A power grid intelligent agent collaborative management and control method and system, electronic equipment and medium

Abstract

The application belongs to the technical field of power grid business intelligent management and control, and specifically discloses a power grid intelligent agent cooperative management and control method, system, electronic equipment and medium, which comprises the following steps: acquiring cooperative basic data of multiple intelligent agents; performing compatibility measurement and mismatch diagnosis based on interface description data and interaction constraint data to obtain a compatibility matrix and mismatch diagnosis results; performing cooperative conflict identification based on operation monitoring data and the mismatch diagnosis results to obtain conflict representation results; constructing a global conflict loss function based on the compatibility matrix and the conflict representation results; generating a mediation strategy based on the global conflict loss function and the mismatch diagnosis results; and issuing interface adaptation mapping to an interface call control object and issuing a resource allocation scheme and behavior coordination rules to a cooperative operation control object. The application can unify interface compatibility influence and cooperative conflict influence into mediation basis, and realize linkage adjustment of interface interaction and cooperative execution process between intelligent agents.

Description

Technical Field

[0001] This invention belongs to the field of intelligent management and control technology for power grid operations, specifically relating to a collaborative management and control method, system, electronic equipment, and medium for intelligent power grid agents. Background Technology

[0002] With the advancement of digitalization and intelligentization in power grid operations, a large amount of structured and unstructured data has accumulated in business systems such as safety production, equipment operation and maintenance, electricity consumption information collection, marketing services, and power supply service command. Existing power grid business management typically utilizes data platforms, interface gateways, business system integration platforms, intelligent analysis tools, or single-point intelligent agent applications to achieve data access, task flow, alarm identification, work order processing, resource scheduling, and business handling. In intelligent agent collaboration scenarios, data transmission, interface calls, and task collaboration between intelligent agents are usually achieved through preset interface specifications, static interface verification, fixed priority scheduling rules, or manually configured business processes.

[0003] However, when multiple agents participate in power grid business collaboration, the interface descriptions, call constraints, and operational states between agents dynamically change with variations in business scenarios, interface versions, task loads, and collaborative relationships. Existing methods struggle to simultaneously characterize the impact of interface compatibility and collaborative conflicts between agents, and also find it difficult to uniformly incorporate interface mismatches, resource contention, behavioral conflicts, and version conflicts into the mediation criteria. This results in a lack of specificity in the generation of mediation strategies, making it difficult to stably manage the interface interactions and collaborative execution processes between agents. Summary of the Invention

[0004] The purpose of this invention is to provide a method, system, electronic device and medium for collaborative management and control of power grid intelligent agents, so as to solve the problems of lack of specificity of the mediation strategy and difficulty in stable management and control of the interface interaction and collaborative execution process between intelligent agents in the prior art.

[0005] To achieve the above objectives, the present invention employs the following technical solution: According to one aspect of the present invention, a method for collaborative management and control of power grid intelligent agents is provided, comprising the following steps: Acquire basic collaborative data from multiple intelligent agents, including interface description data, interaction constraint data, and operational monitoring data. Based on interface description data and interaction constraint data, compatibility measurement and mismatch diagnosis are performed to obtain a compatibility matrix and mismatch diagnosis results. The compatibility matrix is ​​used to characterize the degree of interface compatibility between agents, and the mismatch diagnosis results are used to characterize the reasons for interface mismatch between agents. Based on operational monitoring data and mismatch diagnosis results, collaborative conflict identification is performed to obtain conflict characterization results; A global conflict loss function is constructed based on the compatibility matrix and conflict characterization results. The global conflict loss function is used to characterize the impact of interface compatibility and the impact of collaborative conflict. A mediation strategy is generated based on the global conflict loss function and mismatch diagnosis results. The mediation strategy includes interface adaptation mapping, resource allocation scheme and behavior coordination rules. The interface adaptation mapping is sent to the interface call control object, and the resource allocation scheme and behavior coordination rules are sent to the collaborative operation control object to adjust the interface interaction and collaborative execution process between intelligent agents.

[0006] By adopting the above technical solution, by acquiring interface description data, interaction constraint data, and operation monitoring data, and by using the compatibility matrix, mismatch diagnosis results, and conflict characterization results together to construct a global conflict loss function, the problem of the separation between interface compatibility evaluation and conflict mediation in the existing intelligent agent collaboration process is solved. This achieves the technical effects of unified quantification, linkage mediation, and executable control of interface mismatch and collaboration conflict.

[0007] Furthermore, acquire basic collaborative data from multiple intelligent agents, including: Obtain interface description data from the agent management platform, interface registry, or active probes; Obtain interaction constraint data from interface configuration data or call logs; Runtime monitoring data is collected through runtime tracking points or monitoring agents.

[0008] By acquiring different types of collaborative basic data from the agent management platform, interface registration center, active probes, interface configuration data, call logs, runtime tracking points, or monitoring agents, the problem of scattered sources and inconsistent collection criteria for interface information, constraint information, and operational information in agent collaborative management is solved. This enables the complete collection of data required for compatibility measurement and conflict identification, and improves the data reliability of subsequent diagnostic results.

[0009] Furthermore, the interface description data includes one or more of the following: interface protocol data, parameter definition data, version information, capability declaration data, and return result data; the interaction constraint data includes one or more of the following: call frequency constraint, permission constraint, idempotency constraint, pre-call constraint, execution order constraint, and resource requirement constraint.

[0010] This avoids problems such as unclear scope of collaborative basic data and ambiguous compatibility measurement objects, so that interface structure, parameters, version, capabilities, return results and calling conditions can all be included in the subsequent measurement chain, improving the coverage and completeness of interface compatibility evaluation.

[0011] According to one embodiment of the present invention, the steps of performing compatibility measurement and mismatch diagnosis based on interface description data and interaction constraint data to obtain a compatibility matrix and mismatch diagnosis results include: Semantic constraint encoding is performed on the interaction constraint data to obtain constraint feature vectors; Compatibility measures are performed on the interface description data and constraint feature vectors to obtain the compatibility degree between agents. The inter-agent compatibility of multiple agents is matrixed to obtain a compatibility matrix; The mismatch between intelligent agents is diagnosed and processed to obtain the mismatch diagnosis result.

[0012] By encoding interaction constraint data into constraint feature vectors and matrixing the compatibility between agents, the problem of constraints such as call frequency, permissions, pre-call, and execution order being difficult to participate in unified compatibility calculation is solved. This enables a structured expression of the interface compatibility between agents and a traceable diagnosis of mismatch causes.

[0013] Furthermore, the steps of measuring the compatibility of interface description data and constraint feature vectors to obtain the compatibility between agents include: Perform protocol structure matching on the interface protocol data to obtain the protocol structure compatibility. Perform parameter matching on the parameter definition data to obtain parameter compatibility; Version compatibility is obtained by measuring version differences in version information. Functional matching is performed on the capability declaration data, constraint feature vectors, and returned result data to obtain the functional compatibility. The compatibility between intelligent agents is obtained by integrating the compatibility of protocol structure, parameters, version, and function.

[0014] By calculating the compatibility of protocol structure, parameters, version, and function separately, the problem that static verification based solely on field type or interface format is insufficient to reflect version differences, capability differences, and behavioral constraint differences is solved. This allows the compatibility between agents to simultaneously reflect four dimensions: structure, parameters, version, and function, thus improving the accuracy of compatibility measurement.

[0015] Furthermore, functional compatibility is obtained by performing functional matching processing on the capability declaration data, constraint feature vectors, and returned result data, including: The capability matching degree is obtained by performing capability matching processing on the capability declaration data between intelligent agents. The call constraint matching degree is obtained by performing call constraint matching processing on the constraint feature vectors between agents. The matching degree of the returned results is obtained by performing return structure matching processing on the returned result data between intelligent agents; Functional compatibility is obtained by integrating capability matching degree, call constraint matching degree, and return result matching degree.

