent power grid provided by an embodiment of the present invention;

[0025] Figure 3 It is a schematic structural diagram of a device for calculating safety evaluation indexes provided by an embodiment of the present invention. Specific embodiments

[0026] The following describes the technical solutions in the present invention with reference to the accompanying drawings.

[0027] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as an "example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of the word "example" is intended to present concepts in a specific manner. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two.

[0028] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same. "(of)", "corresponding", and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same.

[0029] In the embodiments of the present invention, sometimes a subscript such as W 1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meanings they express are the same.

[0030] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0031] The embodiments of the present invention provide a method for calculating safety evaluation indexes of a large model of a multi-agent power grid. This method can be implemented by a device for calculating safety evaluation indexes, and this device for calculating safety evaluation indexes can be a terminal or a server. As Figure 1 shown in the flowchart of the method for calculating safety evaluation indexes of a large model of a multi-agent power grid, the processing flow of this method can include the following steps:

[0032] S1: Obtain the voltage change rate and frequency change rate, divide it into four stages according to the disturbance start and extreme points, pair the curve trends, judge the delay and synchronization characteristics, extract the continuous change structure, and generate the disturbance segmented evolution structure.

[0033] Among them, the disturbance segmented evolution structure includes voltage change trend characteristics, frequency change trend characteristics, in-stage parameter response relationships, and structural stage arrangement patterns;

[0034] The trend correlation pattern sequence includes abnormal trend identification points, trend transfer time series, node trend arrangement structures, and trend direction consistency marks;

[0035] The index participation change categories include trend change frequency statistics, trend drift boundary distribution, response duration classification, and control index response types;

[0036] The parameter coupling distribution types include parameter change coupling combinations, cooperation relationship types, response delay characteristics, and propagation path patterns;

[0037] The multi-index linkage sorting rules include control response interval sets, boundary matching priorities, node influence rankings, and adjustment operation records.

[0038] Optionally, obtain the voltage change rate and frequency change rate, divide it into four stages according to the disturbance start and extreme points, pair the curve trends, judge the delay and synchronization characteristics, extract the continuous change structure, and generate the disturbance segmented evolution structure, including:

[0039] S101: Obtain the voltage change rate and frequency change rate of the nodes in the power grid, segment the time axis according to the disturbance start point and extreme point sequences, mark the order of the four time intervals, and generate the stage division interval sequence.

[0040] In a feasible implementation, rely on the actual power grid operation data for time series sampling. For example, taking a certain node in a 220 kV substation as an example, extract the frequency and voltage data from 10 seconds before the disturbance to 20 seconds after the disturbance in the system operation monitoring platform. The sampling period is set to 0.1 second, and a total of 300 data points are collected. Calculate the difference between adjacent voltage values and frequency values divided by the time interval between every two consecutive data points respectively to form the voltage change rate sequence and the frequency change rate sequence, and perform preliminary disturbance identification operations on the two sequences respectively.

[0041] For the identification of the disturbance starting point, search for the section in the voltage change rate sequence where the change rate of three consecutive sampling points first exceeds the change amplitude threshold of 0.05 as the initial disturbance time point. For example, in a certain disturbance, it is found that the voltage change rate at 5.2 seconds exceeds 0.05 for three consecutive points, and this moment is determined as the disturbance starting point. Then, by calculating the change trends of the voltage and frequency change rate sequences in each time period after the disturbance, search for the point where the absolute value of the change rate continuously increases and forms a peak in a local time period as the extreme point. The identification logic is as follows: the difference between the maximum and minimum values among five consecutive points exceeds the change amplitude threshold of 0.1, and the trend reverses significantly after this point. For example, when the voltage change rate gradually decreases from a positive value to a negative value at 7.8 seconds, it is determined as an extreme point. Finally, combining the disturbance starting point and multiple extreme points, divide the overall disturbance process into four stages in chronological order. The first stage is the initial disturbance response stage, the second stage is the response peak stage, the third stage is the system fallback stage, and the fourth stage is the gradual recovery stage. Taking a certain disturbance data as an example, the disturbance starting point is 5.2 seconds, and the extreme points are 6.1 seconds, 8.3 seconds, and 12.0 seconds in sequence, and the system returns to a stable point at 20.0 seconds. Finally, four time intervals are formed, namely: [5.2 seconds, 6.1 seconds], [6.1 seconds, 8.3 seconds], [8.3 seconds, 12.0 seconds], and [12.0 seconds, 20.0 seconds]. Each interval is marked with a number for subsequent analysis and processing. The finally output stage division interval sequence is four sequential intervals.

[0042] S102: Extract the voltage and frequency change rate curves according to the stage division interval sequence, judge the starting time difference between the rising and falling segments of the two curves within the stage, screen the sections with an offset value less than the disturbance detection threshold as the synchronization interval, and combine the voltage direction change corresponding to the frequency peak to obtain the synchronization interval delay value set.

[0043] In a feasible implementation, extract the corresponding segments of the voltage change rate sequence and the frequency change rate sequence respectively in each time period to form the voltage change rate curve and the frequency change rate curve within the stage. Conduct characteristic trend analysis on each segment of the curve to judge whether there is an obvious rising or falling segment. The judgment method is to observe whether the change rates of five consecutive sampling points increase or decrease in the same direction, and whether the average change rate is greater than the judgment threshold of 0.03 for distinguishing the disturbance change trend. For example, in the second stage, the voltage change rate continuously rises from 6.3 seconds to 6.8 seconds, and the frequency change rate starts to show the same rising trend at 6.4 seconds. Then record the starting points of their rising segments as 6.3 seconds and 6.4 seconds respectively, and calculate the starting time difference between the two as 0.1 second. This time difference is less than the system-set disturbance detection threshold of 0.3 seconds. Therefore, this time period is determined as the voltage and frequency synchronous change interval.

[0044] In the stage of extracting the synchronization interval delay value, the time point corresponding to the extreme point of the frequency change rate curve within the synchronization interval is extracted as the peak time. For example, the frequency reaches the maximum value at 6.9 seconds. At the same time, the trend direction of the voltage change rate is analyzed within 0.2 seconds before and after this time point. If the voltage change rate changes from a positive value to a negative value during this period, the direction change is from rising to falling, and the direction is marked as "falling". At this time, record the frequency peak time point of 6.9 seconds and the voltage change direction, and calculate the time interval between the two as the synchronization delay value, that is, 0.3 seconds. This process is repeated in multiple synchronization sections to form multiple synchronization delay values with direction information. For example, the delay values in three typical sections are 0.1 second, 0.2 seconds, and 0.3 seconds respectively, and the directions are "rising", "falling", "rising" respectively, forming the final synchronization interval delay value set for the next stage of disturbance structure analysis.

[0045] S103: Based on the synchronization interval delay value set, sort the voltage and frequency change trends, call the inflection point segment direction sequence, identify the structure arrangement type, and generate the disturbance segmented evolution structure.

