Dispatching operation ticket execution process monitoring method based on digital twinning technology

Abstract

The invention relates to a scheduling operation ticket execution process monitoring method based on a digital twinning technology, and belongs to the field of power system detection. Comprising the following steps: acquiring a step sequence and logic association from an operation order, analyzing to generate a data set, and constructing a logic dependency graph and a digital twinborn instance; analyzing the matching degree of the step precondition and the real-time state, identifying a failure step and generating alarm information; querying an equipment interaction record, and extracting a dynamic change factor list causing state conflicts; analyzing in combination with a logic dependency graph to obtain influence weights of subsequent steps, and identifying high-risk steps; calling digital twin examples to simulate equipment state evolution, and evaluating potential chain reaction risks; adjusting an operation order step sequence according to the protection device triggering probability, and generating a scheduling instruction set meeting a real-time state; monitoring deviation in real time and generating a dynamic correction instruction; and verifying the correction instruction and generating an operation order scheme. The intelligent level and safety of scheduling operation order execution are improved, and power grid accidents caused by misoperation are effectively prevented.

Description

Technical Field

[0001] This invention relates to a method for monitoring the execution process of dispatch operation tickets based on digital twin technology, belonging to the field of power system detection technology. Background Technology

[0002] Power system dispatching is a key link in ensuring the safe and stable operation of the power grid. Its importance lies in achieving the correct switching of equipment status through precise operation instructions, avoiding power grid failures or even large-scale power outages caused by misoperation. As the core tool of dispatching operations, the operation ticket carries the sequence and logical requirements of equipment operation.

[0003] However, existing operation ticket management methods rely heavily on manual experience and static checks, making it difficult to capture dynamic deviations during operations in real time. This is especially true in complex power grids, where the verification of the logical coherence of operation sequences remains insufficient. This leads to serious consequences when operators execute sequences such as "combining before splitting" or "splitting before combining" due to incorrect ordering.

[0004] The current challenge stems from insufficient dynamic monitoring of the logical connections between steps in the operation ticket. Each step in the operation ticket depends on the fulfillment of preconditions; for example, a device must be in a specific state to perform a closing or opening action. If the preconditions fail due to deviations in the operation sequence, the device state will deviate from the expected path. This deviation may further trigger the interlocking mechanism of the protection device, causing operational interruption or even a chain reaction.

[0005] What makes things more complicated is that there is a mapping gap between the real-time status changes of power grid equipment and the static logic of the operation ticket, and existing systems have difficulty dynamically identifying these logical conflicts through real-time data.

[0006] Therefore, how to establish a dynamic mapping between operation ticket steps and real-time equipment status through digital twin technology, and then identify logical conflicts caused by deviations in the operation sequence, has become a key issue in preventing chain reactions of misoperation. Summary of the Invention

[0007] Based on the problems described in the background, the problem to be solved by the present invention is to provide a method for monitoring the execution process of scheduling operation tickets based on digital twin technology, so as to solve the problems mentioned above.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a method for monitoring the execution process of scheduling operation tickets based on digital twin technology, comprising the following steps:

[0009] (1) Obtain the step sequence and logical association rules from the operation ticket, parse the preconditions and expected equipment status of each step, process and generate a dataset containing step numbers, dependencies and status parameters, generate a logical dependency graph based on the dataset, and construct a digital twin instance.

[0010] (2) Obtain the matching degree between the preconditions of the steps in the analysis logic dependency diagram and the real-time device status data, identify the preconditions of the failure steps based on the matching degree, and generate alarm information containing the failure step number and status deviation value.

[0011] (3) Based on the alarm information, query the device interaction records related to the failure steps in the digital twin instance, extract the external disturbances or operation records that cause the deviation, and obtain a list of dynamic change factors that cause state conflicts.

[0012] (4) Combine the logical dependency diagram to analyze the list of dynamic change factors, obtain the influence weight of the preconditions of subsequent steps, and identify high-risk steps based on the influence weight.

[0013] (5) Call the digital twin instance to simulate the evolution of equipment status in high-risk steps, generate an evolution path containing simulated status parameters and the trigger probability of protection devices, and obtain the risk assessment results of potential chain reactions;

[0014] (6) Adjust the logical dependency graph step sequence of the operation ticket according to the trigger probability of the protection device, generate the target operation ticket containing the reordering step and the update precondition, and obtain the scheduling instruction set that meets the real-time status.

[0015] (7) Monitor the deviation between equipment status data and scheduling instruction set in real time. If the deviation exceeds the preset threshold, generate a dynamic correction instruction containing the deviation step number and adjustment suggestions to obtain real-time updated operation guidance.

[0016] (8) Verify the dynamic correction instructions, simulate the execution of the adjusted step sequence, analyze whether the equipment status meets expectations, and generate an operation ticket execution plan.

[0017] Preferably, step (1) includes the following steps:

[0018] (1.1) Based on the step identifier code in the operation ticket, deep semantic extraction is used to obtain the execution condition parameters and device status markers. The execution condition parameters are semantically segmented using text mining methods to obtain a sequence of step dependency relationships.

[0019] (1.2) Construct a state transition matrix for the sequence of steps, wherein the row and column values ​​of the state transition matrix correspond to the step identifier code and the device status mark, and fill the matrix cells with logical dependency variable values ​​and device threshold constraint rules;

[0020] (1.3) The rule definition processor is used to parse the state transition matrix to obtain a logical dependency variable mapping table, and the decision tree algorithm is applied to the mapping table to generate a logical dependency graph;

[0021] (1.4) Extract the device state transition path based on the logical dependency graph, construct the device state prediction model for the state transition path using runtime indicators in the time series database, and obtain the digital twin instance by training the prediction model parameters through the random forest algorithm.

[0022] Preferably, step (2) includes the following steps:

[0023] (2.1) Based on the precondition threshold and monitoring parameters of the logical dependency graph, a data acquisition device is used to read real-time monitoring data from the field equipment, and the validity of the real-time monitoring data is judged by numerical verification rules to obtain the status monitoring dataset;

[0024] (2.2) For the aforementioned precondition threshold and state monitoring dataset, a data standardization processor is used to perform normalization operations, and the condition matching degree is obtained by calculating the state deviation value between the two sets of normalized data through Manhattan distance.

[0025] (2.3) The condition matching degree is classified according to the preset matching degree benchmark value. If the condition matching degree is less than the preset matching degree benchmark value, a support vector machine is used to perform clustering operation on the matching degree data of different levels, and a step failure mark is generated for the clustering results.

[0026] (2.4) For the failure marker of the step, the failure step number and the state deviation value are extracted by the failure marker parser, the failure degree score is calculated by the state evaluation function, and an alarm message containing the failure step number and the state deviation value is generated based on the failure degree score.

[0027] Preferably, step (3) includes the following steps:

[0028] (3.1) Obtain the device interaction sequence from the digital twin instance based on the failure step number in the alarm information, and traverse the state change records before and after the alarm timestamp in the device interaction sequence through the time sequence association algorithm to obtain the state abnormal interval.

[0029] (3.2) For the abnormal state interval, the numerical range of the state change records is verified, and a disturbance propagation link graph is constructed by a depth-first search algorithm;

[0030] (3.3) Extract the sequence of interaction events between devices based on the disturbance propagation link diagram, calculate the event time correlation coefficient, and if the correlation coefficient is greater than a preset threshold, generate a causal association chain;

[0031] (3.4) Perform state change analysis on the causal chain, calculate the state transition probability of each node through Bayesian network, identify key change nodes in the state transition process, and generate a list of dynamic change factors that cause state conflicts.

[0032] Preferably, step (4) includes the following steps:

[0033] (4.1) Extract the set of step nodes from the logical dependency graph, construct a propagation path graph of change factors for the set of step nodes, and quantify the decay rate of change factors in the propagation path graph by a path tracker to obtain the influence propagation matrix;

[0034] (4.2) Extract the precondition group from the set of step nodes according to the influence propagation matrix, calculate the influence probability value of the changing factors on the precondition group through the Markov chain, and perform a weighted superposition of the influence probability value and the influence propagation matrix to obtain the condition influence weight.