[0016] By integrating capability matching degree, call constraint matching degree, and return result matching degree into functional compatibility degree, the problem that functional layer compatibility is difficult to judge solely based on interface fields and parameter formats is solved. This allows the agent's capability adaptation, call condition adaptation, and return structure adaptation to jointly participate in functional compatibility judgment, thereby improving the stability of functional compatibility evaluation.

[0017] According to one embodiment of the present invention, the step of diagnosing the mismatch in compatibility between intelligent agents and obtaining a mismatch diagnosis result includes: Low-compatibility pairs of agents are obtained by filtering the compatibility matrix for low compatibility. Identify the mismatch type of low-compatibility agents in interface description data or interaction constraint data, and obtain the cause of the mismatch. The reasons for mismatch are encoded to obtain incompatibility reason codes; The severity of the mismatch is obtained by quantifying the degree of mismatch corresponding to the cause of mismatch. The incompatibility reason code and severity are correlated to obtain the mismatch diagnosis result.

[0018] By filtering low-compatibility agent pairs from the compatibility matrix and further outputting the incompatibility reason code and severity, the problem of not being able to locate the mismatch type and degree of impact when only the compatibility score is output is solved. This achieves interpretable output of the interface mismatch reason and improves the pertinence and traceability of subsequent mediation strategy matching.

[0019] Furthermore, the incompatibility reason codes include one or more of the following: field missing reason codes, type mismatch reason codes, version mismatch reason codes, capability mismatch reason codes, frequency mismatch reason codes, order non-compliance reason codes, and return field incompatibility reason codes.

[0020] According to one embodiment of the present invention, the step of identifying collaborative conflicts based on operational monitoring data and mismatch diagnosis results to obtain conflict characterization results includes: By identifying target conflicts, resource conflicts, and behavioral conflicts in the operation monitoring data, the target conflicts, resource conflicts, and behavioral conflicts between intelligent agents can be obtained. Version conflict identification is performed on the version mismatch information in the operation monitoring data and mismatch diagnosis results to obtain the version conflict between intelligent agents; The conflict characteristics are summarized and processed to obtain the conflict representation results.

[0021] By simultaneously identifying target conflicts, resource conflicts, behavioral conflicts, and version conflicts, this approach addresses the problem that existing conflict identification methods only focus on single task conflicts or resource conflicts and struggle to cover complex and collaborative conflicts. This achieves a unified representation of multiple types of conflicts and improves the completeness and reliability of conflict identification.

[0022] According to one embodiment of the present invention, the step of constructing a global conflict loss function based on the compatibility matrix and conflict characterization results includes: The compatibility loss is quantized by performing compatibility loss quantization on the compatibility matrix to obtain the compatibility loss term; The conflict characterization results are subjected to collaborative conflict loss quantification to obtain the collaborative conflict loss term. The compatibility loss term and the collaborative conflict loss term are fused to obtain the global conflict loss function.

[0023] By merging the compatibility loss term and the collaborative conflict loss term into a global conflict loss function, the problem of inconsistent mediation criteria when handling interface compatibility risks and collaborative operation conflicts separately is solved. This allows the impact of interface compatibility to participate in strategy generation as part of the conflict mediation objective, thereby improving the consistency of the mediation objective.

[0024] The target conflict in the conflict characterization results is subjected to loss quantification to obtain the target conflict loss. The resource conflict in the conflict characterization results is subjected to loss quantification to obtain the resource conflict loss. The behavioral conflict in the conflict representation results is subjected to loss quantification to obtain the behavioral conflict loss. The version conflict in the conflict representation results is subjected to loss quantification to obtain the version conflict loss. The target conflict loss, resource conflict loss, behavior conflict loss and version conflict loss are merged to obtain the collaborative conflict loss item.

[0025] By quantifying the losses from target conflicts, resource conflicts, behavioral conflicts, and version conflicts separately, the problem of comparing and uniformly calculating the impact of different types of collaborative conflicts is solved. This enables the quantifiable fusion of multiple types of collaborative conflicts and improves the ability to distinguish the severity of conflicts when generating mediation strategies.

[0026] According to one embodiment of the present invention, the step of generating a mediation strategy based on a global conflict loss function and mismatch diagnosis results includes: The interface adaptation matching process is performed on the mismatch causes in the mismatch diagnosis results to obtain the interface adaptation mapping. Resource allocation schemes are obtained by processing resource conflicts in the conflict characterization results. Behavioral conflicts in the conflict representation results are processed for behavioral coordination to obtain behavioral coordination rules. The mediation strategy is obtained by combining interface adaptation mapping, resource allocation scheme and behavior coordination rules.

[0027] By generating interface adaptation mapping, resource allocation scheme and behavior coordination rules separately, and combining the three into a mediation strategy, the problem that a single scheduling rule cannot simultaneously handle interface mismatch, resource contention and behavior conflict is solved. This achieves composite mediation for multiple conflict sources and improves the output completeness of the mediation strategy.

[0028] According to one embodiment of the present invention, the steps of distributing interface adaptation mapping to the interface call control object and distributing resource allocation schemes and behavior coordination rules to the collaborative operation control object include: The interface adaptation mapping is distributed to the interface gateway or proxy layer to adjust the interface call relationship between smart agents; The resource allocation plan is sent to the scheduler or quota manager to adjust the resource quota or call frequency of the agent; The behavior coordination rules are sent to the orchestration engine or workflow scheduler to adjust the execution order, decision scope, or concurrency control method of the agents.

[0029] By distributing interface adaptation mapping, resource allocation schemes, and behavior coordination rules to the interface gateway or proxy layer, scheduler or quota manager, and orchestration engine or workflow scheduler respectively, the problem of lacking a clear execution object after the mediation strategy is generated is solved. This enables interface interaction adjustment, resource quota adjustment, and collaborative execution adjustment to be implemented and improves the closure of the system management chain.

[0030] Furthermore, after the steps of distributing the interface adaptation mapping to the interface call control object and distributing the resource allocation scheme and behavior coordination rules to the collaborative operation control object, the following are also included: Obtain execution feedback data for interface adaptation mapping, resource allocation schemes, and behavior coordination rules; The execution feedback data is processed to evaluate the execution effect, and the changes in compatibility and conflict loss are obtained. The weights of the compatibility change are adjusted to obtain the adjusted compatibility metric weights, or the weights of the conflict loss change are adjusted to obtain the adjusted conflict loss weights.

[0031] By acquiring the execution feedback data of the mediation strategy and correcting the compatibility metric weight or conflict loss weight based on the change in compatibility and the change in conflict loss, the problems of difficulty in transmitting the execution effect of the mediation strategy and difficulty in adjusting the metric parameters according to the running results are solved. This achieves closed-loop optimization of the agent collaborative management process and improves the adaptability of subsequent compatibility metric and conflict loss calculation.

[0032] According to one aspect of the present invention, a power grid intelligent agent collaborative management and control system is provided, comprising: The data acquisition module is used to acquire basic collaborative data of multiple intelligent agents. The basic collaborative data includes interface description data, interaction constraint data, and operation monitoring data. The measurement and diagnosis module is used to perform compatibility measurement and mismatch diagnosis based on interface description data and interaction constraint data, and obtain a compatibility matrix and mismatch diagnosis results. The compatibility matrix is ​​used to characterize the degree of interface compatibility between agents, and the mismatch diagnosis results are used to characterize the reasons for interface mismatch between agents. The conflict identification module is used to perform collaborative conflict identification based on operational monitoring data and mismatch diagnosis results, and obtain conflict characterization results. The loss function construction module is used to construct a global conflict loss function based on the compatibility matrix and conflict characterization results. The global conflict loss function is used to characterize the impact of interface compatibility and the impact of collaborative conflict. The strategy generation module is used to generate mediation strategies based on the global conflict loss function and mismatch diagnosis results. The mediation strategies include interface adaptation mapping, resource allocation schemes and behavior coordination rules. The task distribution module is used to distribute interface adaptation mappings to the interface call control object and distribute resource allocation schemes and behavior coordination rules to the collaborative operation control object in order to adjust the interface interaction and collaborative execution process between intelligent agents.

