A power grid fault disposal method and system based on a large model and multiple agents

Abstract

A power grid fault handling method and system based on a large model and multiple agents. Obtain and collect power grid multi-source data to form power grid fault information; perform semantic analysis on the power grid fault information, generate a fault handling strategy through a fine-tuned power field large model, and decompose the fault handling strategy into an executable subtask sequence to clearly define the input, output and execution order of each subtask; build multiple agents to collaboratively control multiple subtasks for executing the fault handling strategy; after the execution of the fault handling strategy is completed, set multiple-dimensional evaluation indicators to quantitatively evaluate the fault handling effect. The power grid fault handling technical solution of the large model and the multiple agents proposed in the application improves the intelligent level and emergency response efficiency of power grid fault handling.

Description

Technical Field

[0001] This invention belongs to the field of power system automation and artificial intelligence technology, and relates to a power grid fault handling method and system based on a large model and multiple agents. Background Technology

[0002] In modern power systems, with the continuous expansion of power grid scale and the increasing complexity of its structure, the types and impacts of power grid faults are becoming more diverse and complex.

[0003] With the development of artificial intelligence (AI) technology, intelligent fault handling methods have been introduced. For example, patent application CN113193549A uses network analysis mechanism calculation technology to propose a system and method for regulating transaction perception and fault collaborative handling, but it does not employ AI methods, resulting in certain shortcomings in flexibility and in-depth fault mining. Patent application CN118282032A implements a power grid fault handling method based on security analysis, but it only uses security analysis mechanism calculation methods during the handling process without combining AI methods for in-depth analysis of power grid fault information. Furthermore, the static security calculation, load transfer path calculation, and other mechanism calculation modules involved operate independently without a good overall collaborative strategy. Patent application CN115940124A constructs a deep convolutional neural network to achieve auxiliary recommendations for fault measures and electronic contingency plans based on the power grid fault situation. Its focus is on matching and recommending textual handling measures under fault conditions, without generating real-time handling strategies based on the current real-time operation of the power grid.

[0004] To address the aforementioned issues, there is an urgent need for a power grid fault handling method and system based on large models and multiple agents. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a power grid fault handling method and system based on a large model and multiple agents.

[0006] The present invention adopts the following technical solution.

[0007] In a first aspect, this invention provides a power grid fault handling method based on a large model and multiple agents, employing the following technical solution:

[0008] The system acquires and aggregates multi-source data from the power grid to form power grid fault information; the multi-source data includes fault time text information, real-time operating data, power grid topology information, and environmental variables.

[0009] The power grid fault information is semantically parsed, and a fault handling strategy is generated by fine-tuning the trained large power domain model. The strategy is then broken down into an executable sequence of sub-tasks, and the input, output and execution order of each sub-task are clearly defined.

[0010] Multiple intelligent agents are constructed to collaboratively control and execute multiple sub-tasks of the fault handling strategy; the intelligent agents include a risk analysis intelligent agent, a unit adjustment intelligent agent, a load transfer intelligent agent, a coordination intelligent agent, a safety verification intelligent agent, and a control execution intelligent agent;

[0011] After the fault handling strategy is executed, multi-dimensional evaluation indicators are set to quantitatively evaluate the fault handling strategy.

[0012] Furthermore, semantic parsing is performed on the power grid fault information, including:

[0013] Construct a power grid knowledge graph. The entity set includes power grid equipment, sites, and topology nodes, while the relationship set includes equipment affiliation, topology connection, and protection association.

[0014] Power grid fault information is transformed into structured fault features, including fault equipment identification, fault type, and fault equipment status; entity relationship analysis based on power grid knowledge graph is used to determine the fault cause and associated equipment;

[0015] Using a large-scale model in the power sector, fault handling strategies are generated based on the structured fault characteristics and the matching results of historical fault cases.

[0016] Furthermore, the collaborative mechanism of the multiple intelligent agents includes:

[0017] The risk analysis agent generates a risk analysis report based on the real-time operating status of the power grid and sends it to the coordination agent. The unit adjustment agent responds to the risk analysis agent, generates a unit adjustment plan based on multiple optimization objectives of the power grid, and sends it to the coordination agent. The load transfer agent runs parallel to the unit adjustment agent, searches for load transfer paths based on the power grid topology, generates a transfer plan, and sends it to the coordination agent. The coordination agent receives the risk analysis report, the unit adjustment plan, and the load transfer plan, summarizes them, determines the task execution order, and generates a fault handling plan.