[0035] (4.3) A risk propagation network is constructed using a recurrent neural network based on the conditional influence weights. The risk transmission coefficient of each node is calculated for the risk propagation network. A risk accumulation function is generated for the risk transmission coefficient using a risk evaluator to obtain the step risk score.

[0036] (4.4) The logical dependency strength of the step nodes is calculated by the step association analyzer, and the step risk score is associated with the logical dependency strength. The association weighting result is classified by the risk threshold discriminator to obtain the high-risk step number.

[0037] Preferably, step (5) includes the following steps:

[0038] (5.1) Obtain the device parameter curve from the digital twin instance based on the device identification number, obtain the state evolution dataset through the time-series scanning of the parameter curve, and use a state feature extractor to identify key state change points for the state evolution dataset.

[0039] (5.2) For the key state change points, the protection logic parser is used to extract the protection device group configuration information, the protection device trigger probability is calculated according to the configuration information, and a protection action sequence is generated based on the trigger probability value;

[0040] (5.3) Extract the chain trigger source identifier based on the protection action sequence, establish a chain reaction model for the trigger source identifier using a deep reinforcement learning algorithm, and obtain the state evolution prediction value through the chain reaction model;

[0041] (5.4) Calculate the action delay parameter for the predicted state evolution value, construct a chain reaction propagation diagram based on the action delay parameter, obtain the risk propagation intensity through the chain reaction propagation diagram, and obtain the risk assessment result of the potential chain reaction.

[0042] Preferably, step (6) includes the following steps:

[0043] (6.1) Extract the step sequence nodes based on the trigger probability of the protection device in the logical dependency graph, and use the topological sorting algorithm to analyze the dependency relationship between the nodes to obtain the dependency relationship data;

[0044] (6.2) Based on the dependency data, analyze the equipment operating parameters obtained by the real-time status monitor through the risk assessor, and generate a risk level sequence for the equipment operating parameters;

[0045] (6.3) Reorder the step sequence in the time-series execution matrix according to the risk level sequence, and obtain the execution order table through the sequence validator based on the reordering result;

[0046] (6.4) According to the execution order table, call the operation ticket template generator to extract standard instruction items, and use the instruction optimizer to optimize the standard instruction items to obtain a scheduling instruction set that meets the real-time status.

[0047] (6.5) Extract the operation node relationship from the logical dependency graph, readjust the operation step priority in combination with the protection device trigger probability, construct the execution time window parameter table between steps, define the device interaction safety boundary value according to the time window parameter table, compile the operation ticket execution instruction template, map the device action sequence and operator response specifications, embed real-time state adaptation rules, generate a dynamic scheduling instruction set containing multi-branch execution paths, and attach the expected value of device response and fault tolerance range.

[0048] Preferably, step (6.5) includes the following steps:

[0049] (6.5.1) Obtain the operation node association relationship based on the logical dependency graph, construct the node relationship matrix using the association metric algorithm, and generate association metric values ​​for the node relationship matrix;

[0050] (6.5.2) The priority parameter is obtained by normalizing the correlation quantified value using a priority adjuster, and the time window parameter table is obtained by predicting the time window parameter table using a recurrent neural network based on the priority parameter.

[0051] (6.5.3) Extract the boundary values ​​of device state changes according to the time window parameter table, process the boundary values ​​with a boundary value calculator to obtain the device interaction safety threshold, and construct execution constraints for the safety threshold;

[0052] (6.5.4) Construct conditional branch judgment logic for the execution constraints, use an instruction template parser to process the judgment logic to obtain execution instructions, mark the expected value of the device response for the execution instructions, use a fault tolerance calculator to perform interval calculation on the expected value of the device response, generate fault tolerance range parameters based on historical data statistics, and generate a dynamic scheduling instruction set containing multi-branch execution paths.

[0053] Preferably, step (7) includes the following steps:

[0054] (7.1) A data acquisition device is used to acquire equipment status monitoring data, and the difference between the monitoring data and the execution requirement parameters in the scheduling instruction set is calculated. The difference data exceeding the safety threshold range is obtained through a threshold comparator.

[0055] (7.2) Based on the difference data, a deviation analyzer is used to extract key feature parameters, and a long short-term memory network is used to obtain the temporal correlation data of the feature parameters;

[0056] (7.3) For the time-series correlation data, query the adjustment scheme in the adjustment rule base, and obtain the correction instruction content through the rule matcher;

[0057] (7.4) Based on the content of the correction instruction, the random forest algorithm is used to calculate the device response curve, the expected response range of the correction scheme is obtained through the response prediction model, the original scheduling instruction set is updated, and a new operation guidance sequence is generated by the instruction synthesizer to obtain real-time updated operation guidance.

[0058] Preferably, step (8) includes the following steps:

[0059] (8.1) Obtain the standard operation requirement parameters in the scheduling instruction set, and calculate the deviation values ​​of the three key indicators of operating voltage, operating current and energy storage voltage according to the standard operation requirement parameters;

[0060] (8.2) A long short-term memory network is used to perform sliding time window analysis on the operating voltage deviation value, and a voltage compensation scheme is matched from the rule base based on the analysis results;

[0061] (8.3) Generate backup power start command, operation power switching command and energy storage voltage verification command according to the voltage compensation scheme, and determine the execution timing interval based on the equipment action characteristic curve;

[0062] (8.4) The random forest algorithm is used to process the historical correction data to obtain the prediction results of the change trend of operating voltage and energy storage voltage. The prediction results are used to generate the voltage parameter response range and generate the operation ticket execution plan.

[0063] The beneficial effects of this invention are:

[0064] 1. By analyzing the operation ticket step sequence and logical association rules, a digital twin model and logical dependency graph are constructed, realizing the dynamic mapping between operation ticket logic and real-time equipment status.

[0065] 2. Real-time analysis of the matching degree between equipment status and preconditions of steps can accurately identify failed steps and generate alarm information, and promptly detect potential operational risks.

[0066] 3. By combining dynamic change factor analysis, high-risk steps can be effectively identified and the evolution of the state can be simulated using digital twins to accurately assess the risk of potential chain reactions.

[0067] 4. Dynamically adjust the sequence of operation tickets and preconditions based on risk assessment results to generate a scheduling instruction set that adapts to real-time conditions, significantly improving the adaptability and security of operation tickets.

[0068] 5. Real-time monitoring of deviations and provision of dynamic correction instructions, combined with digital twin verification and correction schemes, ensure the safety and reliability of the operation ticket execution process and effectively prevent power grid accidents caused by misoperation.

[0069] 6. It has enabled intelligent optimization and dynamic adjustment of power system operation tickets, improving the efficiency of dispatching operations and the stability of power grid operation. Attached Figure Description

[0070] Figure 1 This is a schematic diagram of the method flow of the present invention; Detailed Implementation

[0071] The embodiments of the present invention will be further described below with reference to the accompanying drawings:

[0072] Example 1

[0073] like Figure 1 As shown, this invention provides a method for monitoring the execution process of scheduling operation tickets based on digital twin technology, comprising the following steps:

[0074] (1) Obtain the step sequence and logical association rules from the operation ticket, parse the preconditions and expected equipment status of each step, process and generate a dataset containing step numbers, dependencies and status parameters, generate a logical dependency graph based on the dataset, and construct a digital twin instance.

[0075] Step (1) includes the following steps:

[0076] (1.1) Based on the step identifier code in the operation ticket, deep semantic extraction is used to obtain the execution condition parameters and device status markers. The execution condition parameters are semantically segmented using text mining methods to obtain a sequence of step dependency relationships.