[0033] According to one aspect of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the power grid intelligent agent collaborative management and control method of any of the above embodiments.

[0034] According to one aspect of the present invention, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the power grid intelligent agent collaborative management and control method of any of the above embodiments.

[0035] The power grid intelligent agent collaborative management and control system, electronic device, and computer-readable storage medium provided by this invention also solve the problems raised in the background section. Attached Figure Description

[0036] The accompanying drawings, which form part of this specification, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a flowchart of the power grid intelligent agent collaborative management and control method according to Embodiment 1 of the present invention; Figure 2 This is a structural block diagram of the power grid intelligent agent collaborative management and control system according to Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the electronic device according to Embodiment 3 of the present invention.

[0037] Reference numerals in the attached figures: Electronic device 100; Memory 101; Processor 102; Computer program 103; Communication bus 104. Detailed Implementation

[0038] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0039] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this invention is for describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0040] Example 1 This embodiment provides a method for collaborative management and control of power grid intelligent agents, applicable to scenarios where multiple intelligent agents participate in collaborative processing of power grid services. For example... Figure 1 As shown, the method includes the following steps: Acquire basic collaborative data from multiple intelligent agents, including interface description data, interaction constraint data, and operational monitoring data. Based on interface description data and interaction constraint data, compatibility measurement and mismatch diagnosis are performed to obtain a compatibility matrix and mismatch diagnosis results. The compatibility matrix is ​​used to characterize the degree of interface compatibility between agents, and the mismatch diagnosis results are used to characterize the reasons for interface mismatch between agents. Based on operational monitoring data and mismatch diagnosis results, collaborative conflict identification is performed to obtain conflict characterization results; A global conflict loss function is constructed based on the compatibility matrix and conflict characterization results. The global conflict loss function is used to characterize the impact of interface compatibility and the impact of collaborative conflict. A mediation strategy is generated based on the global conflict loss function and mismatch diagnosis results. The mediation strategy includes interface adaptation mapping, resource allocation scheme and behavior coordination rules. The interface adaptation mapping is sent to the interface call control object, and the resource allocation scheme and behavior coordination rules are sent to the collaborative operation control object to adjust the interface interaction and collaborative execution process between intelligent agents.

[0041] In this embodiment, the power grid intelligent agent collaborative management and control method can be executed in a power grid business full-process intelligent management and control system. This system can include a source data layer, a data fusion layer, an intelligent agent middleware layer, and a business application layer. The source data layer includes structured and unstructured data sources such as the OMS system, PMS system, electricity consumption information collection system, distribution automation system, power supply service command system, and marketing business system. The data fusion layer includes a data access module, a mixed-granularity data fusion engine, a data quality verification module, and a data cleaning and transformation module, used to output a unified detailed business data wide table. The intelligent agent middleware layer includes an intelligent agent management platform, a basic large-scale model interface adapter, a prompt word engineering management module, a knowledge base management system, and a multi-agent collaborative scheduling engine. The business application layer can include business scenarios such as enterprise safety production prediction and early warning, risk compliance management collaborative supervision, intelligent handling of network security alarms, intelligent review of digital project requirements, power supply service risk management and control, intelligent early warning of omni-channel work orders, intelligent formulation of production and maintenance plans, and intelligent analysis of electricity sales and consumption.

[0042] The source data layer, data fusion layer, and agent platform layer can provide basic collaborative data; the agent platform layer can perform compatibility measurement, mismatch diagnosis, and collaborative conflict identification; the multi-agent collaborative scheduling engine can construct a global conflict loss function and generate mediation strategies; the interface gateway, agent layer, scheduler, quota manager, orchestration engine, or workflow scheduler can act as interface call control objects or collaborative operation control objects to execute corresponding strategies.

[0043] By incorporating the effects of interface compatibility and collaborative conflict between agents into a global conflict loss function, and further generating interface adaptation mapping, resource allocation schemes, and behavior coordination rules, the problem of the fragmentation of interface adaptation, resource scheduling, and behavior coordination in the existing multi-agent collaboration of power grids can be solved, thereby achieving unified mediation and closed-loop management of interface interaction and collaborative execution processes.

[0044] The specific steps of the above-mentioned collaborative management and control method for power grid intelligent agents are as follows: S1. Obtain basic collaborative data from multiple agents.

[0045] Obtain interface description data from the agent management platform, interface registry, or active probes; Obtain interaction constraint data from interface configuration data or call logs; Runtime monitoring data is collected through runtime tracking points or monitoring agents.

[0046] Interface description data includes one or more of the following: interface protocol data, parameter definition data, version information, capability declaration data, and return result data. Interface protocol data may include a list of interface fields, message format, serialization method, and interface path; parameter definition data may include parameter name, parameter type, value range, required parameters, and default values; version information may include protocol version number, API version number, and dependent version numbers; capability declaration data may include the agent's declaration of supported query, control, subscription, notification, audit, analysis, or execution capabilities; and return result data may include return fields, field types, return value range, error codes, and return structure.

[0047] Interaction constraint data includes one or more of the following: call frequency constraints, permission constraints, idempotency constraints, pre-call constraints, execution order constraints, and resource requirement constraints.

[0048] Operational monitoring data includes one or more of the following: call frequency, call sequence, error rate, response latency, and resource usage data such as CPU, memory, bandwidth, and call quota.

[0049] In power grid operations, the time granularity, encoding rules, and entity definitions of data from different source data layers may be inconsistent. To ensure that operational monitoring data and intelligent agent business inputs can be reliably used for subsequent compatibility measurements and conflict identification, a data fusion layer can be used to perform mixed-granularity data fusion on the source data layers. Specifically, through the data platform API gateway, a combination of incremental extraction and full synchronization is used to obtain structured, semi-structured, and unstructured data from the source data layers. Structured data may include equipment ledgers, measurement data, work order records, and project information; semi-structured data may include inspection logs, maintenance records, and meeting minutes; and unstructured data may include policy documents, technical specifications, and images.

[0050] To address the issue of inconsistent time granularity in data across different systems—for example, electricity information collection systems use daily frozen data, distribution automation systems use minute-level real-time data, and marketing systems use monthly statistical data—a time series alignment algorithm based on Dynamic Time Warping (DTW) can be used for alignment. The time series alignment distance can be expressed as: ; in, and These are sequences at different time granularities. To align the path, The distance is Euclidean. The DTW algorithm maps multi-granularity time series onto a unified time axis, providing a unified time benchmark for resource usage, call frequency, response latency, and changes in business status in operational monitoring data.

[0051] To address the difficulty of cross-system entity association, a knowledge graph-based entity semantic matching model can be constructed. The power grid knowledge graph can include entities such as transformer substations, lines, metering points, users, equipment, and work orders, along with their relationships. First, an entity semantic encoder is used to encode entity attribute information into vectors: ; in, name Indicates the entity name. type Indicates entity type, attributes This represents the set of entity attributes. Then, an entity alignment algorithm is used to calculate the similarity between entities across systems: ; in, For learnable parameters, Based on the similarity of knowledge graph structures, This is the balance coefficient. Alignment is learned from data through supervised or semi-supervised learning, typically with the training objective of minimizing the alignment loss function. This aims to ensure high similarity between aligned positive sample pairs and low similarity between aligned negative sample pairs. Alternatively, labeled cross-system entity alignment data can be used as a supervisory signal. (Balance coefficient) Choose non-negative real numbers; a grid search combined with a validation set can be used to determine the value: try a set of candidate values ​​on the validation set, for example: (0, 0.1, 0.3, 0.5, 0.7, 0.9, 1.0), and for each λ, calculate the entity alignment accuracy (or metrics such as Hits@k, MRR, etc.), and select the λ value that optimizes the performance of the validation set. This entity semantic matching result can be used to solve cross-system entity association problems.