[0018] The security verification agent is used to perform risk verification on the fault handling plan. If the handling plan passes the risk verification, it is sent to the control execution agent; if it fails, the verification result is fed back to the coordination agent. The control execution agent is used to issue control commands to the power grid equipment.

[0019] Furthermore, the execution steps of the risk analysis agent include:

[0020] Real-time acquisition of power grid operating parameters, equipment rated parameters, and safe operating thresholds for operating parameters; calculation of the mean value of the power grid operating parameters within a sliding window; if the difference between the power grid operating parameter and the corresponding mean value exceeds the threshold, it is preliminarily determined that there is an over-limit situation.

[0021] Based on the initial assessment, the predicted values ​​of operating parameters are obtained using time-series forecasting methods. If the predicted values ​​exceed the threshold, it is determined that there is a risk of exceeding the limit. Once the risk of exceeding the limit is confirmed, the current load values ​​of all nodes affected by the risk of exceeding the limit and requiring load transfer are weighted and summed to obtain the load to be transferred. A risk analysis report is generated, including the equipment exceeding the limit and its corresponding load to be transferred.

[0022] Furthermore, the execution steps of the unit adjustment agent include:

[0023] An optimization model is constructed with the goals of grid power balance and minimizing economic costs. The objective function includes economic cost, load balance, and frequency stability terms.

[0024] The economic cost item is the total power generation cost of all generator units; the load balance item is the supply-demand deviation penalty for all load nodes; and the frequency stability item is the frequency deviation penalty for all frequency monitoring points.

[0025] The constraints of the objective function include power balance constraints and upper and lower limits of generator output; solving the optimization model yields the unit adjustment scheme.

[0026] Furthermore, the execution steps of the load transfer agent include:

[0027] For the power grid topology, the shortest path is searched based on edge weights as the load transfer path. The factors affecting the edge weights include line impedance and transmission capacity.

[0028] Calculate the load transfer scheme score for the searched load transfer path. The load transfer scheme score is the reciprocal of the load transfer cost; the numerator is 1, and the denominator is the weighted sum of voltage drop, line loss, and load transfer capacity.

[0029] The load transfer path corresponding to the maximum score of the transfer scheme is taken as the optimal transfer scheme.

[0030] Furthermore, the multi-dimensional evaluation indicators include fault handling efficiency, power grid restoration efficiency, resource utilization efficiency, and reduction in economic losses; among which, fault handling time is calculated by the time difference between the fault occurrence time and the fault confirmation and elimination time; and the ratio of fault handling time to the historical handling time of the same type of fault is used as fault handling efficiency.

[0031] The power grid restoration efficiency value is calculated by using the proportion of the total load with restored power to the total load before the fault as the numerator and the weighted average of the power restoration time in each area of ​​the power grid as the denominator.

[0032] The total resource input is calculated by statistically analyzing the number of devices used and the duration of use during the fault handling process, and the amount of power restored is taken as the actual resource output; the ratio of the actual resource output to the total resource input is taken as the resource utilization efficiency.

[0033] The difference between the estimated economic loss under no remedial measures and the actual economic loss under the aforementioned fault handling strategy is used to represent the reduction in economic loss.

[0034] A second aspect of this application provides a power grid fault handling method based on a large model and multiple agents, employing the technical solution described in the first aspect of this application, the system comprising:

[0035] The fault perception and information aggregation module is used to acquire and aggregate multi-source data from the power grid to form structured power grid fault information. The multi-source data includes fault time text information, real-time operating data, power grid topology information, and environmental variables.

[0036] The large-scale model reasoning and task decomposition module is used to perform semantic parsing on the power grid fault information, generate fault handling strategies through a large-scale power model, and decompose them into an executable sequence of sub-tasks, specifying the input, output and execution order of each sub-task.

[0037] The intelligent agent collaborative execution module is used to construct multiple intelligent agents to collaboratively control and execute multiple sub-tasks of the fault handling strategy; the intelligent agents include a risk analysis intelligent agent, a unit adjustment intelligent agent, a load transfer intelligent agent, a coordination intelligent agent, a safety verification intelligent agent, and a control execution intelligent agent;

[0038] The fault handling effect evaluation and feedback optimization module is used to set multi-dimensional evaluation indicators after the fault handling strategy is executed, and to quantitatively evaluate the fault handling strategy.

