A method, device, and medium for cooperative monitoring of cross-level scheduling instructions

Abstract

The application discloses a cross-level dispatching instruction cooperative monitoring method and device, equipment and medium, and relates to the field of power system automation monitoring. The method comprises the following steps: generating a dispatching operation ticket object based on a single dispatching instruction extracted from a source end, and updating the current state of the dispatching operation ticket object to an updated state based on a finite state machine; if the updated state meets a checking state, performing semantic analysis on the dispatching operation ticket object according to a preset neural network model to obtain a target end semantic tuple, and performing consistency checking and operation violation checking on the target end semantic tuple in combination with a source end semantic tuple obtained from the source end to obtain an instruction checking result; wherein the neural network model is constructed based on a bidirectional neural network and a conditional random field; and determining the risk level of the dispatching instruction according to the instruction checking result to monitor the dispatching instruction. Through the implementation of the application, the monitoring accuracy of the cross-level dispatching instruction can be improved to meet the actual dispatching command requirements.

Description

Technical Field

[0001] This application relates to the field of power system automation monitoring, and in particular to a method, device, equipment and medium for collaborative monitoring of cross-level dispatch instructions. Background Technology

[0002] With the continuous expansion of the power grid, power dispatch has gradually evolved into a complex system involving the coordinated operation of multiple dispatch centers. During routine power grid operation and maintenance or fault handling, higher-level dispatch centers need to issue various operational instructions to lower-level centers, requiring strict timing and precise equipment mapping. To ensure power grid security, it is essential to ensure that every instruction, from issuance and circulation to execution and feedback, is transparent and controlled throughout the entire process, and that the semantics of the instructions remain consistent across all levels to avoid misunderstandings that could lead to operational errors. Therefore, it is necessary to audit and monitor the content of instructions to guarantee their safe execution.

[0003] The existing power grid dispatch and command systems at all levels typically use rule engines based on template or keyword matching to conduct preliminary screening of the standardization of terminology in operation tickets for the review and monitoring of instruction content. The operation records, time logs and personnel information corresponding to the operation tickets are stored in a relational database in the form of structured data to meet the needs of daily business query and statistical analysis data. However, for high-risk cross-level interaction scenarios, this workflow has unavoidable defects: (1) There are barriers to data interaction between dispatch and command systems at all levels. After the superior issue the instruction, it is impossible to perceive the specific execution of the subordinate in real time, resulting in a blind spot in supervision. The coordination of dispatch and command and the transparency of the entire instruction process are affected; (2) The existing rule engine can only identify and match simple standard terms, which is difficult to adapt to the diversity of natural language. When the instruction is a non-standard expression or has complex contextual logic, it is easy to miss potential semantic errors; (3) The simple structured storage of specific operation behaviors has trust risks and cannot effectively prevent data tampering. Therefore, it is difficult to provide highly credible data in the process of tracing responsibility. The above-mentioned shortcomings make it difficult for the review and monitoring of existing cross-level dispatch instructions to meet the actual dispatch and command needs. Therefore, how to optimize the review and monitoring of cross-level dispatch instructions to meet the actual dispatch and command needs remains an urgent problem to be solved by existing technologies. Summary of the Invention

[0004] This application provides a collaborative monitoring method, device, equipment, and medium for cross-level dispatching instructions, in order to solve the technical problem that the existing review and monitoring of cross-level dispatching instructions cannot meet the actual dispatching and command needs.

[0005] According to a first aspect of the embodiments of this application, a method for collaborative monitoring of cross-level scheduling instructions is provided, comprising: Based on a single scheduling instruction extracted from the source, a corresponding scheduling operation ticket object is generated, and based on a finite state machine, the current state of the scheduling operation ticket object is updated to determine the updated state of the scheduling operation ticket object. If the updated state meets the preset verification state, then according to the preset neural network model, the scheduling operation ticket object is semantically parsed to obtain the target semantic tuple, and combined with the source semantic tuple obtained from the source end, the target semantic tuple is subjected to consistency verification and operation violation verification to obtain the instruction verification result; wherein, the neural network model is constructed based on a bidirectional neural network and a conditional random field; Based on the instruction verification result, the risk level of the scheduling instruction is determined, and the scheduling instruction is monitored according to the risk level.

[0006] This application first extracts a single scheduling instruction from the source end to generate a scheduling operation ticket object. Then, it updates the current state of the scheduling operation ticket object using a finite state machine to determine the corresponding updated state. When the updated state meets the verification state, it semantically parses the scheduling operation ticket object using a neural network model based on a bidirectional neural network and a conditional random field to obtain a target-end semantic tuple. This target-end semantic tuple is then combined with the source-end semantic tuple for consistency and operation violation checks to obtain the instruction verification result. This result determines the risk level of the scheduling instruction and implements corresponding monitoring. Recording, updating, and maintaining the state of the scheduling instruction using a finite state machine ensures the continuity and accuracy of the state records, providing prerequisite state conditions for subsequent instruction monitoring. The neural network model based on a bidirectional neural network and a conditional random field is used to analyze the scheduling operation... Semantic parsing of voting objects enables the recognition and processing of descriptions, including natural language, avoiding errors and omissions that may occur due to the inability of existing rule engines to adapt to natural language. This improves the accuracy of subsequent monitoring to meet actual dispatch and command needs. Monitoring is implemented by verifying the consistency of semantic tuples at the target and source ends, as well as verifying operational violations. This allows for verification of the target end portion of cross-level and cross-end dispatch instructions from the source end, achieving collaboration between the source and target ends and improving monitoring accuracy to meet actual dispatch and command needs. Furthermore, verification from both consistency and operational violation perspectives ensures comprehensiveness, avoiding errors and omissions that may occur with existing simple rule-based judgments, thus improving the accuracy of subsequent monitoring to meet actual dispatch and command needs.

[0007] In some embodiments of this application, updating the current state of the scheduling operation ticket object based on a finite state machine to determine the updated state of the scheduling operation ticket object specifically includes: Based on the scheduling operation ticket object, the corresponding status tracking instance is found in the preset database, and the current status of the status tracking instance is determined. According to the preset mapping configuration table, the execution status of the scheduling operation ticket object is converted into an input event; Using a finite state machine, based on the input event, the state transition of the state tracking instance is executed so that the state tracking instance transitions from the current state to the updated state, thereby obtaining the updated state of the scheduling operation ticket object.