[0077] The step dependency sequence is constructed for the extracted data, and the validation rules for each step are mapped by the rule definition processor;

[0078] (1.2) Construct a state transition matrix for the sequence of steps, wherein the row and column values ​​of the state transition matrix correspond to the step identifier code and the device status mark, and fill the matrix cells with logical dependency variable values ​​and device threshold constraint rules;

[0079] (1.3) The rule definition processor is used to parse the state transition matrix to obtain a logical dependency variable mapping table, and the decision tree algorithm is applied to the mapping table to generate a logical dependency graph;

[0080] A digital process mapping table can also be established based on resource allocation codes;

[0081] (1.4) Extract the device state transition path based on the logical dependency graph, construct the device state prediction model for the state transition path using runtime indicators in the time series database, and obtain the digital twin instance by training the prediction model parameters through the random forest algorithm.

[0082] Before extracting the equipment state transition path, the equipment runtime state indicators are obtained from the step verification rules. For the runtime state indicators, a time-series database is used to store the state values ​​corresponding to each digital process. The standard deviation of the runtime state indicator data is calculated and normalized using the least squares method. After extracting the equipment state transition path, the digital process is time-series verified using a state transition rule library. A set of equipment operation rule constraints is generated based on the verification results. The digital twin instance is constructed based on the equipment state prediction model and the rule constraint set, and includes the equipment identification code, state parameters, constraint rules, and prediction model.

[0083] In step (1), the monitoring function can be triggered through the user interface of the scheduling system. The specific triggering method can be determined according to the actual scenario and is not subject to too many restrictions. After the monitoring function is started, the system first extracts the step sequence and logical association rules from the operation ticket.

[0084] In step (1.1), deep semantic extraction, based on natural language processing technology, performs word segmentation and semantic segmentation on the operation ticket text, and extracts step identifier codes such as BKR. 001 To BKR 010 The sequence data includes the corresponding execution condition parameters, device status markers including switch position or operating mode, and the sequence data records the logical dependencies between steps.

[0085] In step (1.2), the state transition matrix is ​​arranged with step identifier codes as rows and device status markers as columns. Each cell records logically dependent variables such as voltage and current, as well as threshold constraints. The matrix can intuitively reflect the logical relationships between steps and the device status requirements, providing a basis for subsequent analysis.

[0086] In step (1.3), the rule definition processor performs logical parsing on the matrix to generate a mapping table containing step dependencies. The decision tree algorithm constructs a logical dependency graph based on the mapping table, where nodes represent operation steps and edges represent dependencies between steps. The logical dependency graph clearly shows the execution logic of the operation ticket.

[0087] In step (1.4), the equipment state transition path records the sequence of equipment state changes during operation. Runtime indicators include parameters such as voltage, current, and power, stored in a time-series database at 10-millisecond intervals. The equipment state prediction model is trained using a random forest algorithm to predict future equipment states based on historical operation data. The digital twin instance contains equipment identification codes, state parameters, and constraint rules, reflecting the equipment's operating status in real time.

[0088] Through step (1) of this invention, the scheduling system can extract key information from the operation ticket, generate a logical dependency graph and a digital twin model, and provide support for subsequent dynamic monitoring. The digital twin model reflects the voltage and current changes of the circuit breaker in real time, ensuring the safety of the operation process.

[0089] The application of digital twin technology has significantly improved the monitoring capabilities of the operation ticket execution process. Through logical dependency graphs and digital twin models, it is possible to dynamically map operation steps with equipment status, identify potential logical conflicts and status deviations in advance, and provide higher safety assurance for power grid operations.

[0090] (2) Obtain the matching degree between the preconditions of the steps in the analysis logic dependency diagram and the real-time device status data, identify the preconditions of the failure steps based on the matching degree, and generate alarm information containing the failure step number and status deviation value.

[0091] Step (2) includes the following steps:

[0092] (2.1) Based on the precondition threshold and monitoring parameters of the logical dependency graph, a data acquisition device is used to read real-time monitoring data from the field equipment, and the validity of the real-time monitoring data is judged by numerical verification rules to obtain the status monitoring dataset;

[0093] (2.2) For the aforementioned precondition threshold and state monitoring dataset, a data standardization processor is used to perform normalization operations, and the condition matching degree is obtained by calculating the state deviation value between the two sets of normalized data through Manhattan distance.

[0094] (2.3) The condition matching degree is classified according to the preset matching degree benchmark value. If the condition matching degree is less than the preset matching degree benchmark value, a support vector machine is used to perform clustering operation on the matching degree data of different levels, and a step failure mark is generated for the clustering results.

[0095] (2.4) For the failure marker of the step, the failure step number and the state deviation value are extracted by the failure marker parser, the failure degree score is calculated by the state evaluation function, and an alarm message containing the failure step number and the state deviation value is generated based on the failure degree score.

[0096] Specifically, the failure markers of the steps are graded according to the failure severity score. An alarm level determiner is used to generate corresponding alarm levels, and an alarm information dataset is constructed based on the alarm levels and state deviation values. The failure step number and state deviation value are extracted from the alarm information dataset, and an alarm information generator is used to generate alarm information content according to a preset format. A unique identifier is generated according to alarm coding rules. An alarm record storage table is constructed based on the generated alarm information content and identifier. The alarm records are encoded and stored using a data compression algorithm, and an indexing mechanism is used to establish the correspondence between failure step numbers and alarm records.

[0097] In step (2), to address the dynamic monitoring needs during the execution of the operation ticket, the scheduling system compares the preconditions with the real-time equipment status using a logical dependency graph, calculates the matching degree to identify failed steps, and generates structured alarm information. This step can promptly detect state deviations during the operation process, providing a basis for risk assessment and dynamic adjustment.

[0098] In step (2.1), the precondition thresholds include parameters such as voltage, current, or switch position. The data acquisition module reads real-time data from the field equipment at 100-millisecond intervals. Validity verification removes abnormal data through preset rules to ensure the reliability of the status monitoring dataset. The generated monitoring dataset contains timestamps, device identifiers, and status parameters, providing a foundation for subsequent analysis.

[0099] In step (2.2), the normalization process divides the voltage value by 242 kV and the current value by 2200 amperes to unify the dimensions. The Manhattan distance is calculated as the sum of the absolute differences between the two sets of normalized values. The condition matching degree is mapped according to the deviation value; a deviation value of 0 to 0.05 corresponds to a matching degree of over 90%, and 0.05 to 0.10 corresponds to 80% to 90%. The matching degree reflects the closeness between the real-time state and the preconditions, providing a quantitative basis for failure judgment.

[0100] In step (2.3), the preset matching degree threshold is set to 85%, and steps below this value are considered as failures of the preconditions. The support vector machine algorithm maps the matching degree data to a high-dimensional space through a kernel function, and cluster analysis identifies abnormal steps. The failure marker includes the step number, such as BKR. 003 The deviation value was 0.12, and the specific outlier was recorded for subsequent processing.

[0101] In step (2.4), the state evaluation function calculates the failure severity score by combining the deviation value and the step importance. Alarm levels are determined based on the score: scores below 90 and above 75 are classified as Level 3 alarms, and scores below 75 and above 60 are classified as Level 2 alarms. The alarm information dataset is recorded in the format "Step Number - Deviation Value - Alarm Level", such as "BKR". 003 "-0.12-2" indicates a level 2 alarm. The unique identifier is generated using hash encoding to ensure alarm traceability. Alarm information is stored using a compression algorithm and retrieved quickly using a B-tree index.

[0102] Furthermore, the alarm log table stores alarm content and identification codes, and uses Huffman coding to compress the data, reducing storage space. The index relationship is implemented using a B-tree structure, allowing for rapid location of failure steps such as BKR. 007 The corresponding alarm records. The indexing mechanism allows operators to query anomaly information in real time, improving response efficiency.

[0103] Through step (2) of this invention, and based on the matching degree analysis and alarm generation mechanism, the system can promptly identify state deviations during the execution of operation tickets. For example, in a transformer parallel operation scenario, the system detects that step TR005 generates a first-level alarm "TR005-0.18-1" due to a phase angle deviation of 0.18, prompting the operator to adjust the equipment status. This method significantly improves the dynamic monitoring capability of power grid operation and reduces the risk of misoperation.