[0039] Based on the above technical solution, this application includes at least one of the following beneficial effects:

[0040] 1. This invention can quickly complete key operations such as unit adjustment and load transfer, significantly shortening fault handling time and reducing the impact of faults on the power grid; furthermore, relying on the handling effect evaluation and feedback optimization mechanism, it forms the system's dynamic evolution capability. This overall approach breaks through the limitations of traditional methods' single-point optimization, improving the efficiency of power grid fault handling from the entire process of analysis and decision-making to task execution and iterative optimization, and providing a systematic solution for the safe and stable operation of the power grid.

[0041] 2. By leveraging the powerful logical reasoning capabilities of large models, in-depth analysis and processing of power grid fault information can be achieved, accurately locating the root cause and scope of impact of faults. At the same time, by utilizing the collaborative execution advantages of multiple agents, fault handling tasks can be completed efficiently, thereby significantly improving the safety and stability of power grid operation, shortening fault handling time, and reducing losses caused by faults.

[0042] 3. By training and optimizing the knowledge graph-enhanced large language model, it is made capable of deep semantic understanding of power industry terminology, cross-domain knowledge association reasoning, and logical deduction of complex fault scenarios. On this basis, by constructing a collaborative architecture of multiple intelligent agents such as risk analysis intelligent agents, unit adjustment intelligent agents, and load transfer intelligent agents, and combining power grid topology data and real-time monitoring information, the system can achieve rapid fault location, accurate judgment, and dynamic handling strategy generation, effectively improving the intelligence level of power grid fault handling and emergency response efficiency, and enhancing the overall level of power grid fault handling. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating the methods for handling power grid faults.

[0044] Figure 2 This is a schematic diagram of the intelligent agent collaboration process;

[0045] Figure 3 This is a system architecture diagram of a power grid fault handling system. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this invention clearer and more accurate, the technical solutions of this invention are described in detail below through several specific embodiments. The embodiments used in this invention are merely illustrative and are not intended to limit the scope of this invention.

[0047] As an embodiment of this application, refer to Figure 1 This embodiment provides a specific implementation method for a power grid fault handling method based on a large model and multiple intelligent graphs.

[0048] S1: Fault perception and information aggregation; acquire and aggregate multi-source data to form structured fault-related information.

[0049] The fault event text information is obtained through the integrated intelligent alarm system. An example format is "At xx hour xx minute, xxkV equipment tripped due to protection action, causing a power outage at xx station. The load before the incident was xxMW." This fault event text information provides a time description of the fault occurrence, the equipment involved, the voltage level, the protection action, and the outage area.

[0050] Obtain real-time operating data of equipment and systems from the SCADA system, including electrical quantities (voltage, current, system frequency), power quantities (active power, reactive power), status quantities (circuit breaker opening and closing status, line power flow), and rated parameters (rated voltage, rated current).

[0051] Obtain information on power grid structure and equipment configuration, including node connection relationships, line parameters (line length, impedance), substation main wiring method, and transformer parameters (transformer ratio, operating tap).

[0052] Environmental variables related to power grid operation are obtained from meteorological information systems, including wind speed, wind direction, rainfall, temperature, and humidity.

[0053] As a step in this embodiment, data from different information sources is formatted and transmitted synchronously based on a unified data interface standardization protocol. Fault information, real-time operational data, topology data, and meteorological parameters are aggregated into the fault perception and information aggregation module for centralized management.

[0054] S2: Large-scale model reasoning and task decomposition; semantic understanding and logical analysis of power grid fault information, and decomposition of fault handling process into sub-tasks.

[0055] 2.1: Data Input and Semantic Understanding;

[0056] The fault information (fault information, real-time operation data, topology data, and meteorological parameters) collected by S1 is input into the large model inference and task decomposition steps.

[0057] 2.2: The large model reasoning and task decomposition steps include the large model training and optimization module, the knowledge graph embedding module, and the reasoning and task decomposition module.

[0058] 2.2.1: Large Model Training and Optimization: A combination of supervised learning and reinforcement learning was used to train a large language model adapted to the power industry.

[0059] During the supervised learning phase, the cross-entropy loss function is used to optimize the model parameters, enabling the large language model to accurately classify and identify power grid fault types.

[0060] During the reinforcement learning phase, a policy gradient method is employed, with a reward function designed to evaluate the decision-making effectiveness and update policy parameters, enabling the large language model to gradually learn the optimal fault handling strategy. Throughout the reinforcement learning process, the model continuously interacts with the power grid operating environment, dynamically adjusting its decision-making strategy based on actual fault handling results. This allows the large model to gradually learn the optimal decision-making scheme in power grid fault handling tasks, thereby enhancing its ability to cope with complex fault scenarios.