[0052] After data fusion, the fused data can be quality checked from five dimensions: completeness, standardization, rationality, accuracy, and consistency, using a combination of rule engine verification and statistical detection. Quality verification can be achieved through a combination of rule engine and statistical detection. Specifically, business specifications and standards are transformed into machine-executable logical rules for automatic data validation, performing rule engine verification. This includes checking whether the "equipment voltage level" field is within a preset set of valid values, or whether the "work order creation time" is later than the "task issuance time," etc. Then, statistical methods are used to analyze the macro-distribution and micro-characteristics of the data to identify potential anomalies. This involves calculating the null value rate of a field, identifying significant deviations from historical ranges in the average daily electricity consumption over the past week, or detecting numerous records with non-compliant coding formats in the equipment ledger, thus achieving statistical detection. For abnormal data, a data cleaning process can be triggered. Finally, based on the relationship between data tables and data items, a star schema is used to integrate and output a wide table of detailed business data that meets business analysis needs. This detailed business data wide table can serve as the data foundation for operational monitoring data, agent business inputs, or subsequent task orchestration.

[0053] By clearly defining the sources, types, and optional fusion methods of collaborative basic data, it is possible to ensure that interface description data, interaction constraint data, and operation monitoring data have consistent data standards and traceable sources, thereby improving the accuracy of compatibility measurement, conflict identification, and mediation strategy generation.

[0054] Power grid business capabilities are encapsulated as intelligent agents, and multiple intelligent agents can be defined and registered in the intelligent agent management platform. Each intelligent agent can be represented by the following structure: ; in, ID Identify the intelligent agent; This describes the capabilities of the agent, corresponding to capability declaration data. Define the input data format. Defines the output data format, corresponding to the returned result data in the interface description data; Service quality metrics can include response time, accuracy, or availability. After agents register, an agent directory is created, which can serve as one source of interface description data.

[0055] The system can employ an event-driven mechanism to organize collaboration among multiple intelligent agents. Business event types can include data change events, rule-triggered events, user request events, and timed trigger events. Events can be represented using the following unified format: ; in, ID As an event identifier, Type For event type, Source As the source of the incident, Timestamp For event timestamps, Payload For event load. In the event... Payload It can carry business data, task objectives, calling objects, parameter information, or resource requirement information, and thus become part of the operation monitoring data or interaction constraint data.

[0056] Upon receiving a business event, complex tasks can be decomposed based on business intent recognition, and an intelligent agent workflow can be constructed. A BERT-based intent recognition model is used to classify the intent of the business event. ; in, x Input information for business events or user requests. y For candidate business intent, θ These are the parameters for the intent recognition model. θ This represents all trainable weight parameters of the BERT-based intent recognition model, including the multi-layer Transformer weights of the BERT pre-trained model itself and the weights of the classification layer (usually a fully connected layer + Softmax) added on top of BERT. θ Each weight parameter can take the value of any real number.

[0057] Based on the identified Intent It can retrieve matching agents from the agent catalog and generate agent orchestration sequences based on the workflow template library: ; in, For the set of candidate agents, Constraints This workflow includes task constraints, interface constraints, and operational constraints. It can serve as a basis for subsequent operational monitoring data and the identification of behavioral conflicts.

[0058] For multi-task concurrency scenarios, a multi-level feedback queue can be used for priority queue management. Assume the system has K priority queues. The priority decreases sequentially. Task priority can be dynamically adjusted based on task waiting time and execution history. ; in, Based on priority, Waiting time for the task The number of times the task has been executed is denoted as α, and β are adjustment coefficients. α ranges from [0.1, 1.0]. If the value is too large, new tasks will be quickly promoted, leading to frequent preemption; if the value is too small, the aging effect will be insufficient, and long-waiting tasks may face starvation. β ranges from [0.05, 0.5]. If the value is too large, the priority of a task will quickly drop to zero after a few executions, leading to frequent switching; if the value is too small, the improvement in fairness will be insignificant.

[0059] This priority information can be used to generate target conflict loss or resource allocation plans.

[0060] For critical business decisions, such as power grid risk assessment and maintenance plan approval, a distributed decision-making mechanism based on the Raft consensus algorithm can be adopted. For example, a Leader agent can be elected as the decision coordinator. The Leader collects decision suggestions from participating agents and forms a final decision after receiving consensus confirmations exceeding a preset proportion. The decision result is then broadcast to relevant agents and the execution system. After receiving the decision instructions, the executing agents can use API adapters to call backend business systems such as PMS, marketing systems, and scheduling systems to complete specific operations. The execution results are fed back to the event bus, triggering subsequent processing or forming a business loop.

[0061] The aforementioned agent definition, event format, task orchestration, priority scheduling, and distributed decision-making mechanism are used to provide a structured source for interface description data, interaction constraint data, and operation monitoring data, and to provide a basis for identifying target conflicts, resource conflicts, and behavioral conflicts in subsequent conflict representation results.

[0062] S2. Based on the interface description data and interaction constraint data, perform compatibility measurement and mismatch diagnosis to obtain the compatibility matrix and mismatch diagnosis results.

[0063] The specific execution process is as follows: (1) Semantic constraint encoding is performed on the interactive constraint data to obtain the constraint feature vector.

[0064] This embodiment establishes a semantic constraint encoding table, mapping arbitrary call constraints to a standardized six-tuple structure: ; in, Type The constraint type can include PRECEDENCE (pre-dependency), RATE_LIMIT (frequency limit), PERMISSION (permission requirement), SEQUENCE (execution order), and RESOURCE (resource requirement). Object The constraint object can be a specific interface, intelligent agent, resource, or permission object. Operator Represents relational operators, which can be ≤, ≥, ==, ∈, etc. ValueRepresents constraint values, which can be numbers, strings, or sets; Unit Indicates numerical units, such as times per second, cores, MB, etc. Weight This represents the constraint weight, and its value can be configured in the range of [0,1].

[0065] For any intelligent agent i Its interaction constraint data can be represented as a set of called constraints: .

[0066] The constraint set is encoded into a fixed-dimensional constraint feature vector, denoted as . ,in D The total number of dimensions defined for the encoding table. The encoding strategy can employ a hybrid encoding approach: for... Type Perform type-specific one-hot encoding, allocating a dimension block for each constraint type; Object Perform object embedding encoding to map discrete objects into low-dimensional dense vectors; Value Normalization coding is performed. Min-Max normalization is used for numerical constraints, and multi-hot coding is used for set constraints.

[0067] For including k A constrained intelligent agent i Its constrained eigenvectors can be expressed as: ; in: This indicates vector concatenation; One-hot encoded vector representing the constraint type; The embedding vector representing the constraint object; This represents the normalized constraint value; Indicates the first t The weight factors for each constraint range from [0, 1]; typically, the weights of all constraints are normalized to satisfy... .

[0068] The normalization of numerical constraint values ​​can be expressed as: .

[0069] Through the semantic constraint encoding described above, the originally discrete, heterogeneous, or semi-structured interaction constraint data is transformed into a computable constraint feature vector. This constraint feature vector is then used to calculate the inter-agent compatibility, particularly the call constraint matching degree calculation in functional compatibility.