[0046] In a feasible implementation manner, arrange them in ascending order according to the delay values. After sorting, the delay values are 0.1 second, 0.2 seconds, and 0.3 seconds in sequence. At the same time, keep the corresponding voltage change direction sequence as "rising", "falling", "rising". Calculate the average change rate of the voltage change rate and the frequency change rate in each stage. The calculation method of the average change rate is the difference between the starting value and the ending value of the curve in this stage divided by the stage time length. For example, in the second stage, the voltage change rate changes from 0.05 to 0.20, the stage length is 2.2 seconds, the average change rate is 0.068, and the frequency change rate changes from 0.02 to 0.18, the average change rate is 0.073. After obtaining the change trend rate set of multiple stages, combined with the inflection point information of each direction change in the previous synchronization interval, mark the stage where the voltage or frequency changes from an upward trend to a downward trend or from a downward trend to an upward trend, and construct the inflection point segment direction sequence, such as "rising → falling", "falling → rising", etc.

[0047] Compare the sorted direction sequence with the inflection point segment direction sequence to determine whether the overall disturbance process constitutes a specific disturbance structure type. For example, if the direction sequence is "rising, falling, rising" and the inflection point segment direction is "rising → falling, falling → rising", it means that the overall disturbance process presents a symmetric structure with a depressed middle and high at both ends. Or if the delay times are uneven and the direction changes are chaotic, it is determined as a multi-stage stepped disturbance process. Finally, form a disturbance segmented evolution structure identifier with a clear structure to support subsequent system response modeling or power equipment status prediction analysis.

[0048] S2: According to the segmented evolution structure of the disturbance, extract three types of trend parameters, judge the trend consistency, locate the abnormal segments and turning points, sort out the trend transfer paths, and generate a sequence of trend correlation patterns.

[0049] Optionally, according to the segmented evolution structure of the disturbance, extract three types of trend parameters, judge the trend consistency, locate the abnormal segments and turning points, sort out the trend transfer paths, and generate a sequence of trend correlation patterns, including:

[0050] S201: Obtain the node data of each disturbance stage in the segmented evolution structure of the disturbance, extract the load change rate, voltage fluctuation amplitude, and frequency deviation direction of the nodes within the same stage, and judge the direction consistency state among the three based on the sign of the change slope of the three parameters within the current time period. If the slope directions of the three items are inconsistent, record the corresponding node and time period as an abnormal segment, and generate a set of abnormal trend segment intervals.

[0051] In a feasible implementation manner, for each defined stage, locate the sampling data of each key node in the power grid. The node data should at least include the real-time load value, voltage value, and frequency value. Taking the node A of the 220 kV substation as an example, during a certain disturbance process, the first stage is [5.2 s, 6.1 s]. Extract the load value, voltage value, and frequency value every 0.1 s within this time period, and a total of 9 sets of time series data are collected.

[0052] Calculate the change rate for each parameter sequence, that is, the change amount between every two adjacent data points divided by the time interval of 0.1 s to obtain three sequences of load change rate, voltage volatility, and frequency change rate. Then judge the slope sign of each change rate sequence. A positive slope sign indicates an upward trend, and a negative sign indicates a downward trend. Subsequently, take the values in the three sequences within this time period whose slope signs are greater than zero or less than zero, and mark the trend directions respectively. If at least two of the three parameters have a positive slope and one has a negative slope, or two have a negative slope and one has a positive slope, it is judged that the slope directions of the three items are inconsistent. Taking node A in this stage as an example, the load change rate shows a positive increase, the voltage fluctuation amplitude change rate decreases, and the frequency change rate shows a downward trend. At this time, only the load is positive among the three, and the other two are negative, so it is judged that the directions are inconsistent. If the slope direction of node A remains consistent in other stages, no record is made. The system traverses all nodes and all stages, performs the above direction judgment operation for each stage and each node, and at the same time records all nodes with inconsistent slope directions and their corresponding time periods. Finally, summarize this information to form a set of abnormal trend segment intervals. Each record in this set includes the node number, abnormal stage number, start and end times of the time period, and the slope direction values of the three parameters.

[0053] S202: Invoke the abnormal segment time intervals of the record nodes in the abnormal trend segment interval set, locate the inflection point time of the load change rate curve within each abnormal segment, match the starting position of the slope change before the inflection point as the trend starting point, summarize the time range between the trend starting point and the corresponding inflection point, arrange the segment trend transfer relationships in chronological order, and generate a trend transfer structure sequence.

[0054] In a feasible implementation, perform the inflection point location operation on the load change rate curve in each record.

[0055] Starting from the starting position of the current abnormal time period, observe the change trend of the load change rate value point by point. When multiple consecutive points change from rising to falling or from falling to rising, record the time point of this trend reversal as the inflection point. For example, in an abnormal segment [6.1s, 8.3s] of node B, the load change rate continuously increases from 6.1 to 6.7 seconds and gradually decreases from 6.8 seconds. Then record the inflection point time as 6.8 seconds. At the same time, trace back from the inflection point time to determine the time point when the continuous slope change (positive or negative) first appears as the trend starting point. If a continuous increasing trend appears starting from 6.3 seconds, the trend starting point is 6.3 seconds. The trend change segment within this stage is [6.3s, 6.8s]. Record this time range and mark the direction as rising to falling. The system performs the above operations on each abnormal segment, extracts the trend starting point and inflection point time pairs, generates trend transfer record entries, and arranges all trend transfer entries in chronological order to ensure that a coherent evolution sequence can be formed between adjacent trend segments. For example, node C has three trend transfer segments [6.5s, 7.1s], [7.2s, 7.8s], and [7.9s, 8.4s] in three abnormal segments, and record the direction changes as "rising → falling", "falling → rising", and "rising → falling" respectively, forming a complete trend transfer structure sequence. This sequence reflects the transfer logical relationship of the load change rate during multiple trend turns and is output in chronological order.

[0056] S203: According to the trend transfer structure sequence, extract the trend order combinations of the three parameters within each segment, sort the differential combination sequence based on the time axis, merge the repeated paragraphs with the same structure, count the occurrence times and order positions of the trend repeated structures, establish the trend structure arrangement sequence corresponding to the node, and generate the trend association pattern sequence.

[0057] In a feasible implementation manner, in each trend transition segment, the change rate of the voltage fluctuation amplitude and the change rate sequence of the frequency offset direction corresponding to the time period thereof are extracted, and the trend directions of the three parameters are arranged and combined to form the trend order combination label of this segment. For example, in the trend segment [6.3s, 6.8s], if the load change rate is rising, the voltage fluctuation amplitude is falling, and the frequency offset direction is rising, then record the combination of this segment as "load↑ - voltage↓ - frequency↑". Collect the trend order combinations of all paragraphs in chronological order, and merge the trend segments with exactly the same combination form into the same structure. For example, if the combination "load↑ - voltage↓ - frequency↑" appears 3 times at a certain node, occurring at [6.3s, 6.8s], [7.5s, 8.0s] and [9.2s, 9.8s] respectively, then record the occurrence times of this combination structure as 3, and the sequential positions as the 1st, 4th and 6th segments. At the same time, correspond the node numbers with the combination labels one by one, count the sequence of appearance of all trend combinations at this node, form the trend structure arrangement sequence of the node, then output the set of trend structure arrangement sequences corresponding to each node with each node as an index, and finally, by traversing all node trend combination sequences, collect the structural units of the repeated combinations to form the trend structure type index and the trend association pattern sequence.

[0058] Among them, the occurrence frequency, sorting and distribution of each combination label in the overall node set can be merged and mapped accordingly to obtain the complete trend evolution type structure.