[0061] 2.2.2: Knowledge Graph Embedding and Reasoning;

[0062] A power grid knowledge graph is constructed, with entities including equipment, stations, and topology nodes, and relationships including connectivity and operational status. The TransE model is used to map power grid entities and relationships to a low-dimensional vector space, enabling knowledge representation learning. As an example of this embodiment, a provincial power grid company constructs a knowledge graph containing the following elements:

[0063] Entity type: 500kV substation (entity ID: S1), 220kV main transformer (ID: T1), relay protection device (ID: R1); Relationship type: subordinate (transformer → substation), connected (equipment → line), protected object (device → equipment).

[0064] Construct a set of positive sample triples A negative sample set is generated by randomly replacing the first and last entities. For example, generating incorrect triples. In this application, positive samples represent real power grid topology relationships (such as equipment affiliation relationships), while negative samples refer to illegal relationships that are considered to be constructed (such as random replacement of equipment or substations).

[0065] For triples , and These are the head and tail entities, respectively. For relationships. The TransE model is based on the assumptions The vector is embedded using the model's standard loss function.

[0066] 2.2.3: Generation of fault handling procedures;

[0067] As a technical step in this embodiment, unstructured fault information is first semantically parsed using natural language processing technology and transformed into structured fault feature data. Entity relationship analysis based on knowledge graphs is then employed to determine the root cause of the fault, its scope of impact, and associated power grid equipment. For example, if the input is "the heavy gas protection of the #1 main transformer in 500kV substation A trips," the output would be {equipment: main transformer A1, fault type: heavy gas protection, status: tripped}.

[0068] By employing graph embedding and semantic reasoning, combined with a historical fault case database and handling experience, a fault handling process is generated, encompassing fault event-based grouping and assessment, risk analysis, contingency plan matching, or strategy generation. The historical fault case database includes fault descriptions, initial operating modes, post-fault operating modes, phased fault handling strategies, fault recovery, and information notification. Furthermore:

[0069] As a technical step in this embodiment, based on the historical fault case library, the large language model trained in step 2.2.1 is used to generate the optimal handling strategy using fault feature data as the state.

[0070] The processing flow of large model generation is analyzed, and the Hierarchical Task Network (HTN) planning algorithm is used to decompose the processing flow of large model generation into sub-tasks. Specifically:

[0071] Through task decomposition function According to the original task and system status This generates a set of executable subtasks. The input and output requirements of each subtask, as well as the execution order between subtasks, are defined by constraints imposed by the internal logic of the power system.

[0072] in, This represents the original task, i.e., the optimal handling strategy obtained. This refers to the current system state, i.e., the current power system state, such as the current power grid operating parameters and equipment switching status.

[0073] Break down the task into a set of multiple executable subtasks. and through constraints Clarify this The input and output requirements of each subtask, and the execution order.

[0074] As an example in this embodiment, the original task For the task "Restore power supply to main transformer #1 of 500kV substation A", the system status is "C={Current load: 400MW, standby main transformer A2 capacity: 500MW, adjacent substation can transfer 200MW of load}". Table 1 provides an example of the subtask decomposition results generated by the task decomposition module based on the original task and system status.

[0075] Table 1 Example of Task Decomposition

[0076]

[0077] For example, for a risk analysis task, the inputs are real-time monitoring data and equipment parameters, and the outputs are a list of out-of-limit equipment and the amount of load to be transferred.

[0078] The task of formulating unit adjustment plans takes the results of equipment exceeding limits as input and output power adjustment strategies as input.

[0079] The load transfer task outputs the optimal transfer path and alternative solutions based on the amount of load to be transferred and network topology data.

[0080] The system receives and processes security verification tasks and real-time power grid status data, and outputs security assessment results and risk warnings.

[0081] Coordinate the task and integrate the solutions of various intelligent agents, and output the optimized comprehensive handling strategy and execution priority;

[0082] The control task takes the final disposal plan as input and outputs the execution results and actual effect data.

[0083] This application utilizes a large-scale model inference and task decomposition module to perform semantic understanding and logical analysis of power grid fault information, achieving multi-dimensional fault feature extraction and accurate fault type identification. Based on the inference results, the complex fault handling process is decomposed into multiple interrelated sub-tasks, such as fault location refinement, isolation scheme formulation, and power restoration priority ranking. Simultaneously, by combining the power grid topology and real-time operational data, the task execution order and resource allocation strategy are dynamically optimized, providing a clear execution framework for subsequent multi-agent collaborative operations.