[0104] Meanwhile, the embodiments of this application do not impose too many restrictions on specific algorithm parameters or threshold settings. For example, Manhattan distance can be replaced with Euclidean distance, and the kernel function of support vector machine can be optimized according to the scenario to adapt to the needs of different power grid environments.

[0105] (3) Based on the alarm information, query the device interaction records related to the failure steps in the digital twin instance, extract the external disturbances or operation records that cause the deviation, and obtain a list of dynamic change factors that cause state conflicts.

[0106] Step (3) includes the following steps:

[0107] (3.1) Obtain the device interaction sequence from the digital twin instance based on the failure step number in the alarm information, and traverse the state change records before and after the alarm timestamp in the device interaction sequence through the time sequence association algorithm to obtain the state abnormal interval.

[0108] Abnormal state intervals are identified by an abnormal state detector, and the start time of the abnormal state is also identified.

[0109] (3.2) For the abnormal state interval, the numerical range of the state change records is verified, and a disturbance propagation link graph is constructed by a depth-first search algorithm;

[0110] A data validity validator is used to perform numerical range verification on the status change records. The numerical range verification extracts the device identification code and the associated device list for valid records.

[0111] (3.3) Extract the sequence of interaction events between devices based on the disturbance propagation link diagram, calculate the event time correlation coefficient, and if the correlation coefficient is greater than a preset threshold, generate a causal association chain;

[0112] The event time-series correlation coefficient was calculated using an association rule mining algorithm;

[0113] (3.4) Perform state change analysis on the causal chain, calculate the state transition probability of each node through Bayesian network, identify key change nodes in the state transition process, and generate a list of dynamic change factors that cause state conflicts.

[0114] Specifically, the state change analysis employs a state evolution analyzer. After identifying key change nodes, external disturbance events and equipment operation records are extracted based on these nodes. A spatiotemporal feature clustering algorithm is used to classify the events, and the disturbance source attributes and propagation paths are labeled based on the classification results. A conflict identification processor is used to extract features from the disturbance source attributes, and conflict type labels are generated based on feature similarity calculations. A state conflict feature library is constructed based on the labeling results. Dynamic change factors are categorized based on the state conflict feature library, and a pattern matching algorithm is used to identify typical conflict patterns. A list of dynamic change factors is generated based on the identification results.

[0115] In step (3), for any failed steps discovered during the execution of the operation ticket, the scheduling system uses a digital twin model to trace back the equipment interaction sequence, identify the dynamic factors that caused the state deviation, and generate a structured list of changing factors. This step accurately locates the source of the anomaly through time series analysis and causal deduction, providing a crucial basis for subsequent risk prevention and control.

[0116] In step (3.1), the device interaction sequence records the state changes of the device involved in the failure step. The time-series correlation algorithm, centered on the alarm time, traverses the monitoring records for one hour before and after the alarm time to calculate the trend of state parameter changes. The determination of the abnormal interval provides a precise time window for subsequent analysis and reduces interference from irrelevant data.

[0117] In step (3.2), the data validity validator checks whether the state parameters are within a reasonable range and removes outliers to ensure the accuracy of the analysis. The depth-first search algorithm starts from the failed device, traces the state changes of related devices and voltage transformers, and constructs a disturbance propagation link graph. Nodes in the link graph represent devices, edges represent state influence relationships, and the sequence of interaction events records the specific changes.

[0118] In step (3.3), the association rule mining algorithm calculates the temporal correlation between events, the Bayesian network constructs a conditional probability model based on historical data, and key change nodes are identified as the main trigger points of anomalies, providing focus for subsequent source tracing.

[0119] In step (3.4), external disturbances include voltage fluctuations caused by bus energization switching, and operational records include mechanical vibrations from the charging of the circuit breaker's energy storage springs. A spatiotemporal feature clustering algorithm categorizes events into electrical, mechanical, and control types, with electrical disturbance paths extending to the circuit breaker. The disturbance source attribute is labeled "voltage fluctuation," and the propagation path records the influence chain from the bus to the circuit breaker. A state conflict feature library stores conflict types, reflecting the contradiction between bus switching and circuit breaker voltage requirements.

[0120] Meanwhile, the pattern matching algorithm compares the conflict patterns in the feature library to identify typical patterns of secondary circuit fluctuations caused by primary equipment switching. The dynamic change factor list records the disturbance source, propagation path, and influencing factors, providing operators with clear information on anomaly tracing.

[0121] Through step (3) of this invention, the root cause of the state deviation was located through multi-level analysis. For example, in the transformer TR 005 In alarm scenarios, the electromagnetic disturbances generated by the opening and closing of adjacent circuit breakers are identified and propagated through the busbar, causing distortion of the transformer excitation current. A list of changing factors is generated, including "electromagnetic transients" and "busbar resonance amplification." This method significantly improves the accuracy and efficiency of anomaly tracing.

[0122] (4) Combine the logical dependency diagram to analyze the list of dynamic change factors, obtain the influence weight of the preconditions of subsequent steps, and identify high-risk steps based on the influence weight.

[0123] Step (4) includes the following steps:

[0124] (4.1) Extract the set of step nodes from the logical dependency graph, construct a propagation path graph of change factors for the set of step nodes, and quantify the decay rate of change factors in the propagation path graph by a path tracker to obtain the influence propagation matrix;

[0125] The set of step nodes is extracted from the logical dependency graph based on the list of dynamic changing factors; a forward inference algorithm is used to construct a propagation path graph of the changing factors;

[0126] (4.2) Extract the precondition group from the set of step nodes according to the influence propagation matrix, calculate the influence probability value of the changing factors on the precondition group through the Markov chain, and perform a weighted superposition of the influence probability value and the influence propagation matrix to obtain the condition influence weight.

[0127] The precondition set is extracted using a condition constraint parser; the condition influence weights are generated using a normalization processor.

[0128] (4.3) A risk propagation network is constructed using a recurrent neural network based on the conditional influence weights. The risk transmission coefficient of each node is calculated for the risk propagation network. A risk accumulation function is generated for the risk transmission coefficient using a risk evaluator to obtain the step risk score.

[0129] (4.4) The step association analyzer is used to calculate the logical dependency strength of the step nodes, and the step risk score is associated with the logical dependency strength. The association weighting result is classified by the risk threshold discriminator to obtain the high-risk step number.

[0130] Specifically, after correlation and weighting, a comprehensive risk index is generated based on the weighted results. The comprehensive risk index is then classified according to a preset risk benchmark value. A risk threshold discriminator is used to identify high-risk steps, and the high-risk step numbers are extracted from the identification results. A risk verification processor is used to cross-validate the high-risk steps, and the risk prediction accuracy is calculated based on historical risk data. Risk warning information is generated based on the verification results. A risk warning knowledge base is constructed based on the risk warning information, and a pattern recognition algorithm is used to extract risk feature vectors. A risk tracing index table is then established based on these feature vectors.

[0131] In step (4), the potential impact of dynamically changing factors during the execution of the operation ticket is addressed by quantifying the propagation effect of these factors on subsequent steps using a logical dependency graph and inference algorithm. High-risk operation steps are identified, and structured risk warning information is generated. This step, through multi-level analysis and verification, ensures the accuracy and operability of risk identification, providing a guarantee for the safe operation of the power grid.

[0132] In step (4.1), the set of step nodes includes key steps in the operation ticket, such as insulation checks and opening / closing operations in circuit breaker operation. The forward inference algorithm starts from dynamically changing factors, such as voltage fluctuations, and traces their impact path on subsequent steps. The influence propagation matrix has step nodes as rows and changing factors as columns, with each cell recording the attenuation coefficient. The matrix provides the basis for subsequent quantitative analysis.

[0133] In step (4.2), the preconditions include parameters such as an operating voltage of 220 kV and a current of 2000 amperes. A Markov chain is used to construct a state transition model based on historical data, calculating the probability of voltage fluctuations affecting the operating voltage to be 0.85 and the probability of them affecting the energy storage motor voltage to be 0.72. Combining the attenuation coefficient of the influence propagation matrix, the weights of the conditions are calculated through weighted superposition. These weights reflect the relative importance of the changing factors to the preconditions, providing a quantitative basis for risk assessment.