[0059] S3: Call the trend association pattern sequence, extract the trend transformation segments in adjacent time periods of the node, record the change frequency and judge whether it exceeds the index response range, perform boundary collection on the start and end positions of the drift segment, classify the node state in combination with the boundary distribution quantity and the trend duration, identify the differential response path characteristics, and obtain the index participation change category.

[0060] Optionally, call the trend association pattern sequence, extract the trend transformation segments in adjacent time periods of the node, record the change frequency and judge whether it exceeds the index response range, perform boundary collection on the start and end positions of the drift segment, classify the node state in combination with the boundary distribution quantity and the trend duration, identify the differential response path characteristics, and obtain the index participation change category, including:

[0061] S301: Call the trend sequence of the node in the trend association pattern sequence, extract the trend change start point and the trend holding end point corresponding to the adjacent time period, mark the change occurrence position and the holding interval respectively according to the trend direction change position and the duration length within the adjacent segment, and generate the trend transformation structure information group.

[0062] In a feasible implementation, the time information between two adjacent segments in the trend sequence of each node is extracted, and the end time of the previous segment trend is recorded as the trend retention end point, and the start time of the subsequent segment trend is recorded as the trend change start point, thereby forming trend connection boundary information. Taking node A as an example, its trend sequence segments are [5.2s, 6.3s], [6.3s, 6.9s], [6.9s, 7.5s], then the trend change start points are recorded as 6.3s and 6.9s respectively, and the corresponding trend retention end points are 6.3s and 6.9s. During the structure generation process, it is necessary to identify the trend direction of each segment. The direction is based on the change trends of three parameters: load, voltage, and frequency. If all three parameters of the previous segment are rising and at least two direction changes occur in the subsequent segment, for example, the voltage and frequency change from rising to falling, then mark this position as the trend change occurrence position, and calculate the retention duration for the duration of the previous segment trend. For example, the trend direction of the segment [5.2s, 6.3s] is consistent, and the duration is 1.1 seconds, forming a trend retention section, and the corresponding change position 6.3s is marked as the trend change point. If there are direction changes in three consecutive trend segments and the time periods are 1.1 seconds, 0.6 seconds, and 0.9 seconds respectively, then record the three trend transformation structural units in sequence, each containing the change start point, retention end point, trend duration, and change direction combination, and generate a trend transformation structure information group for subsequent trend change frequency analysis.

[0063] S302: According to the trend direction change positions and time intervals recorded in the trend transformation structure information group, calculate the change frequency index of the node within the trend direction change segment, call the upper limit value of the index response range of the node as the judgment threshold, classify the trend segments whose change frequency exceeds the threshold, and obtain the set of drift trend segment intervals;

[0064] Among them, the calculation formula for the change frequency index of the node within the trend direction change segment is as follows in formula (1):

[0065] (1);

[0066] Among them, represents the node within the trend segment the change frequency index, represents the trend segment the time length of, represents the node at the th sampling point the frequency change direction slope, represents the node at the th sampling point the voltage change direction slope, represents the node at the th sampling point the load change direction slope, , , respectively represent the direction slope weight factors of frequency, voltage and load changes at the th sampling point, is the number of sampling points within the trend segment .

[0067] In a feasible implementation, formula (1) aims to calculate the average change frequency of node within a certain trend transformation segment, where is the total time length of the trend segment , is the number of sampling points within this time period. The three main components in the formula respectively involve frequency change ( ), voltage change ( ), and load change ( ). For each sampling point , the change direction slope of these three parameters will be calculated multiplied by its corresponding weight .

[0068] Set the specific example data as follows: In the trend segment , assume the time length hours (i.e., 3600 seconds). This trend segment contains 360 sampling points, that is . Suppose at a typical sampling point , the change slope of frequency Hz / s, the change slope of voltage % / s, and the change slope of load % / s. To adjust the influence of parameters on the results, set the weights , , , reflecting that in this monitoring, the change of frequency has a greater impact on system stability, followed by voltage, and the impact of load is relatively small.

[0069] Substitute the above values into the formula for calculation:

[0070] ;

[0071] ;

[0072] ;

[0073] ;

[0074] ;

[0075] The result shows that within the entire trend segment, there are 0.0013 changes per second on average, which characterizes the frequency of trend changes of the node during this period. If the set threshold is 0.002 changes per second, it means that the change frequency of the current node does not exceed this threshold, so it is not classified as a drift trend segment. This is of great guiding significance for further trend analysis and node behavior classification.

[0076] S303: Based on the start and end points of the trend segments in the set of drift trend segment intervals, group the boundary positions of each segment by node, count the number of boundary segments corresponding to each node and extract the corresponding time tags, and classify and organize the boundary distribution positions in time series to obtain the trend boundary distribution structure.

[0077] In a feasible implementation, extract the start and end points of all drift segments of each node as boundary position data respectively, and record the corresponding time tags. For example, if the drift segments of node D are [5.5s, 6.2s], [7.1s, 7.8s], and [9.0s, 9.7s], then the number of boundary segments of node D is 3 segments, and the corresponding boundary points are 6 time points: 5.5s, 6.2s, 7.1s, 7.8s, 9.0s, and 9.7s. Arrange these boundary points in time series to obtain the boundary distribution order sequence.

[0078] After performing this operation on all nodes, form the boundary quantity statistical table and the corresponding time tag set of each node. Then, classify all time points in the order of the global time axis, and count how many node boundaries are concentrated in the same time window. If there are more than 5 node trend boundaries in a certain time period, such as [6.0s, 6.5s], mark this segment as a trend boundary concentration segment. After performing time aggregation and distribution statistics on all concentration segments, output the trend boundary distribution structure. This structure is combined and arranged according to time tags, node numbers, boundary quantities, and density, which is convenient for subsequent trend feature classification and status recognition.

[0079] S304: Invoke the trend repetition range and residence time in the trend boundary distribution structure, perform combined classification according to the residence duration interval and the occurrence range of the trend segments, divide the regulatory response status of the nodes according to the differential trend distribution characteristics, and generate the change category of the participation index.

[0080] In a feasible implementation, the residence time of each node within a certain repeated trend section is statistically calculated, that is, the continuous time during which the trend direction remains unchanged. For example, if node E continuously exhibits the combination of "load decrease - voltage increase - frequency decrease" within the section [6.0s, 7.0s], the residence time is 1.0s. The repetition range refers to the total duration of the repeated occurrences of the same trend combination at different nodes or in different time periods. If the total repeated occurrence time of the same trend at multiple nodes exceeds 5 seconds, this trend combination is recorded as a highly repeated trend.

[0081] Different residence time intervals are set for division, such as short residence (<1.5s), medium residence (1.5s - 3.5s), and long residence (>3.5s). Each trend segment is classified into the corresponding interval, and at the same time, its type attribution is judged in combination with its repetition range. For example, if node E has a long residence in the segment [6.0s, 7.0s] and belongs to a highly repeated section, it is marked as a "stable response state". If node F shows variable trends in the segment [7.0s, 7.6s] but each duration is less than 1.0s and the repetition range is scattered, it is classified as a "rapid drift type". Finally, the trend response behaviors classified for each node are combined, and a unified classification label is formed for nodes with the same type of residence time characteristics and trend distribution characteristics, and the output is the set of change categories participated by the indicators.