[0084] S3: Construct multiple agents through the agent collaborative execution module, and control the execution of multiple sub-tasks of the fault handling process through multiple agents.

[0085] Reference Figure 2 , Figure 2 This is a schematic diagram of the intelligent agent collaboration process. The intelligent agents constructed in this application include a risk analysis intelligent agent, a unit adjustment intelligent agent, a load transfer intelligent agent, a coordination intelligent agent, a safety verification intelligent agent, and a control execution intelligent agent. The processing flow and mechanism of each intelligent agent are described below.

[0086] 3.1: Risk Analysis Intelligent Agent;

[0087] Data acquisition; real-time acquisition of power grid operating parameters (line current, voltage, power, etc.), equipment rated parameters, and safe operating thresholds for each parameter;

[0088] Limit exceedance detection: Calculate the mean and standard deviation of real-time power grid operating parameters within a sliding window to determine if any limit exceedances occur;

[0089] When the difference between the real-time running parameter and the corresponding mean exceeds its threshold coefficient, it is preliminarily determined that there is an over-limit situation;

[0090] Trend prediction: Based on the initial judgment, the time series prediction model is used to predict the trend of parameter changes; if the predicted value also exceeds the threshold, it is determined that there is a risk of exceeding the limit.

[0091] Power flow calculation; once the risk of exceeding limits is confirmed, the amount of load to be transferred is calculated based on the power grid topology, load distribution at each node, and current operating status. The calculation formula is as follows:

[0092] ;

[0093] in, The amount of load to be transferred. This refers to the set of nodes that need to undergo load transfer due to exceeding limits. For power grid nodes The current load value; For nodes The load transfer factor is obtained from the sensitivity calculation module of the control system, taking into account factors such as the importance of the node and the transmission capacity limit of the line.

[0094] Output results: Generate a risk analysis report (including out-of-limit equipment and its corresponding pending load) and send it to the coordinating agent.

[0095] 3.2: Unit Adjustment Intelligent Agent;

[0096] An optimization model is constructed with the objectives of achieving power balance in the power grid and minimizing economic costs. Objective function: for:

[0097] ;

[0098] in, For the number of generator sets, For the first The power generation cost function of a generator set, For the first The active power of the generator sets; For the number of load nodes, and They are nodes Maximum load and actual load; The number of frequency monitoring points. For monitoring points frequency, Rated power; and These are the weighting coefficients.

[0099] Constraints include power balance constraints. ;

[0100] Generator output upper and lower limit constraints ;in, and The first The lower and upper limits of the generator set's output.

[0101] The optimization model is solved using the interior point method or genetic algorithm to obtain the unit adjustment scheme, which is then sent to the coordinating agent.

[0102] 3.3: Load Transfer Agent;

[0103] The load transfer path is searched using the shortest path algorithm in graph theory. For the power grid topology graph... ;in, For a set of nodes, Let it be the set of edges. Represents a node and The edge between, edge The weight is determined by factors such as line impedance and transmission capacity limitations.

[0104] The shortest path search algorithm is used to search for load transfer paths based on edge weights. After finding the shortest path, the load transfer evaluation model is used to calculate the solution score. The load transfer evaluation model is expressed as follows:

[0105] ;

[0106] in, To score points for the solution, For voltage drop, For line loss, Due to capacity limitations for resale; , and The weights are respectively for voltage drop, line loss, and transfer capacity limit.

[0107] The optimal load transfer scheme is selected by the scheme score (maximum value), and detailed load transfer operation instructions (such as "close the xx station switch and disconnect the xx station disconnector") are sent to the coordinating agent.

[0108] The weighting coefficients for voltage drop, line loss, and transfer capacity limits can be adjusted according to the actual operational needs of the power grid. Generally, if the power grid has high requirements for voltage stability, the weighting coefficients can be adjusted accordingly. Set it to 0.4-0.6; if you are more focused on reducing line loss, A value of 0.3-0.5 is acceptable; in areas with large load fluctuations, It can be set to 0.2-0.4. This satisfies the requirement. .

[0109] 3.4: Coordinating intelligent agents;

[0110] Information gathering; receiving risk analysis reports, unit adjustment plans, and load transfer plans;

[0111] Priority scheduling; determine the task execution order between the emergency control phase and the mode adjustment phase; if it is the emergency control phase, apply the unit adjustment plan; if it is the mode adjustment phase, apply the load transfer plan.