[0008] This application first determines the current state of the scheduling operation ticket object by searching for the corresponding state tracking instance in a preset database. Then, it converts the execution state of the scheduling operation ticket object into input events through a mapping configuration table. Finally, it executes the state transition of the state tracking instance through a finite state machine to determine the updated state of the scheduling operation ticket object. By searching for the state tracking instance of the scheduling operation ticket object in the database to determine the current state that can be used by the finite state machine, and by determining the input events that can be input into the finite state machine through the mapping configuration table, reliable data can be provided for subsequent state transitions executed through the finite state machine. By executing the state transition through the finite state machine to determine the updated state of the scheduling operation ticket object, the state of the scheduling instruction can be recorded, updated, and maintained in a timely manner to ensure the continuity and accuracy of the state records, providing prerequisite state conditions for subsequent instruction monitoring.

[0009] In some embodiments of this application, the step of semantically parsing the scheduling operation ticket object according to a preset neural network model to obtain the target semantic tuple specifically includes: The scheduling operation ticket object is mapped into a fused feature sequence through the input mapping layer of the neural network model; By using the bidirectional neural network layer of the neural network model, the contextual semantics of the fused feature sequence are extracted to obtain contextual semantic features; The contextual semantic features are decoded through the conditional random field layer of the neural network model to identify the key entities of the scheduling operation ticket object and obtain the target semantic tuple.

[0010] This application first maps scheduling operation ticket objects into fused feature sequences through the input mapping layer of a neural network model. Then, through a bidirectional neural network layer, it extracts the contextual semantics of the fused feature sequences to obtain contextual semantic features. Finally, through a conditional random field layer, it decodes the contextual semantic features, making it easy to identify the key entities of the scheduling operation ticket objects and obtain the target semantic tuples. By using a neural network model based on bidirectional neural networks and conditional random fields to perform semantic parsing of scheduling operation ticket objects, it can recognize and process descriptions including natural language, avoiding errors and omissions that may be caused by existing rule engines' inability to adapt to natural language. This improves the accuracy of subsequent monitoring and meets the actual scheduling and command needs.

[0011] In some embodiments of this application, the step of combining the source-end semantic tuple obtained from the source end with the target-end semantic tuple for consistency verification and operation violation verification to obtain the instruction verification result specifically includes: Based on a preset field matching function, the consistency of each semantic field in the target semantic tuple and the source semantic tuple is calculated to obtain a consistency score. Based on a preset rule engine, the target semantic tuples are matched for violations to obtain a set of violations; The instruction verification result of the scheduling instruction is determined based on the consistency score and the set of violations.

[0012] This application first calculates the consistency score by using a field matching function to evaluate the semantic fields in the target semantic tuple and the source semantic tuple. Then, based on a rule engine, it matches the non-compliant operations in the target semantic tuple to obtain a set of violations. Finally, it combines the consistency score and the set of violations to determine the instruction verification result of the scheduling instruction. By verifying the consistency of each semantic field in the target semantic tuple using the source semantic tuple, the source part of the cross-level and cross-end scheduling instruction can be used to verify the target part, thereby achieving coordination between the source and target ends and improving the accuracy of monitoring to meet the actual scheduling and command requirements. By verifying the consistency and operational violations of the target semantic tuple separately, the comprehensiveness of the verification can be ensured, avoiding errors and omissions that may be caused by existing simple rule judgments, thereby improving the accuracy of subsequent monitoring to meet the actual scheduling and command requirements.

[0013] In some embodiments of this application, the semantic fields include device code, device name, action type, device initial state, device target state, and instruction constraints; the calculation of the consistency of each semantic field in the target semantic tuple and the source semantic tuple based on a preset field matching function to obtain a consistency score specifically includes: Based on a preset discrete field matching function, the consistency of device encoding, action type and device initial state in the target semantic tuple and the source semantic tuple is calculated respectively to obtain the corresponding scoring items; Based on a preset continuous field matching function, the consistency of device name, device target status and instruction constraints in the target semantic tuple and the source semantic tuple is calculated respectively to obtain the corresponding scoring items; Based on a preset weight vector, the scoring items of each semantic field in the target semantic tuple are weighted and calculated to obtain a consistency score.

[0014] This application first calculates the consistency of device codes, action types, and initial device states in the target semantic tuple and the source semantic tuple using a discrete field matching function. Then, it calculates the consistency of device names, target device states, and instruction constraints using a continuous field function. Finally, it performs a weighted calculation on the scoring items reflecting the consistency of each semantic field in the target semantic tuple to obtain a consistency score. This allows for different consistency checks based on the type and characteristics of each semantic field, ensuring the comprehensiveness and accuracy of consistency checks, thereby improving the accuracy of monitoring and meeting the needs of actual dispatch and command.

[0015] In some embodiments of this application, the step of performing violation operation matching on the target semantic tuple based on a preset rule engine to obtain a set of violation items specifically includes: Based on the preset rule engine, forward reasoning is performed according to the target semantic tuple to determine several triggered rules, and these triggered rules are taken as violations to obtain a set of violations.

[0016] This application uses a rule engine to perform forward reasoning on the target semantic tuple, identify several triggered rules, and use these triggered rules as violations to obtain a set of violations. Based on the forward reasoning of the rule engine, it can easily and quickly determine the violations corresponding to the target semantic tuple, providing a corresponding data foundation for subsequent monitoring.

[0017] In some embodiments of this application, determining the risk level of the scheduling instruction based on the instruction verification result specifically includes: If the consistency score is within the preset alarm range, or if the set of violations only contains suggestive violations, then the risk level of the scheduling instruction will be determined as low risk. If the consistency score is within a preset incident range, or if the set of violations contains any blocking violation, then the risk level of the scheduling instruction is determined to be high risk. If the consistency score is not within the alarm interval and the accident interval, and the set of violations is an empty set, then the risk level of the scheduling instruction is determined to be safe.

[0018] This application determines dispatch instructions as low-risk when the consistency score is in the alarm range or the set of violations contains only suggestive violations; it determines dispatch instructions as high-risk when the consistency score is in the accident range or the set of violations contains any blocking violations; otherwise, it determines dispatch instructions as safe. By combining different states of the consistency score and the set of violations, it can accurately match and cover each risk level of dispatch instructions, ensuring the accuracy of risk level determination through verification, thereby improving the accuracy of subsequent monitoring to meet actual dispatch and command needs.

