A workflow implementation method and system for power grid monitoring alarm information disposal based on a large model

Abstract

This invention belongs to the field of power system control and operation, and relates to a workflow implementation method and system for handling power grid monitoring alarm information based on a large model. It includes: a data acquisition step, collecting alarm information and storing it in a structured manner; a business rule filtering step, filtering alarm information and identifying fault events using preset rules; a power grid equipment knowledge base retrieval step, matching the equipment information corresponding to the event with the pre-built knowledge base to retrieve equipment parameters; a message generation step, inputting the filtered alarm information and equipment parameters into the large model to generate fault descriptions and handling suggestions; a daily report generation step, classifying and aggregating alarms within a set time period, and combining historical daily reports with the large model to generate a comprehensive alarm daily report; a message push step, pushing the fault description, handling suggestions, and daily report to the dispatch terminal; and a query and display step, displaying and supporting status confirmation and query. This invention achieves intelligent handling and in-depth analysis of alarm information, improving power grid monitoring efficiency and decision support capabilities.

Description

Technical Field

[0001] This invention belongs to the field of power system control and operation, and relates to a workflow implementation method and system for handling power grid monitoring alarm information based on a large model. Background Technology

[0002] With the continuous development of new power systems based on new energy sources, a large number of devices with highly electronic power characteristics, such as distributed photovoltaics, energy storage devices, and electric vehicle charging piles, are being connected to the distribution network. The complex, highly random, and numerous operating characteristics of these devices have dramatically increased the difficulty of monitoring the distribution network's operational status and identifying faults. Alarm information generated during grid operation is characterized by its massive volume, high dimensionality, and multiple sources. Traditional methods relying on manual interpretation by dispatchers are no longer sufficient for timely and accurate event response and decision-making control, necessitating the construction of a new generation of intelligent alarm information processing systems.

[0003] In recent years, large language models have achieved significant breakthroughs in natural language processing, multimodal information fusion, and logical reasoning tasks. Their excellent contextual understanding, generation capabilities, and knowledge generalization performance provide new technical pathways for automating decision-making in complex systems. However, current large model technology is still mainly focused on general-domain text processing. Its application in specialized fields with high reliability requirements, such as the power industry, has not yet formed a systematic engineering paradigm. Especially in power grid monitoring operations with high real-time requirements and rigorous processing logic, how to deeply integrate the semantic understanding and reasoning capabilities of large models with power system operation rules is a key challenge for achieving business substitution. Summary of the Invention

[0004] This invention aims to address the problems of low efficiency and insufficient intelligence in the manual handling of power grid alarm information in existing technologies. It provides a workflow implementation method and system for power grid monitoring alarm information handling based on a large model, so as to realize real-time handling and in-depth analysis of alarm information and improve the comprehensiveness and foresight of power grid monitoring.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] The first aspect of this invention is to provide a workflow implementation method for handling power grid monitoring alarm information based on a large model, comprising the following steps:

[0007] Data acquisition steps: Real-time acquisition of alarm information generated by the power dispatching system, structured parsing, and storage in the database;

[0008] Business rule filtering steps: Read alarm information from the database, filter and comprehensively analyze the alarm information according to the preset business rules, and identify the alarm information corresponding to the events that meet the rules;

[0009] Power grid equipment knowledge base retrieval steps: Match the equipment information corresponding to the events identified in the business rule filtering step with the pre-built power grid equipment knowledge base to retrieve the corresponding equipment parameter information;

[0010] Message generation step: Input the alarm information filtered by the business rule filtering step and the equipment parameter information obtained by the power grid equipment knowledge base retrieval step into the big language model to generate a natural language fault description and handling suggestions;

[0011] Daily report generation steps: Collect all alarm information within the set time period after the data collection steps, classify and aggregate it, combine it with historical daily report data, and call the big data model to generate a comprehensive alarm daily report;

[0012] Message push step: Push the fault description and handling suggestions generated in the message generation step, as well as the alarm summary daily report generated in the daily report generation step, to the dispatcher's terminal;

[0013] The query and display process involves reading the data generated by the message generation step and the daily report generation step from the database, displaying it through the front-end interface, and allowing users to confirm and query the alarm status.

[0014] Furthermore, in the data acquisition step, the alarm files written by the power dispatching system are obtained through a file listening service, and the key fields in the alarm information are parsed, extracted, and stored in the database.