[0112] A preliminary complete handling plan is formed, and the finalized handling plan is sent to the security verification intelligent agent.

[0113] 3.5: Security Verification Intelligent Agent;

[0114] As a step in this embodiment, the security verification agent receives the handling plan output in 3.4 and adjusts the power grid state in the simulation environment (such as adjusting unit output, switching switches, and transferring loads according to the handling plan); based on the adjusted power grid state, it establishes power flow equations and performs Newton-Raphson power flow calculations.

[0115] Establish active power equations and reactive power equations for each node of the power grid. For each node... Its active power and reactive power They are respectively:

[0116] ;

[0117] ;

[0118] in, and For nodes and voltage amplitude, and For the nodal admittance matrix elements, The phase angle difference of the node voltage. This represents the number of nodes.

[0119] The power flow distribution of the power grid is obtained by using the Newton-Raphson method to calculate the power flow by iteratively solving the balance equations of active and reactive power at the nodes.

[0120] Determine whether there are risks to the power grid under the proposed solution (such as line overload, voltage exceeding limits, or cross-sectional power flow exceeding limits). If there are risks, analyze the risk type, severity, and scope of impact in detail, and generate a safety verification report to feed back to the coordinating agent. If the solution passes the verification, feed back the results to the control execution agent for execution.

[0121] S4: The fault handling effect evaluation and feedback optimization module performs a quantitative evaluation of the fault handling effect based on a pre-set evaluation index system after the fault handling is completed.

[0122] The evaluation index system includes four indicators: fault handling efficiency, power grid restoration efficiency, resource utilization efficiency, and reduction in economic losses. The calculation method is as follows:

[0123] Fault handling efficiency; the fault handling time is calculated by the time difference between the fault occurrence time and the fault confirmation and elimination time; the ratio of the fault handling time to the historical handling time of the same type of fault is used as the fault handling efficiency.

[0124] Power grid restoration efficiency; calculated based on the restoration time, restored load, and total load before the fault in each area, and expressed as:

[0125] ;

[0126] In the formula, To improve power grid restoration efficiency; Indicates the region The amount of power supply to be restored Indicates the total load before the fault; The weighted average of the power restoration time for each area. Indicates the region The time required to restore power. For regional weights.

[0127] Resource utilization efficiency; calculated based on the resources invested and the amount of power restored during fault handling, expressed as:

[0128] ;

[0129] In the formula, For resource utilization rate; This indicates that power has been restored. Indicates equipment Usage duration; This refers to the total amount of resources invested.

[0130] The reduction in economic losses is calculated by estimating the potential economic losses that might occur without effective fault handling measures, such as power outage losses for users and equipment damage losses, and then calculating the actual economic losses that occur. The difference between the two is the reduction in economic losses.

[0131] As an optional implementation step in the embodiment, the four types of indicators are normalized to construct a comprehensive performance evaluation function, output a quantitative report of this fault handling (including the four types of indicators + comprehensive score), and feed the evaluation results back to the large model inference and task decomposition steps and the agent collaborative execution module for model retraining, policy optimization and task allocation adjustment.

[0132] like Figure 3 As shown, Figure 3This is a system architecture diagram of a power grid fault handling system. As an embodiment of this application, a power grid fault handling system based on a large model and multiple agents is provided; the system is implemented using the method described in the first aspect of this invention, including a fault perception and information aggregation module, a large model reasoning and task decomposition module, an agent collaborative execution module, and a handling effect evaluation and feedback optimization module.

[0133] The fault perception and information aggregation module retrieves real-time equipment operation data, network topology information, and meteorological environmental parameters through the power grid dispatch automation system and meteorological information system; it uses standardized data interface protocols to achieve synchronous data transmission and acquisition from various systems; and it presets numerical threshold rules in the built-in data verification module to automatically identify and filter out abnormal values ​​that exceed the normal fluctuation range, and classify and archive them to the data warehouse.

[0134] The large model reasoning and task decomposition module performs semantic understanding and logical analysis on power grid fault information. Based on the reasoning results, it decomposes the complex fault handling process into multiple interrelated sub-tasks. The intelligent agent collaborative execution module combines the power grid topology and real-time operation data to dynamically optimize the task execution order and resource allocation strategy. After the fault handling is completed, the handling effect evaluation and feedback optimization module performs a quantitative evaluation of the fault handling effect based on a pre-set evaluation index system.