[0019] According to a second aspect of the embodiments of this application, a collaborative monitoring device for cross-level scheduling instructions is provided, including an object status update module, an object status verification module, and a scheduling instruction monitoring module; The object state update module is used to generate a corresponding scheduling operation ticket object based on a single scheduling instruction extracted from the source end, and update the current state of the scheduling operation ticket object based on a finite state machine to determine the updated state of the scheduling operation ticket object. The object state verification module is used to perform semantic parsing on the scheduling operation ticket object according to a preset neural network model if the updated state meets the preset verification state, to obtain the target semantic tuple, and to perform consistency verification and operation violation verification on the target semantic tuple in combination with the source semantic tuple obtained from the source end, to obtain the instruction verification result; wherein, the neural network model is constructed based on a bidirectional neural network and a conditional random field; The scheduling instruction monitoring module is used to determine the risk level of the scheduling instruction based on the instruction verification result, and to monitor the scheduling instruction based on the risk level.

[0020] In some embodiments of this application, the object state update module includes a tracking instance lookup unit, an execution state transition unit, and a state transition update unit; The tracking instance lookup unit is used to search for the corresponding status tracking instance in a preset database based on the scheduling operation ticket object, and determine the current status of the status tracking instance. The execution state conversion unit is used to convert the execution state of the scheduling operation ticket object into an input event according to a preset mapping configuration table; The state transition update unit is used to execute the state transition of the state tracking instance based on the input event using a finite state machine, so that the state tracking instance transitions from the current state to the updated state, thereby obtaining the updated state of the scheduling operation ticket object.

[0021] In some embodiments of this application, the object state verification module includes an operation ticket object mapping unit, a context semantic extraction unit, and a semantic feature decoding unit; The operation ticket object mapping unit is used to map the scheduling operation ticket object into a fused feature sequence through the input mapping layer of the neural network model; The context semantic extraction unit is used to extract the context semantics of the fused feature sequence through the bidirectional neural network layer of the neural network model to obtain context semantic features; The semantic feature decoding unit is used to decode the contextual semantic features through the conditional random field layer of the neural network model to identify the key entities of the scheduling operation ticket object and obtain the target semantic tuple.

[0022] In some embodiments of this application, the object status verification module includes a consistency score calculation unit, a violation operation matching unit, and a scheduling instruction verification unit; The consistency score calculation unit is used to calculate the consistency of each semantic field in the target semantic tuple and the source semantic tuple based on a preset field matching function, and obtain a consistency score. The violation matching unit is used to perform violation matching on the target semantic tuple based on a preset rule engine to obtain a set of violation items; The scheduling instruction verification unit is used to determine the instruction verification result of the scheduling instruction based on the consistency score and the set of violations.

[0023] In some embodiments of this application, the semantic fields include device code, device name, action type, device initial state, device target state, and instruction constraints; the consistency score calculation unit includes a first score subunit, a second score subunit, and a weighted score subunit. The first scoring subunit is used to calculate the consistency of device encoding, action type and device initial state in the target semantic tuple and the source semantic tuple based on a preset discrete field matching function, and obtain the corresponding scoring items. The second scoring subunit is used to calculate the consistency of device name, device target status and instruction constraints in the target semantic tuple and the source semantic tuple based on a preset continuity field matching function, and obtain the corresponding scoring items. The weighted scoring subunit is used to perform weighted calculations on the scoring items of each semantic field in the target semantic tuple based on a preset weight vector to obtain a consistency score.

[0024] In some embodiments of this application, the violation matching unit includes a forward reasoning matching subunit; The forward reasoning matching subunit is used to perform forward reasoning based on the target semantic tuples according to the preset rule engine, determine a number of triggered rules, and take the number of triggered rules as violations to obtain a set of violations.

[0025] In some embodiments of this application, the scheduling instruction monitoring module includes a first risk level determination unit, a second risk level determination unit, and a third risk level determination unit; The first risk level determination unit is used to determine the risk level of the scheduling instruction as weak risk if the consistency score is in a preset alarm range, or if the set of violations only contains suggestive violations. The second risk level determination unit is used to determine the risk level of the scheduling instruction as strong risk if the consistency score is in a preset accident range or if the set of violations contains any blocking violation. The third risk level determination unit is used to determine the risk level of the scheduling instruction as safe if the consistency score is not in the alarm interval and the accident interval, and the set of violations is an empty set.

[0026] This application first extracts a single scheduling instruction from the source end to generate a scheduling operation ticket object. Then, it updates the current state of the scheduling operation ticket object using a finite state machine to determine the corresponding updated state. When the updated state meets the verification state, it semantically parses the scheduling operation ticket object using a neural network model based on a bidirectional neural network and a conditional random field to obtain a target-end semantic tuple. This target-end semantic tuple is then combined with the source-end semantic tuple for consistency and operation violation checks to obtain the instruction verification result. This result determines the risk level of the scheduling instruction and implements corresponding monitoring. Recording, updating, and maintaining the state of the scheduling instruction using a finite state machine ensures the continuity and accuracy of the state records, providing prerequisite state conditions for subsequent instruction monitoring. The neural network model based on a bidirectional neural network and a conditional random field is used to analyze the scheduling operation... Semantic parsing of voting objects enables the recognition and processing of descriptions, including natural language, avoiding errors and omissions that may occur due to the inability of existing rule engines to adapt to natural language. This improves the accuracy of subsequent monitoring to meet actual dispatch and command needs. Monitoring is implemented by verifying the consistency of semantic tuples at the target and source ends, as well as verifying operational violations. This allows for verification of the target end portion of cross-level and cross-end dispatch instructions from the source end, achieving collaboration between the source and target ends and improving monitoring accuracy to meet actual dispatch and command needs. Furthermore, verification from both consistency and operational violation perspectives ensures comprehensiveness, avoiding errors and omissions that may occur with existing simple rule-based judgments, thus improving the accuracy of subsequent monitoring to meet actual dispatch and command needs.

[0027] According to a third aspect of the embodiments of this application, a computer device is provided, comprising: a processor; a memory; and a computer program stored in the memory and configured to be executed by the processor; wherein the processor executes the computer program to implement a collaborative monitoring method for cross-level scheduling instructions as described in this application.

[0028] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided, the computer-readable storage medium storing a plurality of instructions adapted for loading by a processor to execute a collaborative monitoring method for cross-level scheduling instructions as described in this application. Attached Figure Description

[0029] Figure 1 This is a flowchart illustrating a collaborative monitoring method for cross-level scheduling instructions according to certain embodiments of this application. Figure 2 This is a module structure diagram of a collaborative monitoring device for cross-level scheduling instructions, as shown in some embodiments of this application. Detailed Implementation

[0030] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below in conjunction with the accompanying drawings are exemplary and are only used to explain some embodiments of this application, and should not be construed as limiting the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments shown in this application without inventive effort are within the protection scope of this application.