[0082] S4: Call the change categories participated by the indicators, screen the synchronous combinations of the three parameters, identify the influencing direction and response order, classify and combine according to the cooperation frequency, and obtain the parameter coupling distribution type.

[0083] Optionally, call the change categories participated by the indicators, screen the synchronous combinations of the three parameters, identify the influencing direction and response order, classify and combine according to the cooperation frequency, and obtain the parameter coupling distribution type, including:

[0084] S401: Call the node classification results in the change categories participated by the indicators, identify the node combinations in which the three parameters of the frequency drift rate, voltage fluctuation amplitude, and load change rate in the same category of nodes change simultaneously within the same time period, extract the node combination list and the corresponding time index, and generate the synchronous change combination set.

[0085] In a feasible implementation, the nodes are clustered and grouped by category. Taking the node set under the same category as the operation object, traverse the change states of the three parameters of the frequency drift rate, voltage fluctuation amplitude, and load change rate of each group of nodes within the same time period, extract the trend direction of each parameter for each time slice. The trend direction is judged to be rising, falling, or remaining unchanged based on the difference between the current time point and the previous time point. If all three parameters change within this time slice, that is, at least two of the three differences have non-zero change signs and the third item is not in the "remaining unchanged" state, it is determined to be "changing simultaneously".

[0086] Group the node numbers that meet the conditions together and record the time index corresponding to this combination. For example, if the three parameters of nodes A, B, and C all change direction at [6.5s], then mark the combination as {A, B, C} and the time as 6.5s. Repeat the above operation to extract all synchronous change points of each group of classified nodes within the full cycle range, form the corresponding node combination list, and mark their respective corresponding time indexes. Each record in the combination list contains the node set, the change direction of the three parameters at this moment, and the time index information. Finally, generate a synchronous change combination set as the basic input for subsequent analysis of the collaborative trend.

[0087] S402: According to the node combinations recorded in the synchronous change combination set, calculate the change starting time difference index between the three parameters of the nodes, judge the order of change response, extract the delay time difference and direction relationship between the parameters, screen the positive and reverse collaborative segments according to the delay direction relationship, and mark the intersection time interval of the consistent change of the three parameters in each group to obtain the collaborative response structure sequence.

[0088] Among them, the calculation formula for the change starting time difference index between the three parameters of the nodes is as follows in formula (2):

[0089] (2);

[0090] Among them, represents the average time difference between the starting times of the changes of the three parameters of frequency, voltage, and load in the node combination , represents the starting time of the change of the frequency parameter in the node combination , represents the starting time of the change of the voltage parameter, represents the starting time of the change of the load parameter.

[0091] In a feasible implementation, formula (2) is used to calculate the change starting time difference index of the three parameters of frequency, voltage, and load in any node combination during the disturbance response process to quantify the time consistency degree of the three responses. All parameters are determined by the actual response data collected by the monitoring equipment.

[0092] Suppose the monitoring object is a combination composed of three key nodes of a 110 kV substation. Among them, the frequency monitoring equipment (based on PMU) collects data at intervals of 50 milliseconds, and the sampling periods of the voltage and load monitoring equipment are both 100 milliseconds. The acquisition time is uniformly aligned with the disturbance occurrence reference time. According to the data record, during a typical disturbance process, the starting points of the changes of the three parameters of the node combination m are: the starting time of the frequency Seconds, calculated based on the detection result of the trend that the frequency decreases from 50.00 Hz to 49.96 Hz and continues to decline. The change threshold is set at ±0.02 Hz, and the data is sourced from the slope analysis of the frequency curve collected by the synchronous measurement device; the starting time of the voltage Seconds, based on the voltage fluctuating from 220 kV to 217.3 kV, with the change amplitude exceeding the stable offset range of ±1.2%. The monitoring device continuously confirms the decreasing slope point by point; the starting time of the load Seconds, based on the load rapidly increasing from 86.5 MW to 92.0 MW and remaining at a relatively high level. In the slope analysis, it is confirmed that the slope is continuously positive and the amplitude exceeds 0.5 MW / s.

[0093] Substitute the above values into the formula for calculation:

[0094] ;

[0095] ;

[0096] The result shows that in this node combination, the average difference in the change response time of the three parameters is 0.1133 seconds, indicating that there is a certain degree of asynchronous offset among them during the change response process. The smaller this value, the more synchronous the response tends to be, and the larger the value, the stronger the degree of separation of the change order. In this step, this time difference index is used to determine the tightness of the order of parameter responses within the collaborative segment, and combined with the direction relationship, it further distinguishes the forward and reverse collaborative structures, providing a time dimension basis for the accurate construction of the subsequent collaborative response structure sequence.

[0097] S403: Call the combined collaborative records in the collaborative response structure sequence, count the number of occurrences of the same collaborative structure in the combination, screen out the combination relationships with a repetition frequency higher than the collaborative structure judgment threshold, summarize the delay paths within the collaborative chain in each type of combination in order, record the parameter change propagation path according to the node order, and generate the parameter coupling distribution type.

[0098] In a feasible implementation, identify and mark the parameter arrangement method, direction consistency, and delay time difference characteristics of the collaborative structure in each record. First, perform a unique encoding according to the direction combination and order method of the three parameters in the combination. For example, when the frequency → voltage → load are all rising, it is encoded as "F↑-V↑-L↑". Count the frequency of occurrence of this structure encoding in all collaborative records. If the frequency of occurrence of a certain type of structure encoding is greater than the collaborative structure judgment threshold, for example, the threshold is set at 3 times. If the structure "F↑-V↑-L↑" appears 5 times, 4 times, and 6 times at nodes E, F, and G respectively, then this structure meets the screening conditions and is retained as a typical structure.

[0099] For the eligible combination relationships, extract the delay paths of the three parameters in their collaborative chains, that is, record the starting sequential time series of parameter changes. Taking node F as an example, the starting point of frequency change is 7.10 s, the voltage is 7.15 s, and the load is 7.25 s. Then the sequential path is frequency → voltage → load. For all combination relationships that meet the conditions, aggregate them according to the node numbers. For each node, record the sequential path of parameter changes in its collaborative structure, and mark its corresponding structure type code and occurrence position. Finally, summarize to form the parameter coupling distribution types of each node, and record the occurrence frequency, the number of involved nodes, and the collaborative path sequence in the network for each type.

[0100] S5: Based on the parameter coupling distribution types, extract the combined response intervals, judge the boundary adjustment directions, sort the combination priorities according to the influence degrees, and generate the multi-index linkage sorting rules.

[0101] Optionally, based on the parameter coupling distribution types, extract the combined response intervals, judge the boundary adjustment directions, sort the combination priorities according to the influence degrees, and generate the multi-index linkage sorting rules, including:

[0102] S501: Based on the parameter coupling distribution types, extract the regulation response intervals of the nodes in the same type during the perturbation process, calculate the length of the effective response time interval, compare the interval boundary values with the currently set response threshold range, judge whether there is an out-of-bounds situation at the boundary, and identify the interval direction change state to generate the response boundary matching results.