[0015] Furthermore, in the business rule filtering step, alarm information within a set time window is filtered and specific fault modes are identified based on a combination of rules set according to one or more conditions in the device type, alarm type, and time window.

[0016] Furthermore, the power grid equipment knowledge base retrieval step includes:

[0017] The specifications and parameters of the power grid equipment are vectorized in advance and stored in a vector database.

[0018] During runtime, the device information corresponding to the alarm information obtained from the business rule filtering step is vectorized, and relevant device information is retrieved from the vector database;

[0019] The search results are reordered to filter out the device parameter information with the highest matching degree.

[0020] Furthermore, in the message generation step, a natural language fault description and operation suggestions for the fault are generated by sending prompt words containing alarm information and device parameter information to the large language model.

[0021] Furthermore, the daily report generation step includes:

[0022] All alarm information within a set time period is categorized, aggregated, and statistically analyzed according to device type or alarm type;

[0023] By using historical daily data as context and interacting with a large language model, a natural language alarm daily report is generated, which includes at least one of the following: overall overview, key fault analysis, and trend prediction.

[0024] Furthermore, in the message push step, the generated fault description, handling suggestions, and alarm daily reports are pushed to the mobile terminal application used by the dispatcher, and the push delay meets the real-time requirements.

[0025] Furthermore, in the query and display step, the front-end interface displays alarm details, time, confirmation status, and alarm daily reports in real time, and supports users to modify the confirmation status and retrieve historical data by time period.

[0026] A second aspect of the present invention is to provide a workflow system for handling power grid monitoring alarm information based on a large model for implementing the above-described method, comprising:

[0027] The data acquisition module is used to collect alarm information generated by the power dispatching system in real time, perform structured parsing, and store it in the database;

[0028] The business rule filtering module is used to read alarm information from the database, filter and comprehensively analyze the alarm information according to preset business rules, and identify the alarm information corresponding to events that meet the rules.

[0029] The power grid equipment knowledge base module is used to match the equipment information corresponding to the events identified in the business rule filtering step with the pre-built power grid equipment knowledge base to retrieve the corresponding equipment parameter information.

[0030] The message generation module is used to input the alarm information filtered by the business rule filtering step and the equipment parameter information obtained by the power grid equipment knowledge base retrieval step into the big language model to generate a natural language fault description and handling suggestions.

[0031] The daily report generation module is used to collect all alarm information within a set time period after the data collection steps, classify and aggregate it, combine it with historical daily report data, and call the big data model to generate a comprehensive alarm daily report.

[0032] The message push module is used to push the fault description and handling suggestions generated in the message generation step, as well as the alarm summary daily report generated in the daily report generation step, to the dispatcher's terminal.

[0033] The query and display module is used to read the data generated by the message generation step and the daily report generation step from the database, display it through the front-end interface, and support users to confirm and query the alarm status.

[0034] Advantages and beneficial effects of the present invention:

[0035] 1. This invention deeply integrates the semantic understanding and text generation capabilities of large models with the business rules of the power system, realizing the intelligent processing of power grid alarm information, replacing manual interpretation, and improving the efficiency and accuracy of processing.

[0036] 2. This invention adopts a modular workflow design, with clear functions, closed-loop logic, and clear data flow between modules, making it easy to deploy and implement in existing power dispatching systems.

[0037] 3. This invention combines business rule filtering with large model generation to achieve real-time analysis and automatic daily report generation of massive alarm information under limited computing power, providing timely and effective decision support for dispatchers.

[0038] 4. The functional modules of this invention are clearly divided, the implementation path is well-defined, and it has good scalability and maintainability. Attached Figure Description

[0039] Figure 1 This is a block diagram of the workflow system structure of the present invention. Detailed Implementation

[0040] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. It should be noted that the embodiments described are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0041] like Figure 1 As shown, the present invention provides a workflow system for handling power grid monitoring alarm information based on a large model, which includes the following modules: data acquisition module, business rule filtering module, power grid equipment knowledge base module, message generation module, daily report generation module, message push module, and query and display module.