[0031] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, unless otherwise explicitly specified, "a plurality of" or "several" means two or more.

[0032] The existing power grid dispatch and command systems at all levels typically use rule engines based on template or keyword matching to conduct preliminary screening of the standardization of terminology in operation tickets for the review and monitoring of instruction content. The operation records, time logs and personnel information corresponding to the operation tickets are stored in a relational database in the form of structured data to meet the needs of daily business query and statistical analysis data. However, for high-risk cross-level interaction scenarios, this workflow has unavoidable defects: (1) There are barriers to data interaction between dispatch and command systems at all levels. After the superior issue the instruction, it is impossible to perceive the specific execution of the subordinate in real time, resulting in a blind spot in supervision. The coordination of dispatch and command and the transparency of the entire instruction process are affected; (2) The existing rule engine can only identify and match simple standard terms, which is difficult to adapt to the diversity of natural language. When the instruction is a non-standard expression or has complex contextual logic, it is easy to miss potential semantic errors; (3) The simple structured storage of specific operation behaviors has trust risks and cannot effectively prevent data tampering. Therefore, it is difficult to provide highly credible data in the process of tracing responsibility. The above-mentioned shortcomings make it difficult for the review and monitoring of existing cross-level dispatch instructions to meet the actual dispatch and command needs. Therefore, how to optimize the review and monitoring of cross-level dispatch instructions to meet the actual dispatch and command needs remains an urgent problem to be solved by existing technologies.

[0033] Based on the above technical background, please refer to Figure 1 This application provides a collaborative monitoring method for cross-level scheduling instructions, including steps S101 to S103, each step of which is as follows: Step S101: Based on the single scheduling instruction extracted from the source, generate the corresponding scheduling operation ticket object, and update the current state of the scheduling operation ticket object based on the finite state machine to determine the updated state of the scheduling operation ticket object.

[0034] Specifically, due to network fluctuations or retry mechanisms during cross-network data transmission, the same scheduling instruction may generate multiple duplicate records. Therefore, when extracting a single scheduling instruction from the source, it is necessary to perform cleaning and deduplication. The preferred implementation method for cleaning and deduplication is a content hash-based deduplication algorithm, which involves extracting the key feature fields of the original record of the current scheduling instruction to calculate its unique feature signature. , ,in For hash functions, This represents the original, unique identifier of the current scheduling instruction at the source end. This indicates the service timestamp generated at the source end of the current scheduling instruction. This text represents the core business content of the current scheduling instruction. This represents a string concatenation operation. The calculated unique signature is compared with historical signatures in the preset cache. If a matching signature exists, the current record is considered duplicate data and discarded. If no matching signature exists, the scheduling instruction corresponding to the current signature is written to the cache and allowed to proceed, so that the corresponding scheduling operation ticket object can be generated later. The preset cache is a memory-based sliding time window cache maintained by the system. Each signature in the cache has a specific time-to-live (TTL), which is set to be greater than the maximum latency of network transmission.

[0035] Specifically, when acquiring scheduling instruction data extracted from the source and deduplicated before release, the scheduling instructions are converted into unified scheduling operation ticket objects based on a predefined field mapping rule table. The scheduling operation ticket object is defined as follows: ;in, This represents a globally unique ID for the scheduling operation ticket object, which is automatically assigned based on generation rules. The business type code represents the scheduling operation ticket object and is used to distinguish business types such as item-by-item order, integrated order, and authorized operation order; The overall description text of the operation task of the dispatch operation ticket object, such as changing the 500kV A line from operation to maintenance; This represents the information of the issuing party, including the name of the issuing party's organization, the name and employee number of the dispatcher; This indicates the recipient information object, including the name of the recipient organization and the name of the recipient personnel; It indicates the current execution status of the scheduling operation ticket, represented in code form; These represent the planned start and end times of the operation task, respectively. This represents the set of specific instruction sequences corresponding to a scheduling operation ticket object, denoted as . ,in The total number of instruction items. Indicates the first A specific switching operation instruction and steps.

[0036] In some embodiments of this application, updating the current state of the scheduling operation ticket object based on a finite state machine to determine the updated state of the scheduling operation ticket object specifically includes: Based on the scheduling operation ticket object, the corresponding status tracking instance is found in the preset database, and the current status of the status tracking instance is determined. According to the preset mapping configuration table, the execution status of the scheduling operation ticket object is converted into an input event; Using a finite state machine, based on the input event, the state transition of the state tracking instance is executed so that the state tracking instance transitions from the current state to the updated state, thereby obtaining the updated state of the scheduling operation ticket object.

[0037] Specifically, a deterministic finite state machine used to perform state transitions. Its mathematical definition is a quintuple. ;in, Given the set of all possible states for a given business type, for direct operations, The sequence corresponds to dispatch, repeater's order, dispatch confirmation, order recipient's reply, dispatcher's receipt of order, and order recipient's final confirmation. Represents a set of input events; Indicates the initial state; This represents the set of termination states, indicating the complete completion of an operation task. This represents the state transition function, which defines the next state to which a user should transition when a specific event is received in the current state. It is denoted as... .

[0038] Specifically, based on a preset mapping configuration table, the execution status of a scheduling operation ticket object can be converted into an input event. Record the current state of the state tracking instance as The update status of the status tracking instance and the scheduling operation ticket object is then... When the state transition function If the returned status is a valid value, assign it to... And obtain the updated state of the state tracking instance; when the state transition function When the returned status is invalid, such as null or error, it indicates that an illegal state transition has occurred. In this case, a process exception alarm is triggered directly, and a traceable "process violation" label is added to the timeline sequence of the corresponding status tracking instance. The current state of the finite state machine remains unchanged (or enters the "exception suspension" state) to wait for manual intervention or forced correction.

[0039] Specifically, after each state transition or event processing is completed, a new node record is appended to the timeline sequence of the corresponding state tracking instance. ;in, Represents the timestamp of the node record; This indicates that the node records the corresponding business status; This indicates the dispatcher or on-site recipient who is performing the operational task; This indicates the core business content of the operational task; This is a hash index pointing to the original data message, used to trace the original credentials.