[0134] In step (4.3), the risk propagation network uses step nodes as nodes and weights as edges, and employs a recurrent neural network to analyze the transmission characteristics of risk between steps. The risk accumulation function calculates the comprehensive risk score for each step using an exponential model. The score reflects the severity of the impact of changing factors on the step, providing a basis for identifying high-risk steps.

[0135] In step (4.4), the logical dependency strength is calculated based on the execution order and conditional dependencies between steps. After weighting the risk scores, a comprehensive risk index is generated, and the risk threshold discriminator extracts its number for subsequent processing. The introduction of dependency strength improves the accuracy of risk assessment.

[0136] Through step (4) of this invention, the scheduling system performs cross-validation on high-risk steps, calculates the prediction accuracy using historical risk data, and uses the validation results to generate risk warning information, which is stored in the risk warning knowledge base. The knowledge base extracts risk feature vectors, such as voltage deviation and propagation path, using pattern recognition algorithms to construct a risk tracing index table. This method significantly improves the reliability of risk identification and the safety of power grid operation through multi-dimensional analysis and validation.

[0137] (5) Call the digital twin instance to simulate the evolution of equipment status in high-risk steps, generate an evolution path containing simulated status parameters and the trigger probability of protection devices, and obtain the risk assessment results of potential chain reactions;

[0138] Step (5) includes the following steps:

[0139] (5.1) Obtain the device parameter curve from the digital twin instance based on the device identification number, obtain the state evolution dataset through the time-series scanning of the parameter curve, and use a state feature extractor to identify key state change points for the state evolution dataset.

[0140] A state trajectory tracker is used to perform a time-series scan of the parameter curves;

[0141] (5.2) For the key state change points, the protection logic parser is used to extract the protection device group configuration information, the protection device trigger probability is calculated according to the configuration information, and a protection action sequence is generated based on the trigger probability value;

[0142] The trigger probability of the protection device is calculated using the Monte Carlo method;

[0143] (5.3) Extract the chain trigger source identifier based on the protection action sequence, establish a chain reaction model for the trigger source identifier using a deep reinforcement learning algorithm, and obtain the state evolution prediction value through the chain reaction model;

[0144] And compare the actual state parameters using a predictive validator;

[0145] (5.4) Calculate the action delay parameter for the predicted state evolution value, construct a chain reaction propagation diagram based on the action delay parameter, obtain the risk propagation intensity through the chain reaction propagation diagram, and obtain the risk assessment result of the potential chain reaction;

[0146] Specifically, a state trajectory analyzer is used to segment the verified state evolution data, identify key evolution nodes based on preset protection action thresholds, and calculate action delay parameters for each node. A chain reaction propagation diagram is constructed based on these delay parameters, and a time-series correlation analyzer is used to extract propagation path features. The risk propagation intensity is then calculated based on these features. A risk quantification processor is used to classify and calculate the risk propagation intensity, generating risk level identifiers based on preset risk benchmarks, and constructing an assessment dataset for each risk level. Key risk propagation paths are extracted from the assessment dataset, and a risk warning generator is used to construct warning information content. A risk response knowledge base is then established based on this warning information.

[0147] In step (5), for high-risk steps identified during the execution of the operation ticket, a digital twin model is used to simulate the dynamic changes in equipment status, analyze the trigger probability of the protection device and the propagation path of the chain reaction, and generate risk assessment results. This step reveals the chain effects that high-risk steps may cause through multi-dimensional simulation and analysis, providing accurate risk warning and prevention basis for power grid operation.

[0148] In step (5.1), the equipment parameter curves include indicators such as voltage, current and mechanical characteristics. The timing scan analyzes the curve changes at 10-millisecond intervals. The state feature extractor identifies key change points based on gradient analysis. The state evolution dataset records timestamps, parameter values ​​and equipment identifiers, providing a data foundation for subsequent protection logic analysis.

[0149] In step (5.2), the protection device group includes voltage protection, overcurrent protection, etc. The configuration information specifies the action thresholds. The Monte Carlo method simulates parameter fluctuations through 1000 random samples, calculating the voltage protection trigger probability to be 0.85 and the overcurrent protection probability to be 0.35. The protection action sequence shows that the voltage protection triggers first at 5.2 seconds, causing the circuit breaker to trip. Chain reaction trigger source analysis identifies the voltage protection action as the primary trigger point, initiating a chain reaction in adjacent equipment such as the busbar.

[0150] In step (5.3), the deep reinforcement learning algorithm optimizes the chain reaction model through a reward mechanism, predicting that after the circuit breaker trips, the bus voltage rises to 235 kV, and the overvoltage protection trigger probability of adjacent circuit breakers is 0.92. The deviation between the predicted value and the actual parameters is less than 3%, verifying the reliability of the model. The state trajectory analyzer divides the evolution process into three nodes: voltage change, protection activation, and circuit breaker action, calculates the action delay, and the chain reaction propagation graph uses nodes to represent equipment and edges to represent action propagation, reflecting the path of fault propagation from the circuit breaker to the bus.

[0151] In step (5.4), the time series correlation analyzer extracts the features of the propagation path, the risk propagation intensity is calculated based on the influence weight of the path nodes, the risk quantification processor divides the intensity into three levels: high, medium and low based on a preset benchmark value, and the risk assessment dataset records the propagation path, intensity and level, providing structured data for the generation of early warning information.

[0152] Through step (5) of this invention, the scheduling system generates early warning information based on the evaluation dataset and stores it in the risk response knowledge base. Early warning information is prioritized for high-risk paths, and the knowledge base records the characteristics of protection actions and corresponding countermeasures. This method significantly improves the accuracy of cascading reaction prediction and the reliability of power grid operation through dynamic simulation and risk quantification.

[0153] (6) Adjust the logical dependency graph step sequence of the operation ticket according to the trigger probability of the protection device, generate the target operation ticket containing the reordering step and the update precondition, and obtain the scheduling instruction set that meets the real-time status.

[0154] Step (6) includes the following steps:

[0155] (6.1) Extract the step sequence nodes based on the trigger probability of the protection device in the logical dependency graph, and use the topological sorting algorithm to analyze the dependency relationship between the nodes to obtain the dependency relationship data;

[0156] (6.2) Based on the dependency data, analyze the equipment operating parameters obtained by the real-time status monitor through the risk assessor, and generate a risk level sequence for the equipment operating parameters;

[0157] Nodes whose trigger probability exceeds the safety threshold are marked with a high-risk identifier, and a risk level sequence is generated using a risk assessor.

[0158] (6.3) Reorder the step sequence in the time-series execution matrix according to the risk level sequence, and obtain the execution order table through the sequence validator based on the reordering result;

[0159] For the adjusted sequence, extract the precondition group, update the constraint parameters through the condition optimizer, and construct the time-series execution matrix based on the updated constraint parameters. After reordering, generate a new execution order table based on the reordering result, and verify the rationality of the execution order through the sequence validator.

[0160] (6.4) According to the execution order table, call the operation ticket template generator to extract standard instruction items, and use the instruction optimizer to optimize the standard instruction items to obtain a scheduling instruction set that meets the real-time status.

[0161] Specifically, before extracting standard instruction items, a constraint checker is used to verify the preconditions in the execution sequence table in real time, and the condition threshold range is adjusted according to the verification results. An execution rule set is then generated for the adjusted constraint conditions.

[0162] Then, based on the execution rule set, the operation ticket template generator is invoked, and a template parsing algorithm is used to extract standard instruction items. A scheduling instruction sequence is then constructed for each instruction item. An instruction verification processor is used to verify the executability of the scheduling instruction sequence, evaluating the instruction execution conditions based on real-time status parameters, and generating an instruction execution list based on the evaluation results. A scheduling instruction set is then constructed based on the instruction execution list, and an instruction optimizer is used to sort the instruction execution order, generating execution sequence numbers based on the sorting results.