[0103] Among them, the formula for calculating the length of the effective response time interval is as follows in formula (3):

[0104] (3);

[0105] Among them, represents the length of the effective response time interval of the parameter on node , represents the starting time of the response of this parameter, represents the termination time of the response of this parameter, represents node at the th sampling point, the change amount of the parameter on it, represents the sampling period, represents the response judgment threshold corresponding to the parameter .

[0106] In a feasible implementation manner, formula (3) is used to calculate the length of the effective change time interval of the parameter on a certain node during the perturbation response process. This interval is determined by the absolute value of the change rate continuously exceeding the response threshold is composed of sampling segments. The time interval between sampling points is , and the response interval starts from the starting point to the ending point and is formed.

[0107] Taking the frequency parameters of a certain 110 kV node as an example, the data is collected by the power grid monitoring system using a PMU device, and the sampling period is set to seconds. The national standard stipulates that the normal fluctuation range of the power grid frequency is ±0.2 Hz, and the minimum effective change slope threshold of this parameter is set to Hz / s for identifying the true disturbance change section.

[0108] During the disturbance process, the frequency sampling records of this node are as follows, with the unit being Hz: The sampling time ranges from 6.00 seconds to 6.80 seconds, and there are a total of 17 sampling points (at 0.05 - second intervals);

[0109] The data segment is: 50.00, 50.01, 50.03, 50.06, 50.10, 50.15, 50.19, 50.22, 50.23, 50.22, 50.18, 50.12, 50.05, 49.96, 49.89, 49.85, 49.84;

[0110] Calculate the change rate slope of each point (difference divided by the interval):

[0111] The slope of the 2nd point = (50.01 - 50.00) / 0.05 = 0.20 Hz / s;

[0112] The slope of the 3rd point = (50.03 - 50.01) / 0.05 = 0.40 Hz / s;

[0113] The slope of the 4th point = (50.06 - 50.03) / 0.05 = 0.60 Hz / s;

[0114] ……;

[0115] The slope of the 17th point = (49.84 - 49.85) / 0.05 = 0.20 Hz / s;

[0116] Judge whether the absolute slope of each of the above points is greater than , and those that meet the condition are marked as 1, and those that do not meet are marked as 0.

[0117] Among them, the point indices that meet the condition are from the 2nd to the 17th point, and a total of 16 consecutive points meet the effective response condition. The starting point of the corresponding time interval is 6.05 seconds, and the ending point is 6.80 seconds. Then the calculation of the response time interval length is:

[0118] ;

[0119] The result shows that the duration of the effective response of node k in this perturbation is 0.80 seconds. This value is used to determine whether the response reaches the regulation intervention threshold and provides a basic time identifier for subsequent boundary state determination. In subsequent steps, this start-stop interval will be jointly judged with the threshold overrun state to determine whether it constitutes a boundary abnormal state and is used for response boundary matching.

[0120] S502: According to the direction state recorded in the response boundary matching result, normalize the change amplitude value of the node, extract the corresponding influence degree score of each node during the perturbation, sort them from high to low according to the score, and sequentially assign adjustment serial number identifiers to the sorted nodes to establish the node adjustment sorting result.

[0121] In a feasible implementation, normalize the change amplitude value of each node. First, calculate the maximum change amplitudes of the three parameters of frequency, voltage, and load within the response section, and perform normalization based on the absolute value of the amplitude. The normalized value is the current change value divided by the maximum change reference value of the corresponding parameter. The reference value is taken from the maximum change record of this type of node within the perturbation influence area. For example, the maximum change reference value of frequency is set to 0.3 Hz, the voltage is 8%, and the load change rate is 15%. If the frequency change of node B is 0.18 Hz, then the normalized value is 0.6. The weighted average of the three normalized values is used to obtain the node influence score, and the weights can be set as 0.4, 0.3, and 0.3 for frequency, voltage, and load respectively. Sort according to the score values of each node from high to low. For example, if the score of node C is 0.82, the score of node A is 0.76, and the score of node F is 0.71, then node C ranks first, A ranks second, and F ranks third. Sequentially assign adjustment serial numbers to the sorted nodes, which are #1, #2, #3 respectively. Store the scores and adjustment serial numbers of all nodes in the sorting result list uniformly as the basis for the sequence of subsequent regulation linkages, and output the node adjustment sorting result.

[0122] S503: Call the node serial numbers in the node adjustment sorting result, match the interval positions in the response boundary matching result, mark the boundary change direction of the sorted nodes within the self-response interval, and extract the continuous paragraphs of the response boundary. Arrange the adjustment order according to the continuous intervals first to generate a multi-index linkage sorting rule.

[0123] In a feasible implementation, locate the corresponding interval positions in the response boundary matching result for the node adjustment sorting result, and search for the boundary state marking information of its regulation section node by node. For example, the direction records of the three response sections of node #1 are "frequency rising out of bounds", "voltage rising out of bounds", and "load dropping normally", then it is marked as having multi-parameter positive boundary change characteristics.

[0124] Identify whether there are consecutive paragraphs in the response section of the node, that is, whether the time interval between the end point of the previous response segment and the start point of the next segment is less than the set continuity determination threshold. The continuity threshold is set according to the minimum adjustment period of the control response. For example, if it is set to 0.5 seconds, and the two segments of node #1 are [6.0s, 6.8s] and [6.9s, 7.4s], with an interval of 0.1s which is less than 0.5s, it is determined to be continuous. If the length of the consecutive paragraphs exceeds 2 segments, the adjustment order is arranged according to the principle of continuous segment priority. After performing the same processing on all sorted nodes, the nodes with continuous response areas are arranged at the front of the priority response list, and the rest are arranged backward according to the scoring sort position. Finally, a multi-index linkage sorting rule structure is established. Each node in this structure is attached with its adjustment serial number, the number of continuous segments, the boundary direction state, and the time position, which is used to support the formulation of the response adjustment plan.

[0125] The present invention proposes a method for calculating safety assessment indicators of a power grid multi-agent large model. By analyzing the changing trends of node voltage and frequency in stages, combining trend consistency judgment and abnormal segment identification, it realizes the dynamic structure reconstruction and trend mapping of the disturbance process. The trend boundary collection and frequency classification improve the coherence and accuracy of index response state identification. The coupling direction and delay analysis of multi-parameter changes between nodes enhance the identification of collaborative characteristics. The regulation response sorting clarifies the adjustment priorities of nodes, constructs an adjustment path under multi-dimensional linkage, and improves the refined management and control ability of disturbance response and the pertinence of control decisions.