[0070] (2) Perform compatibility measurement on the interface description data and constraint feature vectors to obtain the compatibility between agents.

[0071] Specifically, it includes: Perform protocol structure matching on the interface protocol data to obtain the protocol structure compatibility. Perform parameter matching on the parameter definition data to obtain parameter compatibility; Version compatibility is obtained by measuring version differences in version information. Functional matching is performed on the capability declaration data, constraint feature vectors, and returned result data to obtain the functional compatibility. The compatibility between intelligent agents is obtained by integrating the compatibility of protocol structure, parameters, version, and function.

[0072] The compatibility between intelligent agents can be expressed as: ; in, Represents intelligent agents With intelligent agents The overall interface compatibility, with a value range of [0, 1]; Represents intelligent agents With intelligent agents Protocol structure compatibility; Represents intelligent agents With intelligent agents Parameter compatibility; Represents intelligent agents With intelligent agents Version compatibility; Represents intelligent agents With intelligent agents Functional compatibility; These are the weighting coefficients for protocol structure compatibility, parameter compatibility, version compatibility, and functional compatibility, respectively, satisfying... .

[0073] Protocol structure compatibility is calculated using the following formula: ; in, Represents intelligent agents With intelligent agents The number of fields matched in the interface protocol; Represents intelligent agents With intelligent agents The total number of fields involved in the comparison.

[0074] Parameter compatibility is calculated using the following formula: ; in, Represents intelligent agents With intelligent agents The total number of parameters; Indicates the first The matching score of each parameter (based on a comprehensive evaluation of parameter name, type, value range, and mandatory requirement).

[0075] Version compatibility is calculated using the following formula: ; in, Representing intelligent agents respectively and intelligent agents The interface version number; This represents the highest possible version number in the system (used for normalization).

[0076] Functional compatibility The matching of capability declaration data, interaction constraint data, and return result data at the functional level is calculated according to the following formula: ; in, Indicates the degree of ability matching; Indicates the degree of constraint matching during the call; Indicates the matching degree of the returned results; These are the weight coefficients for capability matching, call constraint matching, and return result matching, respectively, satisfying... ;and, .

[0077] Furthermore, capability matching degree is used to measure the agent's ability. i With intelligent agents j The degree of overlap between the sets of capabilities declared by the interfaces is calculated according to the following formula: ; Represents intelligent agents The set of capabilities supported by the interface (such as querying, control, subscription, notification, etc.); Represents intelligent agents The set of capabilities supported by the interface; This indicates the number of elements in the set.

[0078] If both sides have exactly the same ability, then If there is no intersection, then the result is 0.

[0079] Call constraint matching degree It can be calculated based on the constraint feature vector. The constraint feature vector for the caller agent i's requirements... and the capability constraint feature vector of the called agent j The constraint matching degree can be expressed as: ; in: These represent the dot product (inner product) of two constraint feature vectors, used to measure the similarity between constraints; Representing vectors Norm, i.e. ; It is cosine similarity, and its value range is... A larger value indicates a better match between the two constraint sets; Scaling factor ( ), used to adjust the sensitivity of the matching degree; ; is an S-shaped function that maps the matching degree to the interval (0, 1).

[0080] Return result matching degree This can be determined by the returned fields, field type, value range, whether it is optional, and the degree of matching of error codes: ; in, This indicates the total number of fields involved in the matching in the returned results; Represents intelligent agents The returned result is the first The definition of each field (name, type, value range, whether it is optional); This represents the field similarity function, with values ​​ranging from [0, 1]. Field similarity can be further defined as: .

[0081] (3) Perform matrix processing on the inter-agent compatibility of multiple agents to obtain the compatibility matrix.

[0082] By calculating the inter-agent compatibility between each pair of multiple agents, a compatibility matrix can be obtained. C =[ ].

[0083] (4) Diagnose and process the mismatch between intelligent agents to obtain the mismatch diagnosis results.

[0084] Specifically, it includes: Low-compatibility pairs of agents are obtained by filtering the compatibility matrix for low compatibility. Identify the mismatch type of low-compatibility agents in interface description data or interaction constraint data, and obtain the cause of the mismatch. The reasons for mismatch are encoded to obtain incompatibility reason codes; The severity of the mismatch is obtained by quantifying the degree of mismatch corresponding to the cause of mismatch. The incompatibility reason code and severity are correlated to obtain the mismatch diagnosis result.

[0085] Specifically, the compatibility between agents in the compatibility matrix is ​​first compared or sorted by a threshold, and agent pairs with compatibility below the preset compatibility threshold are identified as low-compatibility agent pairs.

[0086] For low-compatibility agent pairs, backtrack their interface description data and interaction constraint data, and identify the mismatch types that lead to reduced compatibility. Mismatch types include one or more of the following: missing fields, type mismatch, version mismatch, capability mismatch, frequency mismatch, order non-compliance, and incompatible return fields. The identified mismatch types are used as mismatch reasons to generate incompatibility reason codes and severity levels in subsequent processes.

[0087] This embodiment establishes a reason code dictionary containing L categories of incompatibility reasons: .

[0088] This represents the i-th incompatibility reason code. Incompatibility reason codes can include at least one or more of the following: missing field reason code, type mismatch reason code, version mismatch reason code, capability mismatch reason code, frequency mismatch reason code, order non-compliance reason code, and returned field incompatibility reason code. For mismatch reasons such as permission mismatch that occur in the interaction constraint data, these can also be recorded as mismatch information in the interaction constraint dimension in the mismatch diagnosis results to assist in the generation of mediation strategies.

[0089] For each activated incompatibility reason code, the corresponding severity can be calculated. s k Severity is used to quantify the impact of the mismatch on agent cooperation. Severity can be expressed as: ; in, For the first k The degree of deviation of the incompatibility reason code, such as the number of missing fields, type difference, and version difference; For the first k The trigger threshold for an incompatible reason code; This is the normalization scaling parameter, typically taken as 0.5 to 10. If d k The magnitude is large (e.g., version number difference 0~100), σ k A value of 5 to 10 is acceptable; if the magnitude is small (e.g., 0 to 10 fields are missing), a value of 0.5 to 2 is acceptable. η is the slope adjustment factor, typically ranging from 1 to 10. k=1 represents a standard Sigmoid curve, indicating low sensitivity; η k =3-5 indicates moderate sensitivity; η k ≥8 indicates high sensitivity (approximately a step function); This is used to map severity to the (0, 1) interval.

[0090] The output of the compatibility matrix is ​​expanded into the following mismatch diagnosis structure: ; in: For intelligent agents i With intelligent agents j Compatibility between intelligent agents; This is the incompatibility reason code vector; This corresponds to the severity vector; This is a mediation suggestion text generated based on the incompatibility reason code and severity. This mismatch diagnostic structure can serve as the mismatch diagnostic result in the claims.

[0091] Furthermore, the mediation strategy generation module can automatically match mediation actions based on the cause code and severity: ; in, Indicates according to the first Incompatibility reason codes and their severity The generated mediation strategy. Specifically, when At lower levels, perform lightweight adjustments, such as automatic type conversion; when In the intermediate stage, standard mediation is performed, such as field mapping and permission requests; when At higher levels, more stringent adjustments are performed, such as agent replacement or workflow re-orchestrating. For example, when a missing field reason code is detected, a field completion mapping can be generated; when an order does not meet a reason code, a workflow re-orchestrating or order adjustment rules can be generated.

[0092] By using incompatibility cause code vectors and severity vectors, low-compatibility issues in the compatibility matrix are transformed into mismatch diagnostic results that can be interpreted, diagnosed, and driven by strategies, thereby improving the stability of interface mismatch localization and the accuracy of mediation strategy matching.

[0093] S3. Based on operational monitoring data and mismatch diagnosis results, collaborative conflict identification is performed to obtain conflict characterization results.