[0163] (6.5) Extract the operation node relationship from the logical dependency graph, readjust the operation step priority in combination with the protection device trigger probability, construct the execution time window parameter table between steps, define the device interaction safety boundary value according to the time window parameter table, compile the operation ticket execution instruction template, map the device action sequence and operator response specifications, embed real-time state adaptation rules, generate a dynamic scheduling instruction set containing multi-branch execution paths, and attach the expected value of device response and fault tolerance range.

[0164] Step (6.5) includes the following steps:

[0165] (6.5.1) Obtain the operation node association relationship based on the logical dependency graph, construct the node relationship matrix using the association metric algorithm, and generate association metric values ​​for the node relationship matrix;

[0166] Among them, the matrix elements are weighted according to the trigger probability of the protection device;

[0167] (6.5.2) The priority parameter is obtained by normalizing the correlation quantified value using a priority adjuster, and the time window parameter table is obtained by predicting the time window parameter table using a recurrent neural network based on the priority parameter.

[0168] Among them, the minimum execution interval between adjacent operation steps is predicted based on the recurrent neural network, and then a time window parameter table is constructed based on the prediction results;

[0169] (6.5.3) Extract the boundary values ​​of device state changes according to the time window parameter table, process the boundary values ​​with a boundary value calculator to obtain the device interaction safety threshold, and construct execution constraints for the safety threshold;

[0170] (6.5.4) Construct conditional branch judgment logic for the execution constraints, use an instruction template parser to process the judgment logic to obtain execution instructions, mark the expected value of the device response for the execution instructions, use a fault tolerance calculator to perform interval calculation on the expected value of the device response, generate fault tolerance range parameters based on historical data statistics, and generate a dynamic scheduling instruction set containing multi-branch execution paths.

[0171] Specifically, an instruction rule generator extracts state judgment rules from execution constraints, constructs branch execution paths based on device action sequences, and embeds a real-time state adaptation mechanism into these paths. Conditional branch judgment logic is built based on this state adaptation mechanism, and an instruction template parser generates corresponding execution instructions, marking the expected device response values ​​for each instruction. A fault tolerance calculator performs range calculations on the expected device response values, generates fault tolerance range parameters based on historical data statistics, and constructs instruction execution verification rules for these ranges. Feasibility verification of the execution instructions is performed based on these verification rules, and a multi-branch scheduling instruction set is constructed using a scheduling instruction generator, with an execution sequence number index table created for this instruction set.

[0172] In step (6), considering the potential impact of high-risk steps during the execution of the operation ticket, the dispatching system dynamically adjusts the step sequence and preconditions using the trigger probability of protection devices and real-time status data to generate an adaptive dispatching instruction set. This step, through multi-level analysis and optimization, ensures real-time matching between the operation ticket and the equipment operating status, reduces the risk of misoperation, and improves the efficiency of power grid dispatching.

[0173] In step (6.1), the step sequence nodes include insulation checks and tripping operations in circuit breaker operation. The topology sorting algorithm generates an ordered sequence based on dependencies, the protection device trigger probability is used for risk assessment, and the risk level sequence records the priority and risk level of each node, providing a quantitative basis for subsequent reordering.

[0174] In step (6.2), the real-time status monitor collects equipment parameters every 100 milliseconds. Based on the risk level sequence, the system adjusts the preconditions for high-risk nodes from an operating voltage of 220 kV to 205 kV, and the energy storage voltage from 220 V to 195 V. The timing execution matrix records the adjusted dependencies with step numbers as rows and execution order as columns, providing a structured representation for sequence optimization.

[0175] In step (6.3), the genetic algorithm iterates through 500 generations to optimize the step sequence, evaluating the safety and efficiency of the sequence based on the fitness function. The optimization result prioritizes energy storage checks, followed by insulation checks, with the tripping operation performed later. The sequence verifier checks whether the new sequence satisfies the dependencies, and the execution sequence table records the optimized step order to ensure the consistency of the operational logic.

[0176] In step (6.4), the operation ticket template generator extracts instruction items from the standard library, such as "check that the energy storage voltage is not lower than 195 volts" and "execute the tripping operation". The instruction optimizer adjusts the instruction priority according to the real-time status, and the scheduling instruction sequence undergoes executability verification to confirm that the instructions meet the current equipment status, such as the energy storage voltage of 200 volts meeting the 195 volt threshold. The verified sequence includes step numbers, conditions, and operation content, providing a basis for the generation of the scheduling instruction set.

[0177] Through step (6) of this invention, the scheduling system constructs an operation node relationship matrix based on a logical dependency graph, adjusts the step priority by combining the trigger probability of the protection device, and generates a time window parameter table to define the safety boundary of device interaction. The system uses a recurrent neural network to predict the execution interval and generate branch instructions that include status judgment and anomaly handling. The instruction set adopts a tree structure, with expected response values ​​such as a tripping time of 40 to 60 milliseconds. The fault tolerance range is based on historical data statistics to ensure safe and controllable operation. This method significantly improves the dynamic adaptability of operation tickets and ensures scheduling safety in complex power grid scenarios.

[0178] (7) Monitor the deviation between equipment status data and scheduling instruction set in real time. If the deviation exceeds the preset threshold, generate a dynamic correction instruction containing the deviation step number and adjustment suggestions to obtain real-time updated operation guidance.

[0179] Step (7) includes the following steps:

[0180] (7.1) A data acquisition device is used to acquire equipment status monitoring data, and the difference between the monitoring data and the execution requirement parameters in the scheduling instruction set is calculated. The difference data exceeding the safety threshold range is obtained through a threshold comparator.

[0181] (7.2) Based on the difference data, a deviation analyzer is used to extract key feature parameters, and a long short-term memory network is used to obtain the temporal correlation data of the feature parameters;

[0182] And identify the deviation steps for the correlation results;

[0183] (7.3) For the time-series correlation data, query the adjustment scheme in the adjustment rule base, obtain the correction instruction content through the rule matcher, and verify the rationality of the correction instruction through the instruction verifier;

[0184] (7.4) Based on the content of the correction instruction, the random forest algorithm is used to calculate the device response curve, the expected response range of the correction scheme is obtained through the response prediction model, the original scheduling instruction set is updated, and a new operation guidance sequence is generated by the instruction synthesizer to obtain real-time updated operation guidance.

[0185] Specifically, a timing analyzer is used to plan the execution timing of correction instructions, calculate the minimum execution interval between instructions based on the device's response characteristics, and generate an instruction timing table for each execution interval. A real-time correction scheme is constructed based on the instruction timing table, and a random forest algorithm is used to predict the device's response curve after correction, generating a response expectation range based on the prediction results. An effectiveness evaluator is used to verify the feasibility of the correction scheme, calculate the correction success rate based on historical data, and generate execution suggestions based on the verification results. The original scheduling instruction set is updated based on the execution suggestions, and a new operation guidance sequence is generated using an instruction synthesizer. A real-time monitoring and feedback mechanism is established for the guidance sequence.

[0186] In step (7), in response to the dynamic changes in equipment status during the execution of the operation ticket, the dispatching system identifies steps that do not conform to the dispatching instructions through real-time monitoring and deviation analysis, and generates targeted correction instructions to ensure that the operation guidance is consistent with the actual operating conditions. This step significantly improves the adaptability and safety of power grid operation through multi-level data analysis and instruction optimization.

[0187] In step (7.1), the data acquisition device collects equipment parameters at a period of 100 milliseconds. The deviation data records the timestamp, parameter value and step number, providing accurate input for subsequent analysis.

[0188] In step (7.2), the deviation analyzer extracts the operating voltage and energy storage voltage as key features.

[0189] In step (7.3), the rule base stores various contingency plans for abnormal scenarios, such as "voltage insufficiency" corresponding to "voltage compensation plan". The rule matcher extracts correction instructions containing "start backup power", "switch operating power", and "verify energy storage voltage" based on deviation characteristics. The instruction verifier checks whether the correction instructions meet safety specifications, such as confirming that switching the backup power will not cause an overload. Verified instructions provide a reliable basis for subsequent timing planning.