[0040] This application first determines the current state of the scheduling operation ticket object by searching for the corresponding state tracking instance in a preset database. Then, it converts the execution state of the scheduling operation ticket object into input events through a mapping configuration table. Finally, it executes the state transition of the state tracking instance through a finite state machine to determine the updated state of the scheduling operation ticket object. By searching for the state tracking instance of the scheduling operation ticket object in the database to determine the current state that can be used by the finite state machine, and by determining the input events that can be input into the finite state machine through the mapping configuration table, reliable data can be provided for subsequent state transitions executed through the finite state machine. By executing the state transition through the finite state machine to determine the updated state of the scheduling operation ticket object, the state of the scheduling instruction can be recorded, updated, and maintained in a timely manner to ensure the continuity and accuracy of the state records, providing prerequisite state conditions for subsequent instruction monitoring.

[0041] Step S102: If the updated state satisfies the preset verification state, then according to the preset neural network model, the scheduling operation ticket object is semantically parsed to obtain the target semantic tuple, and combined with the source semantic tuple obtained from the source end, the target semantic tuple is subjected to consistency verification and operation violation verification to obtain the instruction verification result; wherein, the neural network model is constructed based on a bidirectional neural network and a conditional random field.

[0042] In some embodiments of this application, the preset verification state is preset based on all possible states of all service types of the scheduling instruction; if the updated state does not meet the preset verification state, the verification monitoring process of steps S102 and S103 is not performed after executing step S101 to obtain the updated state and completing the state tracking.

[0043] In some embodiments of this application, the step of semantically parsing the scheduling operation ticket object according to a preset neural network model to obtain the target semantic tuple specifically includes: The scheduling operation ticket object is mapped into a fused feature sequence through the input mapping layer of the neural network model; By using the bidirectional neural network layer of the neural network model, the contextual semantics of the fused feature sequence are extracted to obtain contextual semantic features; The contextual semantic features are decoded through the conditional random field layer of the neural network model to identify the key entities of the scheduling operation ticket object and obtain the target semantic tuple.

[0044] Specifically, the scheduling operation ticket object is an unstructured text, denoted as... The semantic parsing operation of the scheduling operation ticket object by the neural network model built based on bidirectional neural networks and conditional random fields is denoted as... Then the target semantic tuple ;in, Indicates a unique device code; The original description indicating the device name; The action type is represented by a standardized encoding obtained by mapping synonyms such as disconnect, pull-out, and trip to a unified code in an action mapping table. This describes the initial state of the equipment before operation. This describes the target state of the equipment after operation. This represents the set of constraints attached to the scheduling instruction.

[0045] Specifically, when decoding the contextual semantic features through the conditional random field layer of the neural network model, the joint probability of all possible label sequences is first calculated using the contextual semantic features and a predefined label transition matrix. Then, the label sequence with the highest probability is solved using the Viterbi algorithm to predict the label for each text position. Based on the optimal label sequence, specific labels are filtered out, and the text fragments corresponding to consecutive labels are aggregated to obtain preliminary entity candidates. These candidates are then post-processed using a dictionary from the power dispatching domain to determine key entities, and finally, the target semantic tuple is obtained.

[0046] More specifically, the bidirectional neural network layer of the neural network model can be constructed based on the Bi-LSTM model, the Bi-RNN model, and their improved models, preferably the Bi-LSTM model; by combining the bidirectional neural network with the Conditional Random Field (CRF), the context feature extraction capability of the bidirectional neural network and the label dependency modeling capability of the CRF can be integrated to obtain a better named entity recognition capability, thereby improving the recognition accuracy of key entities and obtaining more accurate target semantic tuples.

[0047] This application first maps scheduling operation ticket objects into fused feature sequences through the input mapping layer of a neural network model. Then, through a bidirectional neural network layer, it extracts the contextual semantics of the fused feature sequences to obtain contextual semantic features. Finally, through a conditional random field layer, it decodes the contextual semantic features, making it easy to identify the key entities of the scheduling operation ticket objects and obtain the target semantic tuples. By using a neural network model based on bidirectional neural networks and conditional random fields to perform semantic parsing of scheduling operation ticket objects, it can recognize and process descriptions including natural language, avoiding errors and omissions that may be caused by existing rule engines' inability to adapt to natural language. This improves the accuracy of subsequent monitoring and meets the actual scheduling and command needs.

[0048] In some embodiments of this application, the step of combining the source-end semantic tuple obtained from the source end with the target-end semantic tuple for consistency verification and operation violation verification to obtain the instruction verification result specifically includes: Based on a preset field matching function, the consistency of each semantic field in the target semantic tuple and the source semantic tuple is calculated to obtain a consistency score. Based on a preset rule engine, the target semantic tuples are matched for violations to obtain a set of violations; The instruction verification result of the scheduling instruction is determined based on the consistency score and the set of violations.

[0049] This application first calculates the consistency score by using a field matching function to evaluate the semantic fields in the target semantic tuple and the source semantic tuple. Then, based on a rule engine, it matches the non-compliant operations in the target semantic tuple to obtain a set of violations. Finally, it combines the consistency score and the set of violations to determine the instruction verification result of the scheduling instruction. By verifying the consistency of each semantic field in the target semantic tuple using the source semantic tuple, the source part of the cross-level and cross-end scheduling instruction can be used to verify the target part, thereby achieving coordination between the source and target ends and improving the accuracy of monitoring to meet the actual scheduling and command requirements. By verifying the consistency and operational violations of the target semantic tuple separately, the comprehensiveness of the verification can be ensured, avoiding errors and omissions that may be caused by existing simple rule judgments, thereby improving the accuracy of subsequent monitoring to meet the actual scheduling and command requirements.

[0050] In some embodiments of this application, the semantic fields include device code, device name, action type, device initial state, device target state, and instruction constraints; the calculation of the consistency of each semantic field in the target semantic tuple and the source semantic tuple based on a preset field matching function to obtain a consistency score specifically includes: Based on a preset discrete field matching function, the consistency of device encoding, action type and device initial state in the target semantic tuple and the source semantic tuple is calculated respectively to obtain the corresponding scoring items; Based on a preset continuous field matching function, the consistency of device name, device target status and instruction constraints in the target semantic tuple and the source semantic tuple is calculated respectively to obtain the corresponding scoring items; Based on a preset weight vector, the scoring items of each semantic field in the target semantic tuple are weighted and calculated to obtain a consistency score.

[0051] Specifically, the consistency of each semantic field in the target semantic tuple is determined by field matching: for the consistency of device code, action type and device initial state, the discrete field matching function returns 1 when the target field is completely equal to the source field, otherwise it returns 0; for the consistency of device name, device target state and instruction constraints, the continuous field matching function judges based on text similarity algorithm, returns 1 when the similarity is higher than the preset threshold, otherwise it returns 0.