[0094] Specifically, it includes: By identifying target conflicts, resource conflicts, and behavioral conflicts in the operation monitoring data, the target conflicts, resource conflicts, and behavioral conflicts between intelligent agents can be obtained. Version conflict identification is performed on the version mismatch information in the operation monitoring data and mismatch diagnosis results to obtain the version conflict between intelligent agents; The conflict characteristics are summarized and processed to obtain the conflict representation results.

[0095] Goal conflicts can be identified through consistency between agent task goals, directional differences between business goal vectors, or conflicts in task priorities. Resource conflicts can be identified through the relationship between the agent's requests for resources such as CPU, memory, bandwidth, and call quotas and the total system capacity. Behavioral conflicts can be identified through call chains, execution order, concurrent access status, decision scope, and response latency. Version conflicts can be identified through call failures, exception returns, version fields in runtime monitoring data, and version mismatch information in mismatch diagnosis results.

[0096] The conflict characterization results can include the conflict type, the agents involved, the time of occurrence, the duration, the severity, the scope of impact, and the corresponding operational monitoring data. These results are used to construct the global conflict loss function. Compared to the mismatch diagnosis results, the conflict characterization results focus more on the collaborative state during operation, while the mismatch diagnosis results focus more on the mismatch state at the interface and interaction constraint levels. Both are incorporated into the global conflict loss function, enabling a unified representation of the impact of interface compatibility and collaborative conflict.

[0097] S4. Construct a global conflict loss function based on the compatibility matrix and conflict characterization results.

[0098] First, the compatibility loss is quantized on the compatibility matrix to obtain the compatibility loss term.

[0099] Then, the conflict representation results are subjected to collaborative conflict loss quantification to obtain the collaborative conflict loss term. Specifically, the target conflict in the conflict representation results is quantified to obtain the target conflict loss; the resource conflict in the conflict representation results is quantified to obtain the resource conflict loss; the behavioral conflict in the conflict representation results is quantified to obtain the behavioral conflict loss; the version conflict in the conflict representation results is quantified to obtain the version conflict loss; and the target conflict loss, resource conflict loss, behavioral conflict loss, and version conflict loss are fused to obtain the collaborative conflict loss term.

[0100] The global conflict loss function is used to characterize the impact of interface compatibility and cooperative conflict, and can be defined as a weighted sum of the cooperative conflict loss term and the compatibility loss term: ; in: This represents the value of the global conflict loss function; Losses due to conflict with objectives; Losses due to resource conflicts; Losses due to behavioral conflict; Losses due to version conflicts; This results in a loss of compatibility. The weight coefficients for target conflict loss, resource conflict loss, behavior conflict loss, version conflict loss, and compatibility loss are respectively assigned, and the following conditions are met: .

[0101] Target conflict loss measures the degree of conflict between the task objectives of multiple agents. Let there be N agents in the system, and the task objective of agent i be represented as a target vector. ( (where is the dimension of the target space), then the target conflict loss can be expressed as: ; in, The target alignment is represented as: ; in, , which is the inner product of the two target vectors; , is a vector Norm; Cosine similarity, with values ​​ranging from -1 to 1. Higher values ​​indicate greater similarity between the targets. For two intelligent agents i With intelligent agents j The task priority difference vector; The norm of the priority difference; This is a priority decay factor used to control the degree to which priority differences affect alignment.

[0102] Resource conflict loss is used to measure the degree of competition among multiple agents for resources in a finite system. Suppose the system has... M For a given resource, the loss from resource conflict can be expressed as: ; in, Index for resource types; Total number of resource types; This indicates whether there is contention for resource r; it is set to 1 if there is contention, and 0 otherwise. Represents intelligent agents Resources The number of requests, such as the number of CPU cores and the number of MB of memory; Representing resources The maximum total system capacity; The resource over-limit index. This is used to control the nonlinear amplification effect of exceeding limits; Represents intelligent agents Resources The expected duration of occupancy; To normalize the time base, the maximum task execution time of the system is usually taken; For resources The business value weight is determined by the importance of the business.

[0103] Behavioral conflict loss measures conflicts between agents caused by execution order, concurrent execution, overlapping decision scope, and abnormal response latency. It can be expressed as: ; in, This is a sequential conflicting sub-item; For concurrent conflicting sub-items; For overlapping sub-items in the decision-making scope; For response delay exception sub-items; , , , These are the weights corresponding to the sequential conflict sub-items, concurrent conflict sub-items, overlapping decision scope sub-items, and abnormal response delay sub-items.

[0104] In this embodiment, a call dependency graph between agents is constructed based on the agent workflow, the call sequence in the operation monitoring data, and the pre-call constraints in the interaction constraint data. Nodes in the call dependency graph represent agents or agent interfaces, and edges represent the call dependency relationship between one agent or agent interface and another agent or agent interface. Pre-call constraints indicate other agent or interface calls that must be completed before a certain agent or interface can execute; the order of pre-calls is determined by the pre-call constraints and execution order constraints in the interaction constraint data.

[0105] Conflicting sub-items are represented as follows: ; in, To access the set of cyclic dependency edges detected in the dependency graph; This represents the severity of circular dependency edges, with values ​​ranging from [0, 1]. A set of intelligent agents with pre-existing constraints; For the indicator function, the agent The value is 1 if the preceding call order is violated, and 0 otherwise.

[0106] Concurrent conflicting sub-items are represented as: ; in, For a collection of shared resources; For currently accessing shared resources A collection of intelligent agents; For intelligent agents and For shared resources The degree of overlap in access times; For indicator functions, if the operation and The value is 1 if the elements are mutually exclusive, and 0 otherwise. Overlapping sub-items in the decision scope are represented as follows: ; in, For intelligent agents The set of decision-making scope, such as the set of managed equipment or the set of responsible areas; The size of the intersection of the decision-making ranges of the two agents; Let be the size of the union of the decision ranges of the two agents; is the Jaccard similarity coefficient; a larger value indicates a greater overlap in decision-making scope.

[0107] The response delay exception sub-item is represented as follows: ; in, The total number of agent pairs with a calling relationship; For calling a set of relational pairs; For intelligent agents Call The actual response delay afterward; This is a preset delay threshold; This is the latency sensitivity factor, indicating how sensitive the call chain is to latency. The function ensures that the loss is zero when the actual latency is below the threshold.

[0108] Version conflict loss is used to measure conflicts caused by version mismatches of interfaces or components. In this embodiment, agent pairs with calling relationships can be obtained from the call chain or call log in the runtime monitoring data; the major version number, minor version number, and revision version number corresponding to the agent can be obtained from the version information in the interface description data; the version mapping rules can be obtained from the mismatch diagnosis results, historical interface adaptation mappings, or pre-configured version mapping rules. Let P be the set of agent pairs with calling relationships, then the version conflict loss can be expressed as: ; in, For a set of agent pairs that have a calling relationship, Size of the set; For intelligent agents and The number of version mapping rules available between them; This represents the theoretically maximum number of mapping rules required. To map the coverage attenuation factor, ; For an exponentially decaying function, the more complete the mapping rule ( The larger the value, the smaller the decay term, and the lower the version conflict loss.

[0109] The version difference between agent i and agent j can be represented as: ; ; ; ; in, and These are the major version numbers of agents i and j, respectively; and These are the minor version numbers of agents i and j, respectively; and These are the revision numbers for agents i and j, respectively; For the weighting coefficients, satisfying ,generally .

[0110] The compatibility loss term is used to directly incorporate the compatibility matrix into the global conflict loss function, and can be expressed as: ; The overall interface compatibility after timing decay correction is expressed as: ; in, The compatibility matrix represents the inter-agent compatibility between agent i and agent j. The time elapsed since the last compatibility evaluation between agents i and j; As a compatibility decay time reference, This is the compatibility attenuation coefficient. This factor also incorporates the freshness of compatibility information into the global conflict loss function.