[0190] In step (7.4), the time series analyzer plans the instruction interval based on the device response characteristics, the random forest algorithm predicts the correction effect through 100 decision trees, and the feasibility verification is based on historical data to ensure the reliability of the implementation of the scheme.

[0191] Through step (7) of this invention, the dispatching system updates the dispatching instruction set based on the verification results and generates a new operation guidance sequence using an instruction synthesizer. The corrected instructions include "start forced cooling" and "adjust paralleling timing." The operation guidance sequence adopts a tree structure, integrating standard paths and corrected branches, and provides real-time feedback display of circuit breaker voltage recovery and transformer temperature changes. This method, through dynamic deviation identification and instruction optimization, achieves real-time adjustment of the operation process, ensuring the safety and stability of power grid operation.

[0192] (8) Verify the dynamic correction instructions, simulate the execution of the adjusted step sequence, analyze whether the equipment status meets expectations, and generate an operation ticket execution plan.

[0193] Step (8) includes the following steps:

[0194] (8.1) Obtain the standard operation requirement parameters in the scheduling instruction set, and calculate the deviation values ​​of the three key indicators of operating voltage, operating current and energy storage voltage according to the standard operation requirement parameters;

[0195] (8.2) When the operating voltage deviation exceeds the preset safety threshold range, a long short-term memory network is used to perform sliding time window analysis on the operating voltage deviation value, and a voltage compensation scheme is matched from the rule base based on the analysis results.

[0196] (8.3) Generate backup power start command, operation power switching command and energy storage voltage verification command according to the voltage compensation scheme, and determine the execution timing interval based on the equipment action characteristic curve;

[0197] The command verification results meet electrical operation safety standards and mechanical action specifications. The revised command execution timing plan is based on the equipment's action characteristic curve, determining the minimum execution interval between commands according to the mechanical action response time and the rate of change of electrical parameters.

[0198] (8.4) The random forest algorithm is used to process the historical correction data to obtain the prediction results of the change trend of operating voltage and energy storage voltage. The prediction results are used to generate the voltage parameter response range and generate the operation ticket execution plan.

[0199] Specifically, the random forest algorithm, combined with historical correction data, predicts the changing trends of operating voltage and energy storage voltage, generating the expected response range. The prediction results show that the voltage parameters have recovered to the standard operating range. A standard correction scheme library is established for different operating conditions, and a tree-indexed structure is used to organize multi-level correction instructions. Response timing constraints and state verification rules are established between nodes to obtain a verified operation ticket execution scheme. A real-time monitoring and feedback mechanism continuously tracks the execution effect of correction instructions and supports dynamic adjustment of correction strategies.

[0200] In step (8), to address the reliability requirements of dynamically correcting operation tickets, the dispatching system uses a digital twin model to simulate and verify the corrected sequence of steps, ensuring that the equipment status is consistent with the instruction requirements and generating a safe and executable operation plan. This step, through multi-dimensional data analysis, timing prediction, and verification mechanisms, ensures the accuracy of instruction adjustments and the safety of power grid operations.

[0201] In step (8.1), the data acquisition device acquires the equipment status in real time, calculates the voltage deviation and energy storage voltage deviation, and records the step number and timestamp of the deviation data that exceeds the standard, providing an accurate basis for analysis.

[0202] In step (8.2), the Long Short-Term Memory (LSTM) network samples every 10 milliseconds with a 1-minute sliding window, totaling 6000 data points. The trend of voltage deviation increasing from 3 kV to 10 kV is analyzed, with a temporal correlation of 0.93. The rule base is then adjusted to match a "voltage compensation scheme" based on the deviation characteristics.

[0203] In step (8.3), three correction instructions are generated: start the backup power supply, switch the operating power supply, and verify the energy storage voltage. The timing of each instruction is determined based on the equipment's operating characteristic curve, and the instruction sequence ensures the coordination of electrical and mechanical actions.

[0204] In step (8.4), the random forest algorithm is trained based on 10,000 sets of historical data and predicted using 100 decision trees. The confidence interval of the prediction model is based on the statistical distribution of historical data. The generation and verification of the response curve provide quantitative support for the feasibility of the correction scheme.

[0205] Through step (8) of this invention, the execution effect of the correction instruction is simulated based on the digital twin model to verify whether the equipment status meets the requirements of the logical dependency graph. A real-time monitoring feedback mechanism tracks the execution process, recording key parameters such as the energy storage motor current and the displacement of the operating mechanism to ensure status recovery. In the scenario of abnormal transformer temperature, the system generates instructions to "start the cooling fan" and "adjust the parallel connection timing." The verification data of the correction scheme is stored in a knowledge base, organized using a tree-like index structure, layered according to equipment type and fault scenario, improving the efficiency of subsequent instruction generation. This method, through dynamic simulation and real-time feedback, significantly improves the reliability of operation ticket adjustments and the stability of power grid operation.

Claims

1. A method for monitoring the execution process of scheduling operation tickets based on digital twin technology, characterized in that, Includes the following steps: (1) Obtain the step sequence and logical association rules from the operation ticket, parse the preconditions and expected equipment status of each step, process and generate a dataset containing step numbers, dependencies and status parameters, generate a logical dependency graph based on the dataset, and construct a digital twin instance. (2) Obtain the matching degree between the preconditions of the steps in the analysis logic dependency diagram and the real-time device status data, identify the preconditions of the failure steps based on the matching degree, and generate alarm information containing the failure step number and status deviation value. (3) Based on the alarm information, query the device interaction records related to the failure steps in the digital twin instance, extract the external disturbances or operation records that cause the deviation, and obtain a list of dynamic change factors that cause state conflicts. (4) Combine the logical dependency diagram to analyze the list of dynamic change factors, obtain the influence weight of the preconditions of subsequent steps, and identify high-risk steps based on the influence weight. (5) Call the digital twin instance to simulate the evolution of equipment status in high-risk steps, generate an evolution path containing simulated status parameters and the trigger probability of protection devices, and obtain the risk assessment results of potential chain reactions; (6) Adjust the logical dependency graph step sequence of the operation ticket according to the trigger probability of the protection device, generate the target operation ticket containing the reordering step and the update precondition, and obtain the scheduling instruction set that meets the real-time status. (7) Monitor the deviation between the equipment status data and the scheduling instruction set in real time. If the deviation exceeds the preset threshold, generate a dynamic correction instruction containing the deviation step number and adjustment suggestions to obtain real-time updated operation guidance. (8) Verify the dynamic correction instructions, simulate the execution of the adjusted step sequence, analyze whether the equipment status meets expectations, and generate an operation ticket execution plan.

2. The method for monitoring the execution process of a scheduling operation ticket based on digital twin technology according to claim 1, characterized in that, Step (1) includes the following steps: (1.1) Based on the step identifier code in the operation ticket, deep semantic extraction is used to obtain the execution condition parameters and device status markers. The execution condition parameters are semantically segmented using text mining methods to obtain a sequence of step dependency relationships. (1.2) Construct a state transition matrix for the sequence of steps, wherein the row and column values ​​of the state transition matrix correspond to the step identifier code and the device status mark, and fill the matrix cells with logical dependency variable values ​​and device threshold constraint rules; (1.3) The rule definition processor is used to parse the state transition matrix to obtain a logical dependency variable mapping table, and the decision tree algorithm is applied to the mapping table to generate a logical dependency graph; (1.4) Extract the device state transition path based on the logical dependency graph, construct the device state prediction model for the state transition path using runtime indicators in the time series database, and obtain the digital twin instance by training the prediction model parameters through the random forest algorithm.