[0052] Specifically, the consistency score and calculation are expressed by the following formula: ; in, For consistency scoring; These represent the source semantic tuple and the target semantic tuple, respectively. Total number of semantic fields; For the first The weights of each semantic field, and satisfying ; For the corresponding number The field matching function for each semantic field outputs the corresponding scoring item. The source semantic tuple and the target semantic tuple are respectively in the th... The corresponding field content of each semantic field.

[0053] This application first calculates the consistency of device codes, action types, and initial device states in the target semantic tuple and the source semantic tuple using a discrete field matching function. Then, it calculates the consistency of device names, target device states, and instruction constraints using a continuous field function. Finally, it performs a weighted calculation on the scoring items reflecting the consistency of each semantic field in the target semantic tuple to obtain a consistency score. This allows for different consistency checks based on the type and characteristics of each semantic field, ensuring the comprehensiveness and accuracy of consistency checks, thereby improving the accuracy of monitoring and meeting the needs of actual dispatch and command.

[0054] In some embodiments of this application, the step of performing violation operation matching on the target semantic tuple based on a preset rule engine to obtain a set of violation items specifically includes: Based on the preset rule engine, forward reasoning is performed according to the target semantic tuple to determine several triggered rules, and these triggered rules are taken as violations to obtain a set of violations.

[0055] Specifically, the rules engine has a built-in security rule base. , target semantic tuple As factual data input, traverse the security rule base. Matching is performed through forward reasoning using a rule engine, and its formulaic representation is as follows: The output is the target semantic tuple. The corresponding set of violations ,in For security rules that should have been met but were not, This represents the semantic tuple for the target end. Safety rules should be met. The corresponding conditions.

[0056] For example, security rule base Includes ground knife locking rules: if the operation action To close the grounding switch, the associated main disconnect switch status must be met. To disconnect; if the target semantic tuple If the main disconnect switch is in the closed position, the grounding switch interlocking rule is triggered, and this triggering rule should be added as a violation to the violation set. .

[0057] This application uses a rule engine to perform forward reasoning on the target semantic tuple, identify several triggered rules, and use these triggered rules as violations to obtain a set of violations. Based on the forward reasoning of the rule engine, it can easily and quickly determine the violations corresponding to the target semantic tuple, providing a corresponding data foundation for subsequent monitoring.

[0058] Step S103: Determine the risk level of the scheduling instruction based on the instruction verification result, and monitor the scheduling instruction based on the risk level.

[0059] In some embodiments of this application, determining the risk level of the scheduling instruction based on the instruction verification result specifically includes: If the consistency score is within the preset alarm range, or if the set of violations only contains suggestive violations, then the risk level of the scheduling instruction will be determined as low risk. If the consistency score is within a preset incident range, or if the set of violations contains any blocking violation, then the risk level of the scheduling instruction is determined to be high risk. If the consistency score is not within the alarm interval and the accident interval, and the set of violations is an empty set, then the risk level of the scheduling instruction is determined to be safe.

[0060] Specifically, when determining the risk level of a scheduling instruction through consistency scoring, a warning threshold is relied upon. and safety threshold ,in Then, by using the warning threshold and safety threshold of the consistency score, the three intervals corresponding to the consistency score can be determined: the warning interval. , denoted as a1; accident interval , denoted as a2; safe interval Let a3 be the set of violations. When determining the risk level of a scheduling instruction through the set of violations, it depends on the number and type of violations in the set. Violations include warning violations and blocking violations. Then, based on the number and type of violations in the set, we can determine the three intervals corresponding to the set of violations: containing at least 1 warning violation and at most 0 blocking violations (i.e., containing only warning violations), denoted as b1; containing at least 0 warning violations and at least 1 blocking violation (i.e., containing any blocking violation), denoted as b2; containing 0 warning violations and 0 blocking violations, in which case the set of violations is empty, denoted as b3.

[0061] More specifically, when the consistency score is in the alarm interval, or the set of violations contains only suggestive violations, with the corresponding interval combinations being a1b1, a1b3, and a3b1, the scheduling instruction is determined to be of a low-risk level. The corresponding monitoring operation is: do not interrupt the operation, but generate an alarm signal and mark the discrepancies, awaiting manual confirmation for subsequent operations. When the consistency score is in the incident interval, or the set of violations contains any blocking violations, with the corresponding interval combinations being a1b2, a2b1, a2b2, a2b3, and a3b2, the scheduling instruction is determined to be of a high-risk level. The corresponding monitoring operation is: interrupt the operation and generate a severe alarm signal. If the consistency score is not in the alarm interval or the incident interval, and the set of violations is an empty set, with the corresponding interval combination being a3b3, the scheduling instruction is determined to be of a safe level. The corresponding monitoring operation is: do not perform any interruption operation and continue to execute the operation task.

[0062] This application determines dispatch instructions as low-risk when the consistency score is in the alarm range or the set of violations contains only suggestive violations; it determines dispatch instructions as high-risk when the consistency score is in the accident range or the set of violations contains any blocking violations; otherwise, it determines dispatch instructions as safe. By combining different states of the consistency score and the set of violations, it can accurately match and cover each risk level of dispatch instructions, ensuring the accuracy of risk level determination through verification, thereby improving the accuracy of subsequent monitoring to meet actual dispatch and command needs.

[0063] Compared to existing technologies, this application first extracts a single scheduling instruction from the source end to generate a scheduling operation ticket object. Then, it uses a finite state machine to update the current state of the scheduling operation ticket object to determine the corresponding updated state. When the updated state meets the verification state, a neural network model based on a bidirectional neural network and a conditional random field semantically parses the scheduling operation ticket object to obtain a target-end semantic tuple. This target-end semantic tuple is then combined with the source-end semantic tuple for consistency and operation violation checks to obtain the instruction verification result. This allows for the determination of the risk level of the scheduling instruction and corresponding monitoring. Recording, updating, and maintaining the state of the scheduling instruction using a finite state machine ensures the continuity and accuracy of the state records, providing prerequisite state conditions for subsequent instruction monitoring. The neural network model based on a bidirectional neural network and a conditional random field further enhances this effect. Semantic parsing of scheduling operation tickets enables the recognition and processing of descriptions, including natural language, avoiding errors and omissions that may occur due to the inability of existing rule engines to adapt to natural language. This improves the accuracy of subsequent monitoring to meet actual scheduling and command requirements. Monitoring is implemented by verifying the consistency of semantic tuples at the target and source ends, as well as verifying operational violations. This allows for verification of the target end portion of cross-level and cross-end scheduling instructions from the source end, achieving collaboration between the source and target ends and improving monitoring accuracy to meet actual scheduling and command requirements. Furthermore, verification from both consistency and operational violation perspectives ensures comprehensiveness, avoiding errors and omissions that may occur with existing simple rule-based judgments, thus improving the accuracy of subsequent monitoring to meet actual scheduling and command requirements.