[0111] S5. Generate a mediation strategy based on the global conflict loss function and mismatch diagnosis results.

[0112] Specifically, it includes: The interface adaptation matching process is performed on the mismatch causes in the mismatch diagnosis results to obtain the interface adaptation mapping. Resource allocation schemes are obtained by processing resource conflicts in the conflict characterization results. Behavioral conflicts in the conflict representation results are processed for behavioral coordination to obtain behavioral coordination rules. The mediation strategy is obtained by combining interface adaptation mapping, resource allocation scheme and behavior coordination rules.

[0113] Interface adaptation mapping is used to handle mismatches in interface description data or interaction constraint data. When the mismatch is due to missing fields, field completion mapping is performed to obtain a field completion map; when the mismatch is due to type mismatch, type conversion mapping is performed to obtain a type conversion map; when the mismatch is due to version mismatch, version mapping is performed to obtain a version map; combining at least one of field completion mapping, type conversion mapping, and version mapping yields an interface adaptation map. Field completion mapping can set default values, extract field values ​​from the context, or map field values ​​from the results returned by the upstream agent; type conversion mapping can convert strings, numbers, enumerations, booleans, or time formats to the type required by the target interface; version mapping can establish field correspondences, parameter correspondences, and return structure correspondences between different version interfaces.

[0114] Resource allocation schemes can be generated based on the resource request volume, resource occupancy status, and global conflict loss function corresponding to resource conflicts. Resource status is identified from operational monitoring data to obtain the resource request volume and resource occupancy status of the agents experiencing resource conflicts. Resource allocation optimization is then performed on the resource request volume, resource occupancy status, and global conflict loss function to obtain the resource allocation result. A resource allocation scheme is determined based on the resource allocation result. For example, resource status identification can be performed on the agents experiencing resource conflicts to obtain their CPU, memory, bandwidth, or call quota request volume and occupancy status. Then, a resource allocation result is generated with the goal of reducing the global conflict loss function. Resource allocation schemes may include adjusting resource quotas, adjusting call frequency, delaying the execution of low-value tasks, adjusting concurrency, or reallocating computing resources.

[0115] Behavioral coordination rules can be generated based on the type of behavioral conflict. The conflict representation results are analyzed to identify the type of behavioral conflict, such as execution order conflict, concurrent execution conflict, or overlapping decision scope between agents. Based on these conflicts, coordination rules are generated to create execution order adjustment rules, concurrency control rules, or decision scope adjustment rules. At least two of these rules are combined to obtain the behavioral coordination rules.

[0116] When generating a mediation strategy, the goal is to reduce the global conflict loss function, and the interface adaptation direction is determined using the mismatch diagnosis results. Specifically, the mismatch cause and severity in the mismatch diagnosis results are used to determine the type and strength of the interface adaptation mapping; resource conflicts and their corresponding losses in the conflict characterization results are used to determine the resource allocation scheme; and behavioral conflicts and their corresponding losses in the conflict characterization results are used to determine the behavioral coordination rules. One or more candidate mediation strategies can be generated as needed, and the global conflict loss function after executing the candidate mediation strategies is estimated. The candidate mediation strategy that can reduce the global conflict loss function is selected as the final mediation strategy.

[0117] S6. Send the interface adaptation mapping to the interface call control object, and send the resource allocation scheme and behavior coordination rules to the collaborative operation control object.

[0118] Specifically, interface adaptation mappings can be distributed to the interface gateway or proxy layer to adjust the interface call relationships between agents; resource allocation schemes can be distributed to the scheduler or quota manager to adjust the resource quotas or call frequencies of agents; and behavior coordination rules can be distributed to the orchestration engine or workflow scheduler to adjust the execution order, decision scope, or concurrency control methods of agents.

[0119] An interface gateway or proxy layer can reside in the agent's call chain, adapting request parameters, request bodies, version identifiers, return fields, and error codes. A scheduler or quota manager can control the agent's computing resources, call quotas, concurrency levels, or call frequency. An orchestration engine or workflow scheduler can control the agent's execution order, task flow paths, concurrent execution conditions, and decision-making scope.

[0120] After the strategy is executed, execution feedback data can be obtained for interface adaptation mapping, resource allocation schemes, and behavior coordination rules. This feedback data can include strategy issuance status, strategy effectiveness status, interface call success rate, error rate, response latency, compatibility changes, resource utilization changes, conflict loss changes, and new conflict occurrences. After evaluating the execution feedback data, the changes in compatibility and conflict loss can be obtained. The change in compatibility can be obtained by comparing the compatibility matrix before and after strategy execution; the change in conflict loss can be obtained by comparing the global conflict loss function before and after strategy execution.

[0121] Furthermore, the weights of the compatibility change can be adjusted to obtain the adjusted compatibility metric weights; similarly, the weights of the conflict loss change can be adjusted to obtain the adjusted conflict loss weights. For example, when the compatibility improvement is insufficient after the interface adaptation mapping is executed, the weights of protocol structure compatibility, parameter compatibility, version compatibility, or functional compatibility can be adjusted; when the conflict loss decreases insufficient after the resource allocation scheme or behavior coordination rule is executed, the weights of target conflict loss, resource conflict loss, behavior conflict loss, version conflict loss, or compatibility loss can be adjusted.

[0122] The power grid intelligent agent collaborative management and control method of this embodiment obtains interface description data, interaction constraint data and operation monitoring data, and uses the compatibility matrix, mismatch diagnosis results and conflict characterization results together to construct a global conflict loss function. This solves the problem of the separation between interface compatibility evaluation and conflict mediation in the existing intelligent agent collaboration process, thereby achieving the technical effects of unified quantification, linkage mediation and executable management of interface mismatch and collaborative conflict.

[0123] Example 2 A power grid intelligent agent collaborative management and control system, such as Figure 2 As shown, it includes: The data acquisition module is used to acquire basic collaborative data of multiple intelligent agents. The basic collaborative data includes interface description data, interaction constraint data, and operation monitoring data. The measurement and diagnosis module is used to perform compatibility measurement and mismatch diagnosis based on interface description data and interaction constraint data, and obtain a compatibility matrix and mismatch diagnosis results. The compatibility matrix is ​​used to characterize the degree of interface compatibility between agents, and the mismatch diagnosis results are used to characterize the reasons for interface mismatch between agents. The conflict identification module is used to perform collaborative conflict identification based on operational monitoring data and mismatch diagnosis results, and obtain conflict characterization results. The loss function construction module is used to construct a global conflict loss function based on the compatibility matrix and conflict characterization results. The global conflict loss function is used to characterize the impact of interface compatibility and the impact of collaborative conflict. The strategy generation module is used to generate mediation strategies based on the global conflict loss function and mismatch diagnosis results. The mediation strategies include interface adaptation mapping, resource allocation schemes and behavior coordination rules. The task distribution module is used to distribute interface adaptation mappings to the interface call control object and distribute resource allocation schemes and behavior coordination rules to the collaborative operation control object in order to adjust the interface interaction and collaborative execution process between intelligent agents.

[0124] Example 3 like Figure 3 As shown, the present invention also provides an electronic device 100 for a collaborative management and control method for power grid intelligent agents; The electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on at least one processor 102, and at least one communication bus 104.

[0125] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the method for analyzing and finding abnormal power line losses in Embodiment 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101.

[0126] The memory 101 may primarily include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created based on the use of the electronic device 100 (such as audio data), etc. In addition, the memory 101 may include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other non-volatile solid-state storage device.