3. The method for monitoring the execution process of a scheduling operation ticket based on digital twin technology according to claim 1, characterized in that, Step (2) includes the following steps: (2.1) Based on the precondition threshold and monitoring parameters of the logical dependency graph, a data acquisition device is used to read real-time monitoring data from the field equipment, and the validity of the real-time monitoring data is judged by numerical verification rules to obtain the status monitoring dataset; (2.2) For the aforementioned precondition threshold and state monitoring dataset, a data standardization processor is used to perform normalization operations, and the condition matching degree is obtained by calculating the state deviation value between the two sets of normalized data through Manhattan distance. (2.3) The condition matching degree is classified according to the preset matching degree benchmark value. If the condition matching degree is less than the preset matching degree benchmark value, a support vector machine is used to perform clustering operation on the matching degree data of different levels, and a step failure mark is generated for the clustering results. (2.4) For the failure marker of the step, the failure step number and the state deviation value are extracted by the failure marker parser, the failure degree score is calculated by the state evaluation function, and an alarm message containing the failure step number and the state deviation value is generated based on the failure degree score.

4. The method for monitoring the execution process of a scheduling operation ticket based on digital twin technology according to claim 1, characterized in that, Step (3) includes the following steps: (3.1) Obtain the device interaction sequence from the digital twin instance based on the failure step number in the alarm information, and traverse the state change records before and after the alarm timestamp in the device interaction sequence through the time sequence association algorithm to obtain the state abnormal interval. (3.2) For the abnormal state interval, the numerical range of the state change records is verified, and a disturbance propagation link graph is constructed by a depth-first search algorithm; (3.3) Extract the sequence of interaction events between devices based on the disturbance propagation link diagram, calculate the event time correlation coefficient, and if the correlation coefficient is greater than a preset threshold, generate a causal association chain; (3.4) Perform state change analysis on the causal chain, calculate the state transition probability of each node through Bayesian network, identify key change nodes in the state transition process, and generate a list of dynamic change factors that cause state conflicts.

5. The method for monitoring the execution process of a scheduling operation ticket based on digital twin technology according to claim 1, characterized in that, Step (4) includes the following steps: (4.1) Extract the set of step nodes from the logical dependency graph, construct a propagation path graph of change factors for the set of step nodes, and quantify the decay rate of change factors in the propagation path graph by a path tracker to obtain the influence propagation matrix; (4.2) Extract the precondition group from the set of step nodes according to the influence propagation matrix, calculate the influence probability value of the changing factors on the precondition group through the Markov chain, and perform a weighted superposition of the influence probability value and the influence propagation matrix to obtain the condition influence weight. (4.3) A risk propagation network is constructed using a recurrent neural network based on the conditional influence weights. The risk transmission coefficient of each node is calculated for the risk propagation network. A risk accumulation function is generated for the risk transmission coefficient using a risk evaluator to obtain the step risk score. (4.4) The logical dependency strength of the step nodes is calculated by the step association analyzer, and the step risk score is associated with the logical dependency strength. The association weighting result is classified by the risk threshold discriminator to obtain the high-risk step number.

6. The method for monitoring the execution process of a scheduling operation ticket based on digital twin technology according to claim 1, characterized in that, Step (5) includes the following steps: (5.1) Obtain the device parameter curve from the digital twin instance based on the device identification number, obtain the state evolution dataset through the time-series scanning of the parameter curve, and use a state feature extractor to identify key state change points for the state evolution dataset. (5.2) For the key state change points, the protection logic parser is used to extract the protection device group configuration information, the protection device trigger probability is calculated according to the configuration information, and a protection action sequence is generated based on the trigger probability value; (5.3) Extract the chain trigger source identifier based on the protection action sequence, establish a chain reaction model for the trigger source identifier using a deep reinforcement learning algorithm, and obtain the state evolution prediction value through the chain reaction model; (5.4) Calculate the action delay parameter for the predicted state evolution value, construct a chain reaction propagation diagram based on the action delay parameter, obtain the risk propagation intensity through the chain reaction propagation diagram, and obtain the risk assessment result of the potential chain reaction.

7. The method for monitoring the execution process of a scheduling operation ticket based on digital twin technology according to claim 1, characterized in that, Step (6) includes the following steps: (6.1) Extract the step sequence nodes based on the trigger probability of the protection device in the logical dependency graph, and use the topological sorting algorithm to analyze the dependency relationship between the nodes to obtain the dependency relationship data; (6.2) Based on the dependency data, analyze the equipment operating parameters obtained by the real-time status monitor through the risk assessor, and generate a risk level sequence for the equipment operating parameters; (6.3) Reorder the step sequence in the time-series execution matrix according to the risk level sequence, and obtain the execution order table through the sequence validator based on the reordering result; (6.4) According to the execution order table, call the operation ticket template generator to extract standard instruction items, and use the instruction optimizer to optimize the standard instruction items to obtain a scheduling instruction set that meets the real-time status. (6.5) Extract the operation node relationship from the logical dependency graph, readjust the operation step priority in combination with the protection device trigger probability, construct the execution time window parameter table between steps, define the device interaction safety boundary value according to the time window parameter table, compile the operation ticket execution instruction template, map the device action sequence and operator response specifications, embed real-time state adaptation rules, generate a dynamic scheduling instruction set containing multi-branch execution paths, and attach the expected value of device response and fault tolerance range.

8. The method for monitoring the execution process of a scheduling operation ticket based on digital twin technology according to claim 7, characterized in that, Step (6.5) includes the following steps: (6.5.1) Obtain the operation node association relationship based on the logical dependency graph, construct the node relationship matrix using the association metric algorithm, and generate association metric values ​​for the node relationship matrix; (6.5.2) The priority parameter is obtained by normalizing the correlation quantified value using a priority adjuster, and the time window parameter table is obtained by predicting the time window parameter table using a recurrent neural network based on the priority parameter. (6.5.3) Extract the boundary values ​​of device state changes according to the time window parameter table, process the boundary values ​​with a boundary value calculator to obtain the device interaction safety threshold, and construct execution constraints for the safety threshold; (6.5.4) Construct conditional branch judgment logic for the execution constraints, use an instruction template parser to process the judgment logic to obtain execution instructions, mark the expected value of the device response for the execution instructions, use a fault tolerance calculator to perform interval calculation on the expected value of the device response, generate fault tolerance range parameters based on historical data statistics, and generate a dynamic scheduling instruction set containing multi-branch execution paths.

9. The method for monitoring the execution process of a scheduling operation ticket based on digital twin technology according to claim 1, characterized in that, Step (7) includes the following steps: (7.1) A data acquisition device is used to acquire equipment status monitoring data, and the difference between the monitoring data and the execution requirement parameters in the scheduling instruction set is calculated. The difference data exceeding the safety threshold range is obtained through a threshold comparator. (7.2) Based on the difference data, a deviation analyzer is used to extract key feature parameters, and a long short-term memory network is used to obtain the temporal correlation data of the feature parameters; (7.3) For the time-series correlation data, query the adjustment scheme in the adjustment rule base, and obtain the correction instruction content through the rule matcher; (7.4) Based on the content of the correction instruction, the random forest algorithm is used to calculate the device response curve, the expected response range of the correction scheme is obtained through the response prediction model, the original scheduling instruction set is updated, and a new operation guidance sequence is generated by the instruction synthesizer to obtain real-time updated operation guidance.

10. The method for monitoring the execution process of a scheduling operation ticket based on digital twin technology according to claim 1, characterized in that, Step (8) includes the following steps: (8.1) Obtain the standard operation requirement parameters in the scheduling instruction set, and calculate the deviation values ​​of the three key indicators of operating voltage, operating current and energy storage voltage according to the standard operation requirement parameters; (8.2) A long short-term memory network is used to perform sliding time window analysis on the operating voltage deviation value, and a voltage compensation scheme is matched from the rule base based on the analysis results; (8.3) Generate backup power start command, operation power switching command and energy storage voltage verification command according to the voltage compensation scheme, and determine the execution timing interval based on the equipment action characteristic curve; (8.4) The random forest algorithm is used to process the historical correction data to obtain the prediction results of the change trend of operating voltage and energy storage voltage. The prediction results are used to generate the voltage parameter response range and generate the operation ticket execution plan.

Images