[0126] Figure 2 It is a block diagram of a device for calculating safety assessment indicators of a power grid multi-agent large model shown according to an exemplary embodiment. This device is used for the method of calculating safety assessment indicators of a power grid multi-agent large model. Refer to Figure 2 , this device includes a disturbance segmented evolution structure generation module 210, a trend correlation pattern sequence generation module 220, an index response path feature recognition module 230, a parameter coupling distribution type acquisition module 240, and a multi-index linkage sorting rule generation module 250. Among them:

[0127] The disturbance segmented evolution structure generation module 210 is used to obtain the voltage change rate and the frequency change rate, divide it into four stages according to the disturbance start and extreme points, pair the curve trends, judge the delay and synchronization characteristics, extract the continuous change structure, and generate the disturbance segmented evolution structure;

[0128] The trend correlation pattern sequence generation module 220 is used to extract three types of trend parameters according to the disturbance segmented evolution structure, judge the trend consistency, locate the abnormal segments and turning points, sort out the trend transfer paths, and generate the trend correlation pattern sequences;

[0129] The index response path feature recognition module 230 is used to call the trend correlation pattern sequence, extract the trend transformation segments in adjacent time periods of nodes, record the change frequency and determine whether it exceeds the index response range, perform boundary aggregation on the start and end positions of the drift segments, classify the node states based on the boundary distribution quantity and the trend duration, identify the differential response path features, and obtain the index participation change categories;

[0130] The parameter coupling distribution type acquisition module 240 is used to call the index participation change categories, screen the three-parameter synchronous combinations, identify the influence direction and response order, classify and combine according to the cooperation frequency, and obtain the parameter coupling distribution types;

[0131] The multi-index linkage sorting rule generation module 250 is used to extract the combined response intervals based on the parameter coupling distribution types, judge the boundary adjustment direction, sort the combination priorities according to the influence degree, and generate the multi-index linkage sorting rules.

[0132] Among them, the disturbance segmented evolution structure includes voltage change trend features, frequency change trend features, in-phase parameter response relationships, and structural phase arrangement patterns;

[0133] The trend correlation pattern sequence includes abnormal trend identification points, trend transfer time series, node trend arrangement structures, and trend direction consistency markers;

[0134] The index participation change categories include trend change frequency statistics, trend drift boundary distribution, response duration classification, and regulation index response types;

[0135] The parameter coupling distribution types include parameter change coupling combinations, cooperation relationship types, response delay characteristics, and propagation path patterns;

[0136] The multi-index linkage sorting rules include regulation response interval sets, boundary matching priorities, node influence rankings, and adjustment operation records.

[0137] Optionally, the disturbance segmented evolution structure generation module 210 is further used for:

[0138] S101: Obtain the voltage change rate and frequency change rate of nodes in the power grid, segment the time axis according to the disturbance start point and extreme value point sequences, mark the order of four time intervals, and generate a stage division interval sequence;

[0139] S102: Extract the voltage and frequency change rate curves according to the stage division interval sequence, judge the start time difference between the rising and falling segments of the two curves in the stage, screen the segments with an offset value less than the disturbance detection threshold as the synchronous intervals, and obtain the synchronous interval delay value set in combination with the voltage direction change corresponding to the frequency peak;

[0140] S103: Based on the synchronization interval delay value set, the voltage and frequency change trends are sorted, the inflection point segment direction sequence is called, the structure arrangement type is identified, and the disturbance segment evolution structure is generated.

[0141] Optionally, the trend association pattern sequence generating module 220 is further used to:

[0142] S201: Obtain node data of each disturbance stage in the disturbance segmented evolution structure, extract the load change rate, voltage fluctuation amplitude and frequency offset direction of the node in the same stage, and judge the direction consistency state between the three parameters based on the change slope signs of the three parameters in the current time period. If the directions of the three slopes are inconsistent, the corresponding node and time period are recorded as abnormal segments, and an abnormal trend segment interval set is generated;

[0143] S202: calling the abnormal trend segment interval set to record the node abnormal segment time interval, locating the inflection point time of the load change rate curve in each abnormal segment, matching the starting position of the slope change before the inflection point as the trend starting point, summarizing the time range between the trend starting point and the corresponding inflection point, arranging the segment trend transfer relationship in chronological order, and generating a trend transfer structure sequence;

[0144] S203: According to the trend transfer structure sequence, the trend sequence combination of the three parameters in each segment is extracted, the differentiated combination sequence is sorted according to the time axis and the repeated segments with the same structure are merged, the number of occurrences and sequential positions of the trend repeated structure are counted, the trend structure arrangement sequence corresponding to the node is established, and the trend association pattern sequence is generated.

[0145] Optionally, the indicator response path feature identification module 230 is further used to:

[0146] S301: Call the trend sequence of the nodes in the trend association pattern sequence, extract the trend change starting point and trend retention end point corresponding to the adjacent time periods, mark the change location and retention interval according to the trend direction change position and the duration length in the adjacent segments, and generate a trend transformation structure information group;

[0147] S302: Calculate the change frequency index of the node in the trend direction change segment according to the trend direction change position and time interval recorded in the trend transformation structure information group, call the upper limit value of the node's index response range as a judgment threshold, classify the trend segment whose change frequency exceeds the threshold, and obtain a drift trend segment interval set;

[0148] S303: Based on the starting point and the ending point of the trend segment in the drift trend segment interval set, the boundary position of each segment is aggregated by node, the number of boundary segments corresponding to each node is counted and the corresponding time label is extracted, and the boundary distribution position is classified and sorted according to the time series to obtain the trend boundary distribution structure;

[0149] S304: Call the trend repetition range and residence time in the trend boundary distribution structure, perform combined classification based on the residence duration interval and the occurrence range of the trend segment, divide the control response status of the nodes according to the differential trend distribution characteristics, and generate the change category of the participation index.

[0150] Among them, the calculation formula for the change frequency index of the node within the trend direction change segment is as follows in Equation (1):

[0151] (1);

[0152] Among them, represents the node within the trend segment the trend change frequency index, represents the time length of the trend segment ; represents the node at the th sampling point, the slope of the frequency change direction, represents the node at the th sampling point, the slope of the voltage change direction, represents the node at the th sampling point, the slope of the load change direction, , , respectively represent the slope weight factors of the frequency, voltage, and load changes in the direction at the th sampling point, is the number of sampling points within the trend segment .

[0153] Optionally, the parameter coupling distribution type acquisition module 240 is further used for:

[0154] S401: Call the node classification result in the change category of the participation index, identify the node combinations in which the three parameters of the frequency drift rate, voltage fluctuation amplitude, and load change rate change simultaneously within the same time period among the nodes of the same category, extract the node combination list and the corresponding time index, and generate a synchronous change combination set;

[0155] S402: According to the node combinations recorded in the synchronous change combination set, calculate the change start time difference index between the three parameters of the nodes, judge the sequence of the change responses, extract the delay time difference and direction relationship between the parameters, screen the positive and negative cooperative segments according to the delay direction relationship, and mark the intersection time interval of the consistent changes of the three parameters in each group to obtain the cooperative response structure sequence;

[0156] S403: Invoke the combined collaboration records in the collaborative response structure sequence, count the occurrences of the same collaborative structure in the combination, filter out the combination relationships with a repetition frequency higher than the collaborative structure judgment threshold, sequentially summarize the delay paths within the collaborative chain in each type of combination, record the parameter change propagation path according to the node order, and generate the parameter coupling distribution type.

[0157] Among them, the calculation formula for the change start time difference index between the three parameters of the node is as follows in formula (2):

[0158] (2);

[0159] Among them, represents the average time difference between the start times of the changes in the three parameters of frequency, voltage, and load in the node combination , represents the start time of the change in the frequency parameter in the node combination , represents the start time of the change in the voltage parameter, represents the start time of the change in the load parameter.