[0064] For a method corresponding to the one described above, please refer to [link / reference]. Figure 2 The present application provides a collaborative monitoring device for cross-level scheduling instructions, including an object status update module 210, an object status verification module 220 and a scheduling instruction monitoring module 230. The object state update module 210 is used to generate a corresponding scheduling operation ticket object based on a single scheduling instruction extracted from the source end, and update the current state of the scheduling operation ticket object based on a finite state machine to determine the updated state of the scheduling operation ticket object. The object state verification module 220 is used to, if the updated state meets the preset verification state, perform semantic parsing on the scheduling operation ticket object according to the preset neural network model to obtain the target semantic tuple, and combine it with the source semantic tuple obtained from the source end to perform consistency verification and operation violation verification on the target semantic tuple to obtain the instruction verification result; wherein, the neural network model is constructed based on a bidirectional neural network and a conditional random field; The scheduling instruction monitoring module 230 is used to determine the risk level of the scheduling instruction based on the instruction verification result, and to monitor the scheduling instruction based on the risk level.

[0065] In some embodiments of this application, the object state update module 210 includes a tracking instance lookup unit, an execution state transition unit, and a state transition update unit; The tracking instance lookup unit is used to search for the corresponding status tracking instance in a preset database based on the scheduling operation ticket object, and determine the current status of the status tracking instance. The execution state conversion unit is used to convert the execution state of the scheduling operation ticket object into an input event according to a preset mapping configuration table; The state transition update unit is used to execute the state transition of the state tracking instance based on the input event using a finite state machine, so that the state tracking instance transitions from the current state to the updated state, thereby obtaining the updated state of the scheduling operation ticket object.

[0066] In some embodiments of this application, the object state verification module 220 includes an operation ticket object mapping unit, a context semantic extraction unit, and a semantic feature decoding unit; The operation ticket object mapping unit is used to map the scheduling operation ticket object into a fused feature sequence through the input mapping layer of the neural network model; The context semantic extraction unit is used to extract the context semantics of the fused feature sequence through the bidirectional neural network layer of the neural network model to obtain context semantic features; The semantic feature decoding unit is used to decode the contextual semantic features through the conditional random field layer of the neural network model to identify the key entities of the scheduling operation ticket object and obtain the target semantic tuple.

[0067] In some embodiments of this application, the object status verification module 220 includes a consistency score calculation unit, a violation operation matching unit, and a scheduling instruction verification unit; The consistency score calculation unit is used to calculate the consistency of each semantic field in the target semantic tuple and the source semantic tuple based on a preset field matching function, and obtain a consistency score. The violation matching unit is used to perform violation matching on the target semantic tuple based on a preset rule engine to obtain a set of violation items; The scheduling instruction verification unit is used to determine the instruction verification result of the scheduling instruction based on the consistency score and the set of violations.

[0068] In some embodiments of this application, the semantic fields include device code, device name, action type, device initial state, device target state, and instruction constraints; the consistency score calculation unit includes a first score subunit, a second score subunit, and a weighted score subunit. The first scoring subunit is used to calculate the consistency of device encoding, action type and device initial state in the target semantic tuple and the source semantic tuple based on a preset discrete field matching function, and obtain the corresponding scoring items. The second scoring subunit is used to calculate the consistency of device name, device target status and instruction constraints in the target semantic tuple and the source semantic tuple based on a preset continuity field matching function, and obtain the corresponding scoring items. The weighted scoring subunit is used to perform weighted calculations on the scoring items of each semantic field in the target semantic tuple based on a preset weight vector to obtain a consistency score.

[0069] In some embodiments of this application, the violation matching unit includes a forward reasoning matching subunit; The forward reasoning matching subunit is used to perform forward reasoning based on the target semantic tuples according to the preset rule engine, determine a number of triggered rules, and take the number of triggered rules as violations to obtain a set of violations.

[0070] In some embodiments of this application, the scheduling instruction monitoring module 230 includes a first risk level determination unit, a second risk level determination unit, and a third risk level determination unit; The first risk level determination unit is used to determine the risk level of the scheduling instruction as weak risk if the consistency score is in a preset alarm range, or if the set of violations only contains suggestive violations. The second risk level determination unit is used to determine the risk level of the scheduling instruction as strong risk if the consistency score is in a preset accident range or if the set of violations contains any blocking violation. The third risk level determination unit is used to determine the risk level of the scheduling instruction as safe if the consistency score is not in the alarm interval and the accident interval, and the set of violations is an empty set.

[0071] This application first extracts a single scheduling instruction from the source end to generate a scheduling operation ticket object. Then, it updates the current state of the scheduling operation ticket object using a finite state machine to determine the corresponding updated state. When the updated state meets the verification state, it semantically parses the scheduling operation ticket object using a neural network model based on a bidirectional neural network and a conditional random field to obtain a target-end semantic tuple. This target-end semantic tuple is then combined with the source-end semantic tuple for consistency and operation violation checks to obtain the instruction verification result. This result determines the risk level of the scheduling instruction and implements corresponding monitoring. Recording, updating, and maintaining the state of the scheduling instruction using a finite state machine ensures the continuity and accuracy of the state records, providing prerequisite state conditions for subsequent instruction monitoring. The neural network model based on a bidirectional neural network and a conditional random field is used to analyze the scheduling operation... Semantic parsing of voting objects enables the recognition and processing of descriptions, including natural language, avoiding errors and omissions that may occur due to the inability of existing rule engines to adapt to natural language. This improves the accuracy of subsequent monitoring to meet actual dispatch and command needs. Monitoring is implemented by verifying the consistency of semantic tuples at the target and source ends, as well as verifying operational violations. This allows for verification of the target end portion of cross-level and cross-end dispatch instructions from the source end, achieving collaboration between the source and target ends and improving monitoring accuracy to meet actual dispatch and command needs. Furthermore, verification from both consistency and operational violation perspectives ensures comprehensiveness, avoiding errors and omissions that may occur with existing simple rule-based judgments, thus improving the accuracy of subsequent monitoring to meet actual dispatch and command needs.