[0127] At least one processor 102 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 102 may be a microprocessor or any conventional processor. Processor 102 is the control center of electronic device 100, connecting various parts of electronic device 100 via various interfaces and lines.

[0128] The memory 101 in the electronic device 100 stores multiple instructions to implement a collaborative management and control method for a power grid intelligent agent, and the processor 102 can execute multiple instructions to achieve the following: Acquire basic collaborative data from multiple intelligent agents, including interface description data, interaction constraint data, and operational monitoring data. Based on interface description data and interaction constraint data, compatibility measurement and mismatch diagnosis are performed to obtain a compatibility matrix and mismatch diagnosis results. The compatibility matrix is ​​used to characterize the degree of interface compatibility between agents, and the mismatch diagnosis results are used to characterize the reasons for interface mismatch between agents. Based on operational monitoring data and mismatch diagnosis results, collaborative conflict identification is performed to obtain conflict characterization results; A global conflict loss function is constructed based on the compatibility matrix and conflict characterization results. The global conflict loss function is used to characterize the impact of interface compatibility and the impact of collaborative conflict. A mediation strategy is generated based on the global conflict loss function and mismatch diagnosis results. The mediation strategy includes interface adaptation mapping, resource allocation scheme and behavior coordination rules. The interface adaptation mapping is sent to the interface call control object, and the resource allocation scheme and behavior coordination rules are sent to the collaborative operation control object to adjust the interface interaction and collaborative execution process between intelligent agents.

[0129] Example 4 If the modules / units integrated in the electronic device 100 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, and read-only memory (ROM).

[0130] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0131] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0132] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0133] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0134] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0135] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for collaborative management and control of power grid intelligent agents, characterized in that, include: Acquire basic collaborative data from multiple intelligent agents, including interface description data, interaction constraint data, and operational monitoring data. Based on interface description data and interaction constraint data, compatibility measurement and mismatch diagnosis are performed to obtain a compatibility matrix and mismatch diagnosis results. The compatibility matrix is ​​used to characterize the degree of interface compatibility between agents, and the mismatch diagnosis results are used to characterize the reasons for interface mismatch between agents. Based on operational monitoring data and mismatch diagnosis results, collaborative conflict identification is performed to obtain conflict characterization results; A global conflict loss function is constructed based on the compatibility matrix and conflict characterization results. The global conflict loss function is used to characterize the impact of interface compatibility and the impact of collaborative conflict. A mediation strategy is generated based on the global conflict loss function and mismatch diagnosis results. The mediation strategy includes interface adaptation mapping, resource allocation scheme and behavior coordination rules. The interface adaptation mapping is sent to the interface call control object, and the resource allocation scheme and behavior coordination rules are sent to the collaborative operation control object.

2. The power grid intelligent agent collaborative management and control method according to claim 1, characterized in that, The steps of performing compatibility measurement and mismatch diagnosis based on interface description data and interaction constraint data to obtain a compatibility matrix and mismatch diagnosis results include: The interaction constraint data is semantically constrained to obtain a constraint feature vector; The compatibility of the interface description data and the constraint feature vector is measured to obtain the compatibility between the agents. The compatibility degree between multiple agents is matrixed to obtain the compatibility degree matrix; The mismatch between the intelligent agents is diagnosed and processed to obtain the mismatch diagnosis result.

3. The power grid intelligent agent collaborative management and control method according to claim 2, characterized in that, The steps for diagnosing the mismatch in compatibility between the intelligent agents and obtaining the mismatch diagnosis result include: The compatibility matrix is ​​subjected to low compatibility filtering to obtain low-compatibility agent pairs; Identify the mismatch type of the low-compatibility agent in the interface description data or interaction constraint data to obtain the cause of the mismatch. The mismatch reasons are encoded to obtain incompatibility reason codes; The severity of the mismatch corresponding to the cause of mismatch is quantified. The incompatibility reason code and the severity are correlated to obtain the mismatch diagnosis result.

4. The power grid intelligent agent collaborative management and control method according to claim 1, characterized in that, The steps for identifying collaborative conflicts based on operational monitoring data and mismatch diagnosis results to obtain conflict characterization results include: The operational monitoring data is subjected to target conflict identification, resource conflict identification, and behavioral conflict identification to obtain target conflict, resource conflict, and behavioral conflict between intelligent agents. Version conflict identification is performed on the version mismatch information in the operation monitoring data and the mismatch diagnosis results to obtain version conflicts between intelligent agents; The target conflict, resource conflict, behavior conflict, and version conflict are summarized and processed to obtain the conflict characterization result.

5. The power grid intelligent agent collaborative management and control method according to claim 1, characterized in that, The steps for constructing a global conflict loss function based on the compatibility matrix and conflict characterization results include: The compatibility loss is quantized by performing compatibility loss quantization on the compatibility matrix to obtain the compatibility loss term; The conflict characterization results are subjected to collaborative conflict loss quantification to obtain a collaborative conflict loss term; The compatibility loss term and the collaborative conflict loss term are fused to obtain the global conflict loss function.

6. The power grid intelligent agent collaborative management and control method according to claim 1, characterized in that, The steps for generating a mediation strategy based on the global conflict loss function and mismatch diagnosis results include: The interface adaptation matching process is performed on the mismatch cause in the mismatch diagnosis result to obtain the interface adaptation mapping. The resource conflicts in the conflict characterization results are processed by resource allocation to obtain the resource allocation scheme. The behavioral conflicts in the conflict representation results are subjected to behavioral coordination processing to obtain the behavioral coordination rules. The mediation strategy is obtained by combining the interface adaptation mapping, the resource allocation scheme, and the behavior coordination rules.

7. The power grid intelligent agent collaborative management and control method according to claim 1, characterized in that, The steps of distributing the interface adaptation mapping to the interface call control object and distributing the resource allocation scheme and behavior coordination rules to the collaborative operation control object include: The interface adaptation mapping is sent to the interface gateway or proxy layer to adjust the interface call relationship between intelligent agents; The resource allocation scheme is sent to the scheduler or quota manager to adjust the resource quota or call frequency of the intelligent agent; The behavior coordination rules are sent to the orchestration engine or workflow scheduler to adjust the execution order, decision scope, or concurrency control method of the agents.

8. A power grid intelligent agent collaborative management and control system, characterized in that, include: The data acquisition module is used to acquire basic collaborative data of multiple intelligent agents. The basic collaborative data includes interface description data, interaction constraint data, and operation monitoring data. The measurement and diagnosis module is used to perform compatibility measurement and mismatch diagnosis based on interface description data and interaction constraint data, and obtain a compatibility matrix and mismatch diagnosis results. The compatibility matrix is ​​used to characterize the degree of interface compatibility between agents, and the mismatch diagnosis results are used to characterize the reasons for interface mismatch between agents. The conflict identification module is used to perform collaborative conflict identification based on operational monitoring data and mismatch diagnosis results, and obtain conflict characterization results. The loss function construction module is used to construct a global conflict loss function based on the compatibility matrix and conflict characterization results. The global conflict loss function is used to characterize the impact of interface compatibility and the impact of collaborative conflict. The strategy generation module is used to generate mediation strategies based on the global conflict loss function and mismatch diagnosis results. The mediation strategies include interface adaptation mapping, resource allocation schemes and behavior coordination rules. The task distribution module is used to distribute interface adaptation mappings to the interface call control object and distribute resource allocation schemes and behavior coordination rules to the collaborative operation control object in order to adjust the interface interaction and collaborative execution process between intelligent agents.

9. An electronic device, characterized in that, It includes a processor and a memory, the processor being used to execute a computer program stored in the memory to implement the power grid intelligent agent collaborative management and control method 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 at least one instruction, which, when executed by a processor, implements the power grid intelligent agent collaborative management and control method as described in any one of claims 1 to 7.

Images