[0160] Optionally, the multi-index linkage sorting rule generation module 250 is further used for:

[0161] S501: Based on the parameter coupling distribution type, extract the regulation response interval of the nodes in the same type during the disturbance process, calculate the length of the effective response time interval, compare the interval boundary values with the currently set response threshold range, determine whether there is an out-of-bounds situation at the boundary, and identify the interval direction change state to generate the response boundary matching result;

[0162] S502: According to the direction state recorded in the response boundary matching result, normalize the change amplitude value of the node, extract the influence degree score corresponding to each node during the disturbance process, sort them from high to low according to the score, and sequentially assign adjustment serial number identifiers to the sorted nodes to establish the node adjustment sorting result;

[0163] S503: Invoke the node serial numbers in the node adjustment sorting result, match the interval positions in the response boundary matching result, mark the boundary change direction of the sorted nodes within the self-response interval, extract the continuous paragraphs of the response boundary, and arrange the adjustment order according to the continuous interval priority to generate the multi-index linkage sorting rule.

[0164] Among them, the calculation formula for the length of the effective response time interval is as follows in formula (3):

[0165] (3);

[0166] Among them, represents the node Upper parameter The length of the effective response time interval of Indicates the starting time of the parameter response Indicates the termination time of the parameter response Indicates the node At the th sampling point, the change of the parameter is Indicates the sampling period Indicates the parameter The corresponding response judgment threshold

[0167] The present invention proposes a method for calculating safety assessment indicators of a power grid multi-agent large model. By analyzing the changing trends of node voltage and frequency in stages, combining trend consistency judgment and abnormal segment identification, dynamic structure reconstruction and trend mapping of the disturbance process are realized. The coherence and accuracy of index response state identification are improved through trend boundary aggregation and frequency classification. The coupling direction and delay analysis of multi-parameter changes between nodes enhance the identification of collaborative characteristics. The adjustment response sorting clarifies the node adjustment priority, constructs an adjustment path under multi-dimensional linkage, and improves the refined management and control ability of disturbance response and the pertinence of control decisions

[0168] Figure 3 FIG. is a schematic structural diagram of a safety assessment index calculation device provided by an embodiment of the present invention. As Figure 3 shown, the safety assessment index calculation device may include the above-mentioned Figure 2 shown power grid multi-agent large model safety assessment index calculation device. Optionally, the safety assessment index calculation device 310 may include a first processor 2001

[0169] Optionally, the safety assessment index calculation device 310 may further include a memory 2002 and a transceiver 2003

[0170] Among them, the first processor 2001, the memory 2002, and the transceiver 2003 may be connected through a communication bus

[0171] Next, in combination with Figure 3 Specific introductions will be made to the various components of the safety assessment index calculation device 310

[0172] Among them, the first processor 2001 is the control center of the security evaluation index calculation device 310, which can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 is one or more central processing units (CPUs), or can be an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. For example: one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs).

[0173] Optionally, the first processor 2001 can execute various functions of the security evaluation index calculation device 310 by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0174] In a specific implementation, as an embodiment, the first processor 2001 can include one or more CPUs, such as Figure 3 the CPU0 and CPU1 shown in

[0175] In a specific implementation, as an embodiment, the security evaluation index calculation device 310 can also include multiple processors, such as Figure 3 the first processor 2001 and the second processor 2004 shown in

[0176] Each of these processors can be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). Here, the processor can refer to one or more devices, circuits, and / or processing cores for processing data (such as computer program instructions).

[0177] Optionally, the memory 2002 may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or may also be an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently and be coupled to the first processor 2001 through an interface circuit ( Figure 3 not shown) of the security assessment metric calculation device 310. The embodiments of the present invention do not make specific limitations thereto.

[0178] The transceiver 2003 is used to communicate with a network device or with a terminal device.

[0179] Optionally, the transceiver 2003 may include a receiver and a transmitter ( Figure 3 not shown separately). Among them, the receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function.

[0180] Optionally, the transceiver 2003 may be integrated with the first processor 2001 or may exist independently and be coupled to the first processor 2001 through an interface circuit ( Figure 3 not shown) of the security assessment metric calculation device 310. The embodiments of the present invention do not make specific limitations thereto.

[0181] It should be noted that Figure 3 the structure of the security assessment metric calculation device 310 shown does not constitute a limitation to the router. The actual knowledge structure recognition device may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0182] In addition, the technical effects of the security assessment metric calculation device 310 may refer to the technical effects of the power grid multi-agent large model security assessment metric calculation method described in the above method embodiments, and will not be elaborated here.

[0183] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0184] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0185] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0186] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. Additionally, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood by referring to the context before and after.

[0187] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "At least one of the following" or its similar expressions refer to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, and c can be single or multiple.

[0188] It should be understood that in various embodiments of the present invention, the magnitudes of the sequence numbers of the above processes do not imply the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0189] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in connection with the embodiments disclosed herein can be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0190] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the devices, apparatuses, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0191] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be in an electrical, mechanical, or other form.

[0192] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0193] In addition, the functional units in various embodiments of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0194] When the above-mentioned function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0195] As described above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for calculating safety assessment indicators of a large multi-agent model of a power grid, characterized in that: The method comprises: S1: Obtain the voltage change rate and frequency change rate, divide the disturbance into four stages according to the disturbance start and extreme point, pair the curve trends, determine the delay and synchronization characteristics, extract the continuous change structure, and generate the disturbance segmented evolution structure; S2: According to the disturbance segment evolution structure, extract three types of trend parameters, judge the trend consistency, locate the abnormal segment and turning point, sort out the trend transfer path, and generate the trend association pattern sequence; S3: Call the trend association pattern sequence, extract the trend change segments of the node adjacent time periods, record the change frequency and determine whether it exceeds the indicator response range, perform boundary aggregation on the start and end positions of the drift segment, classify the node status based on the boundary distribution quantity and trend duration, identify the differentiated response path characteristics, and obtain the indicator participation change category; S4: Call the indicator participation change category, screen the three-parameter synchronization combination, identify the impact direction and response order, classify and combine according to the coordination frequency, and obtain the parameter coupling distribution type; S5: Based on the parameter coupling distribution type, extract the combined response interval, determine the boundary adjustment direction, sort the combination priority by impact, and generate multi-indicator linkage sorting rules.

2. The method for calculating the safety assessment index of a power grid multi-agent large model according to claim 1 is characterized in that: The disturbance segmented evolution structure includes voltage change trend characteristics, frequency change trend characteristics, intra-stage parameter response relations and structural stage arrangement patterns; The trend association pattern sequence includes abnormal trend identification points, trend transfer time series, node trend arrangement structure and trend direction consistency mark; The indicator participation change categories include trend change frequency statistics, trend drift boundary distribution, response duration classification and regulation indicator response type; The parameter coupling distribution types include parameter change coupling combinations, synergy relationship types, response delay characteristics, and propagation path patterns; The multi-indicator linkage sorting rules include a set of control response intervals, boundary matching priorities, node influence sorting, and adjustment operation records.

3. The method for calculating the safety assessment index of a power grid multi-agent large model according to claim 1 is characterized in that: The voltage change rate and frequency change rate are obtained, and four stages are divided according to the disturbance start and extreme point, and the curve trends are paired, and the delay and synchronization characteristics are determined, and the continuous change structure is extracted to generate the disturbance segmented evolution structure, including: S101: Obtain the voltage change rate and frequency change rate of the nodes in the power grid, segment the time axis according to the disturbance starting point and extreme point sequence, mark the sequence of four time intervals, and generate a stage division interval sequence; S102: extracting the voltage and frequency change rate curves according to the stage division interval sequence, determining the starting time difference between the rising segment and the falling segment of the two curves in the stage, selecting the segment with an offset value less than the disturbance detection threshold as the synchronization interval, and combining the voltage direction change corresponding to the frequency peak to obtain the synchronization interval delay value set; S103: Based on the synchronization interval delay value set, the voltage and frequency change trends are sorted, the inflection point segment direction sequence is called, the structure arrangement type is identified, and the disturbance segment evolution structure is generated.