[0072] It should be understood that the apparatus provided in this application is corresponding to the aforementioned method. The collaborative monitoring apparatus for cross-level scheduling instructions provided in this application can realize the collaborative monitoring method for cross-level scheduling instructions provided in any embodiment of this application.

[0073] Adaptively, embodiments of this application also provide a computer device and a computer-readable storage medium.

[0074] The computer device includes: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor; The processor, when executing the computer program, implements a collaborative monitoring method for cross-level scheduling instructions as described in this application.

[0075] The computer-readable storage medium stores multiple instructions that are adapted for a processor to load in order to execute a collaborative monitoring method for cross-level scheduling instructions according to this application.

[0076] The above description represents some embodiments of this application, providing a further detailed explanation of the purpose, technical solution, and beneficial effects of this application. It should be understood that the above-described embodiments of this application should not be construed as limiting this application. In particular, any changes, modifications, equivalent substitutions, and variations made by those skilled in the art within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A collaborative monitoring method for cross-level scheduling instructions, characterized in that, include: Based on a single scheduling instruction extracted from the source, a corresponding scheduling operation ticket object is generated, and based on a finite state machine, the current state of the scheduling operation ticket object is updated to determine the updated state of the scheduling operation ticket object. If the updated state meets the preset verification state, then according to the preset neural network model, the scheduling operation ticket object is semantically parsed to obtain the target semantic tuple, and combined with the source semantic tuple obtained from the source end, the target semantic tuple is subjected to consistency verification and operation violation verification to obtain the instruction verification result; wherein, the neural network model is constructed based on a bidirectional neural network and a conditional random field; Based on the instruction verification result, the risk level of the scheduling instruction is determined, and the scheduling instruction is monitored according to the risk level.

2. The collaborative monitoring method for cross-level scheduling instructions according to claim 1, characterized in that, The step of updating the current state of the scheduling operation ticket object based on a finite state machine to determine the updated state of the scheduling operation ticket object specifically includes: Based on the scheduling operation ticket object, the corresponding status tracking instance is found in the preset database, and the current status of the status tracking instance is determined. According to the preset mapping configuration table, the execution status of the scheduling operation ticket object is converted into an input event; Using a finite state machine, based on the input event, the state transition of the state tracking instance is executed so that the state tracking instance transitions from the current state to the updated state, thereby obtaining the updated state of the scheduling operation ticket object.

3. The collaborative monitoring method for cross-level scheduling instructions according to claim 1, characterized in that, The step of semantically parsing the scheduling operation ticket object according to a preset neural network model to obtain the target semantic tuple specifically includes: The scheduling operation ticket object is mapped into a fused feature sequence through the input mapping layer of the neural network model; By using the bidirectional neural network layer of the neural network model, the contextual semantics of the fused feature sequence are extracted to obtain contextual semantic features; The contextual semantic features are decoded through the conditional random field layer of the neural network model to identify the key entities of the scheduling operation ticket object and obtain the target semantic tuple.

4. The collaborative monitoring method for cross-level scheduling instructions according to claim 1, characterized in that, The step of combining the source-end semantic tuple obtained from the source end with the target-end semantic tuple for consistency verification and operation violation verification to obtain the instruction verification result specifically includes: Based on a preset field matching function, the consistency of each semantic field in the target semantic tuple and the source semantic tuple is calculated to obtain a consistency score. Based on a preset rule engine, the target semantic tuples are matched for violations to obtain a set of violations; The instruction verification result of the scheduling instruction is determined based on the consistency score and the set of violations.

5. The collaborative monitoring method for cross-level scheduling instructions according to claim 4, characterized in that, The semantic fields include device code, device name, action type, device initial state, device target state, and instruction constraints. The consistency score is calculated based on a preset field matching function, comparing the semantic fields of the target semantic tuple with those of the source semantic tuple to obtain a consistency score. Specifically, this includes: Based on a preset discrete field matching function, the consistency of device encoding, action type and device initial state in the target semantic tuple and the source semantic tuple is calculated respectively to obtain the corresponding scoring items; Based on a preset continuous field matching function, the consistency of device name, device target status and instruction constraints in the target semantic tuple and the source semantic tuple is calculated respectively to obtain the corresponding scoring items; Based on a preset weight vector, the scoring items of each semantic field in the target semantic tuple are weighted and calculated to obtain a consistency score.

6. The collaborative monitoring method for cross-level scheduling instructions according to claim 4, characterized in that, The preset rule engine performs violation matching on the target semantic tuples to obtain a set of violation items, specifically including: Based on the preset rule engine, forward reasoning is performed according to the target semantic tuple to determine several triggered rules, and these triggered rules are taken as violations to obtain a set of violations.

7. The collaborative monitoring method for cross-level scheduling instructions according to claim 4, characterized in that, The step of determining the risk level of the scheduling instruction based on the instruction verification result specifically includes: If the consistency score is within the preset alarm range, or if the set of violations only contains suggestive violations, then the risk level of the scheduling instruction will be determined as low risk. If the consistency score is within a preset incident range, or if the set of violations contains any blocking violation, then the risk level of the scheduling instruction is determined to be high risk. If the consistency score is not within the alarm interval and the accident interval, and the set of violations is an empty set, then the risk level of the scheduling instruction is determined to be safe.

8. A collaborative monitoring device for cross-level scheduling instructions, characterized in that, It includes an object status update module, an object status verification module, and a scheduling instruction monitoring module; The object state update module is used to generate a corresponding scheduling operation ticket object based on a single scheduling instruction extracted from the source end, and update the current state of the scheduling operation ticket object based on a finite state machine to determine the updated state of the scheduling operation ticket object. The object state verification module is used to perform semantic parsing on the scheduling operation ticket object according to a preset neural network model if the updated state meets the preset verification state, to obtain the target semantic tuple, and to perform consistency verification and operation violation verification on the target semantic tuple in combination with the source semantic tuple obtained from the source end, to obtain the instruction verification result; wherein, the neural network model is constructed based on a bidirectional neural network and a conditional random field; The scheduling instruction monitoring module is used to determine the risk level of the scheduling instruction based on the instruction verification result, and to monitor the scheduling instruction based on the risk level.

9. A computer device, characterized in that, include: processor; Memory; A computer program stored in the memory and configured to be executed by the processor; The processor, when executing the computer program, implements a collaborative monitoring method for cross-level scheduling instructions as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores multiple instructions adapted for loading by a processor to execute a collaborative monitoring method for cross-level scheduling instructions as described in any one of claims 1 to 7.

Images