[0135] A power grid fault handling system based on a large model and multiple agents includes a perception layer, a data layer, an intelligent decision-making layer, an intelligent execution layer, an application layer, and a feedback optimization layer.

[0136] The perception layer consists of existing power grid monitoring equipment and edge computing nodes. The monitoring equipment collects information such as the operating parameters of faulty equipment, power grid topology, and meteorological environmental data in real time. The edge computing nodes perform preliminary cleaning and preprocessing of the collected data and then upload key fault information to the data layer.

[0137] The data layer includes a data warehouse and a data governance module. The data warehouse uses distributed storage technology to store and manage massive amounts of fault data, while the data governance module standardizes the data, constructs a power grid fault knowledge graph, and integrates data such as equipment information, historical fault cases, and emergency response plans, providing data support for large-scale model reasoning and agent decision-making.

[0138] The intelligent decision-making layer comprises a large-scale model inference module and a task decomposition module. The large-scale model inference module integrates large-scale model training and optimization algorithms with knowledge graph embedding algorithms. It selects a large-scale language model adapted to the power industry, and through training, acquires a deep understanding and logical reasoning ability for power grid fault information. This enables it to infer the root cause and scope of impact of faults, and generate fault handling procedures and strategies. The task decomposition module parses the handling procedures generated by the large-scale model, breaks them down into specific sub-tasks, and allocates them according to the tasks and the capabilities of the intelligent agent.

[0139] The intelligent execution layer comprises risk analysis agents, unit adjustment agents, load transfer agents, coordination agents, safety verification agents, and control execution agents. Each agent executes corresponding fault handling tasks according to its task assignment and achieves information exchange and collaborative operation through communication protocols, implementing corresponding theoretical formulas and methods.

[0140] The application layer provides a visual operation interface for power grid dispatchers and maintenance personnel, displaying data such as real-time fault information, handling process progress, and intelligent agent execution status. It supports manual intervention functions, allowing maintenance personnel to adjust handling strategies and issue instructions to intelligent agents when the system automatically handles anomalies or when human decision-making is required.

[0141] The feedback optimization layer collects data and results during the fault handling process, evaluates and analyzes the inference accuracy, agent execution efficiency, and collaborative effect of the large model based on the evaluation index system, and optimizes the training data and algorithm of the large model based on the evaluation results, and adjusts the agent functions and collaborative mechanisms.

[0142] As an embodiment of this application, a terminal is disclosed, including a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to perform the steps of the method described in the first aspect of the present invention.

[0143] As one embodiment of this application, it relates to a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method described in the first aspect of the invention.

[0144] 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 the technical solutions of the present invention still include content that can be modified or equivalently replaced in the specific implementation of the present invention.

Claims

1. A power grid fault handling method based on a large model and multiple agents, characterized in that, The method comprises the following steps: Collecting power grid multi-source data to form power grid fault information; the multi-source data includes fault time text information, real-time operation data, power grid topology information and environmental variables; Performing semantic analysis on the power grid fault information to construct a power grid knowledge graph; the entity set includes power grid equipment, stations and topology nodes; the relationship set includes equipment affiliation, topology connection and protection association; the power grid fault information is converted into structured fault features, including fault equipment identification, fault type and fault equipment state; the fault cause and associated equipment are determined based on entity relationship analysis of the power grid knowledge graph; Using a power domain large model to generate a fault handling strategy based on the structured fault features and historical fault case matching results, and to decompose the fault handling strategy into an executable subtask sequence, and to clearly define the input, output and execution sequence of each subtask; Building multiple agents to cooperatively control multiple subtasks of the fault handling strategy; The cooperative mechanism of the multiple agents includes: a risk analysis agent for generating a risk analysis report based on the real-time operation state of the power grid, including the over-limit equipment and the corresponding load to be transferred, and sending the report to a coordination agent; a unit adjustment agent for responding to the risk analysis agent, generating a unit adjustment scheme based on multiple optimization objectives of the power grid, and sending the scheme to the coordination agent; a load transfer agent parallel to the unit adjustment agent, for searching a load transfer path based on the power grid topology; For the power grid topology graph, the shortest path based on edge weight is searched as the load transfer path, and the influencing factors of the edge weight include line impedance and transmission capacity; the load transfer path with the maximum transfer scheme score is selected as the optimal transfer scheme, and the score is the inverse of the transfer cost; the transfer cost is the weighted sum of voltage drop, line loss and transfer capacity; the load transfer path with the maximum transfer scheme score is sent to the coordination agent as the optimal transfer scheme; The coordination agent is used to receive the risk analysis report, the unit adjustment scheme and the load transfer scheme, determine the task execution sequence after summarizing, and generate a fault handling scheme; A safety check agent is used to perform risk checking on the fault handling scheme; if the handling scheme passes the risk checking, it is sent to a control execution agent; if it does not pass, the checking result is fed back to the coordination agent; the control execution agent is used to issue control instructions to the power grid equipment; After the execution of the fault handling strategy is completed, a multi-dimensional evaluation index is set to quantitatively evaluate the fault handling strategy.