4. The method for calculating the safety assessment index of a power grid multi-agent large model according to claim 1 is characterized in that: The method extracts three types of trend parameters based on the disturbance segment evolution structure, determines the trend consistency, locates the abnormal segment and turning point, sorts out the trend transfer path, and generates a trend association pattern sequence, including: S201: Obtain node data of each disturbance stage in the disturbance segmented evolution structure, extract the load change rate, voltage fluctuation amplitude and frequency offset direction of the node in the same stage, and judge the direction consistency state between the three parameters based on the change slope signs of the three parameters in the current time period. If the directions of the three slopes are inconsistent, the corresponding node and time period are recorded as abnormal segments, and an abnormal trend segment interval set is generated; S202: calling the abnormal trend segment interval set to record the node abnormal segment time interval, locating the inflection point time of the load change rate curve in each abnormal segment, matching the starting position of the slope change before the inflection point as the trend starting point, summarizing the time range between the trend starting point and the corresponding inflection point, arranging the segment trend transfer relationship in chronological order, and generating a trend transfer structure sequence; S203: According to the trend transfer structure sequence, the trend sequence combination of the three parameters in each segment is extracted, the differentiated combination sequence is sorted according to the time axis and the repeated segments with the same structure are merged, the number of occurrences and sequential positions of the trend repeated structure are counted, the trend structure arrangement sequence corresponding to the node is established, and the trend association pattern sequence is generated.

5. The method for calculating the safety assessment index of a power grid multi-agent large model according to claim 1 is characterized in that: The trend association pattern sequence is called, the trend change segments of the adjacent time periods of the nodes are extracted, the change frequency is recorded and it is determined whether it exceeds the indicator response range, the boundary collection is performed on the start and end positions of the drift segment, the node status is classified in combination with the boundary distribution quantity and the trend duration, and the differentiated response path characteristics are identified to obtain the indicator participation change category, including: S301: Call the trend sequence of the nodes in the trend association pattern sequence, extract the trend change starting point and trend retention end point corresponding to the adjacent time periods, mark the change location and retention interval according to the trend direction change position and the duration length in the adjacent segments, and generate a trend transformation structure information group; S302: Calculate the change frequency index of the node in the trend direction change segment according to the trend direction change position and time interval recorded in the trend transformation structure information group, call the upper limit value of the node's index response range as a judgment threshold, classify the trend segment whose change frequency exceeds the threshold, and obtain a drift trend segment interval set; S303: Based on the starting point and the ending point of the trend segment in the drift trend segment interval set, the boundary position of each segment is aggregated by node, the number of boundary segments corresponding to each node is counted and the corresponding time label is extracted, and the boundary distribution position is classified and sorted according to the time series to obtain the trend boundary distribution structure; S304: Call the trend repetition range and residence time in the trend boundary distribution structure, perform combined classification according to the residence time interval and the trend segment appearance range, divide the node's regulation response state according to the differentiated trend distribution characteristics, and generate the indicator participation change category.

6. The method for calculating the safety assessment index of a power grid multi-agent large model according to claim 5 is characterized in that: The calculation formula of the node change frequency index in the trend direction change segment is as follows (1): （1）； in, Representative Node In the trend segment The frequency of trend changes within Representing trend segment length of time, Representative Node In the The frequency change direction slope of the sampling point, Representative Node In the The voltage change direction slope of the sampling point, Representative Node In the The slope of the load change direction at each sampling point, , , Respectively represent The weight factor of the direction slope of frequency, voltage and load changes at each sampling point, For trend segment The number of sampling points within .

7. The method for calculating safety assessment index of a power grid multi-agent large model according to claim 1 is characterized in that: The calling indicator participates in the change category, screens the three-parameter synchronization combination, identifies the impact direction and response sequence, and obtains the parameter coupling distribution type according to the collaborative frequency classification combination, including: S401: calling the node classification results in the indicator participation change category, identifying the node combination in which the frequency drift rate, voltage fluctuation amplitude and load change rate of the same category of nodes change simultaneously in the same time period, extracting the node combination list and the corresponding time index, and generating a synchronous change combination set; S402: According to the node combination recorded in the synchronous change combination set, the change starting point time difference index between the three parameters of the node is calculated, the sequence of change response is determined, the delay time difference and direction relationship between the parameters are extracted, the forward and reverse coordination segments are selected according to the delay direction relationship, and the intersection time interval of the consistent change of the three parameters in each group is marked to obtain the coordinated response structure sequence; S403: Call the combined collaborative records in the collaborative response structure sequence, count the number of occurrences of the same collaborative structure in the combination, screen the combination relationship with a repetition frequency higher than the collaborative structure judgment threshold, sequentially summarize the delay paths in the collaborative chain in each type of combination, record the parameter change propagation path in node order, and generate the parameter coupling distribution type.

8. The method for calculating safety assessment index of a power grid multi-agent large model according to claim 7 is characterized in that: The calculation formula of the starting time difference index of the change of the three parameters of the node is as follows (2): （2）； in, Represents a node combination The average time difference between the start time of the change of the three parameters of medium frequency, voltage and load, Represents a node combination The starting time of the change of the medium frequency parameter, Indicates the starting time of the voltage parameter change, Indicates the starting time of load parameter change.

9. The method for calculating the safety assessment index of a power grid multi-agent large model according to claim 1 is characterized in that: The method of extracting the combined response interval based on the parameter coupling distribution type, determining the boundary adjustment direction, sorting the combined priority by the degree of influence, and generating the multi-indicator linkage sorting rules includes: S501: Based on the parameter coupling distribution type, extract the control response interval of the nodes of the same type in the disturbance process, calculate the effective response time interval length, and compare the interval boundary value with the currently set response threshold range to determine whether there is an excess of the boundary, identify the interval direction change state, and generate a response boundary matching result; S502: According to the direction state recorded in the response boundary matching result, the change amplitude value of the node is normalized, the impact degree score corresponding to each node in the disturbance process is extracted, and the scores are sorted from high to low, and the sorted nodes are assigned adjustment sequence numbers in turn to establish the node adjustment sorting result; S503: Call the node sequence number in the node adjustment sorting result, match the interval position in the response boundary matching result, mark the boundary change direction of the sorting node in the self-response interval, and extract the response boundary continuity paragraph, arrange the adjustment order according to the continuous interval priority, and generate a multi-indicator linkage sorting rule.

10. The method for calculating safety assessment index of a power grid multi-agent large model according to claim 9 is characterized in that: The calculation formula of the effective response time interval length is as follows (3): （3）； in, Representation Node Upper parameters The length of the effective response time interval, Indicates the starting time of the parameter response. Indicates the parameter response end time. Representation Node In the Parameters at sampling points The amount of change, represents the sampling period, Representation parameters The corresponding response judgment threshold.

Images