2. The power grid fault handling method based on a large model and multiple agents according to claim 1, wherein The execution steps of the risk analysis agent include: Real-time acquisition of power grid operation parameters, equipment rated parameters and safety operation thresholds of operation parameters; calculation of the mean value of the power grid operation parameters in a sliding window; if the difference between the power grid operation parameters and the corresponding mean value exceeds the threshold, it is preliminarily determined that there is an over-limit situation; Based on the preliminary determination, the time series prediction method is used to obtain the predicted value of the operation parameter; if the predicted value exceeds the threshold, it is determined that there is an over-limit risk. When it is confirmed that there is an over-limit risk, the current load values of all nodes affected by the over-limit risk and requiring load transfer are weighted and summed to obtain the amount of load to be transferred; A risk analysis report is generated, including the over-limit equipment and its corresponding amount of load to be transferred.

3. The power grid fault handling method based on a large model and multiple agents according to claim 1, characterized in that, the execution step of the unit adjustment agent comprises: an optimization model is constructed with the goal of power grid power balance and economic cost minimization, and the objective function includes an economic cost term, a load balance term, and a frequency stability term; the economic cost term is the total generation cost of all generating units, the load balance term is the supply-demand deviation penalty of all load nodes, and the frequency stability term is the frequency deviation penalty of all frequency monitoring points; the constraints of the objective function include power balance constraints and upper and lower limits of generating unit output; the unit adjustment scheme is obtained by solving the optimization model.

4. The power grid fault handling method based on a large model and multiple agents according to claim 1, characterized in that, the multi-dimensional evaluation index includes fault handling efficiency, power grid recovery efficiency, resource utilization efficiency, and economic loss reduction amount; wherein the fault handling time is calculated by the time difference between the fault occurrence time and the fault confirmation elimination time, and the fault handling efficiency is the ratio of the fault handling time to the historical handling time of the same type of fault; the power grid recovery efficiency value is calculated by taking the proportion of the total amount of recovered load to the total load before the fault as the numerator and the weighted average of the recovery power supply time of each region in the power grid as the denominator; the resource input total amount is calculated by counting the number and usage time of the equipment during the fault handling process, and the resource actual output is the amount of recovered power supply; the resource utilization efficiency is the ratio of the resource actual output to the resource input total amount; the economic loss reduction amount is the difference between the estimated economic loss under the condition of not taking the handling measures and the actual economic loss under the condition of taking the fault handling strategy.

5. A large model and multi-agent based power grid fault handling system, performing the power grid fault handling method of any one of claims 1-4, characterized in that, The system comprises: a fault perception and information aggregation module for collecting power grid multi-source data to form structured power grid fault information; the multi-source data includes fault time text information, real-time operation data, power grid topology information, and environmental variables; a large model reasoning and task decomposition module for performing semantic analysis on the power grid fault information, generating a fault handling strategy through a power field large model, and decomposing it into an executable subtask sequence to clearly define the input and output and execution sequence of each subtask; an agent collaborative execution module for constructing multiple agents to collaboratively control and execute multiple subtasks of the fault handling strategy; the agents include a risk analysis agent, a unit adjustment agent, a load transfer agent, a coordination agent, a safety verification agent, and a control execution agent; a handling effect evaluation and feedback optimization module for setting multi-dimensional evaluation indexes after the execution of the fault handling strategy to quantitatively evaluate the fault handling strategy.

6. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The computer program is loaded into the processor to realize the power grid fault handling method according to any one of claims 1-4.

7. A computer-readable storage medium storing a computer program, wherein the computer program comprises the following steps of: receiving a request for a resource from a client; determining whether the client is authorized to access the resource; and if the client is authorized to access the resource, providing the resource to the client. The computer program, which is executed by a processor, implements the power grid failure handling method according to any one of claims 1-4.

Images