[0042] System Architecture Overview:

[0043] This system constructs an end-to-end, fully automated alarm processing system based on the large amount of alarm information generated by the power dispatching system. The data acquisition module is responsible for the access and standardization of alarm information; the business rule filtering module enables rapid fault assessment based on preset rules; the power grid equipment knowledge base module provides equipment parameter support for fault analysis; the message generation module and daily report generation module respectively undertake the dual tasks of real-time alarm handling and periodic in-depth analysis; the message push module promptly delivers analysis results to dispatchers; and the query and display module provides a human-computer interaction interface.

[0044] Example

[0045] This embodiment uses the alarm handling of a power distribution network in a certain region as an example to illustrate the workflow of the present invention.

[0046] 1. Data Acquisition Module

[0047] Deploy a file monitoring service on the workflow server. During operation, the power dispatching system writes all anomaly information generated by each monitoring point to a file and sends the file to a specified directory on the workflow server via SFTP.

[0048] The data acquisition module reads newly added files in a specified directory to obtain alarm information generated by the power dispatching system, extracts key information such as time, device ID, device name, and alarm details from the alarm information, deduplicates the data, processes it into a unified object, and stores it in the database of the workflow server, so that all subsequent business processes operate based on the same data model.

[0049] 2. Business Rule Filtering Module

[0050] The business rule filtering module periodically reads alarm information from the workflow server's database, for example, searching for all alarm information within the last minute. It then filters relevant alarm events based on rules set during programming to comprehensively assess the power grid fault situation.

[0051] Taking the "distribution network automation switch tripping" event as an example, the business rule is set as follows: A "switch tripping signal" alarm and an "overcurrent protection action," "zero-sequence protection action," "overcurrent fault," or "fault merging" alarm occur simultaneously for the same distribution network automation switch corresponding to the same device ID, and the time difference between the two alarms does not exceed 30 seconds. When the above conditions are met, it is determined that a tripping event has occurred in the distribution network automation switch corresponding to this device ID. The business rule filtering module then extracts the alarm information for use by subsequent modules.

[0052] 3. Power Grid Equipment Knowledge Base Module

[0053] The power grid equipment knowledge base module requires dispatchers to upload specifications and parameter documents (such as equipment manuals and technical specifications) of power grid equipment to this workflow beforehand. The program vectorizes the specifications and parameter documents using an embedded model (such as the bge-m3 model) and stores them in a local vector database file using a vector database (such as Python FAISS).

[0054] During workflow operation, the device information corresponding to the alarms obtained by the business rule filtering module is vectorized through the same embedding model. Then, multiple most relevant device information (e.g., 10) are searched from the vector database. The re-ranking model (e.g., bge-reranker-v2-m3 model) re-ranks the search results and filters out the device parameter information with the highest accuracy for use by the message generation module.

[0055] 4. Message Generation Module

[0056] The message generation module sends the alarm information filtered by the business rule filtering module and the equipment parameter information filtered by the power grid equipment knowledge base module to the locally deployed large language model (such as qwen3-30b-a3b-instruct-2507) through the API interface.

[0057] Taking the "distribution network automation switch tripping" event as an example, the prompt message is: "You are a power grid operation and maintenance expert. Please summarize the distribution network automation switch tripping event based on the following alarm content, including the time, location, and line where the tripping occurred, and provide handling suggestions based on equipment information. Alarm content: {alarm information provided by the filtering module}; Equipment information: {equipment information provided by the power grid equipment knowledge base module}."

[0058] Based on alarm information and device parameter information, the large language model generates a concise and easy-to-read fault description and operation suggestions for the fault, helping dispatchers quickly grasp the key points of the fault.

[0059] 5. Daily Report Generation Module

[0060] The daily report generation module collects all alarm information within a set time period from the data acquisition module, such as data from 8:00 yesterday to 8:00 today, a total of 24 hours. All alarm information is categorized into main network, distribution network, transformer equipment, communication equipment, etc., and alarms with the same content but different times are statistically summarized.

[0061] Simultaneously, historical daily data, such as daily reports generated in the previous 7 days, is read from the database as contextual information. The categorized and aggregated alarm information, along with the historical daily data, is sent to the locally deployed large language model (qwen3-30b-a3b-instruct-2507) via the API interface. This constructs prompt words to guide the model in generating a complete and formally worded natural language alarm report.

[0062] The prompt may include: "You are a power grid operation and maintenance expert. Please write a daily report based on the power grid equipment alarms of the past 24 hours. The report should include an overall overview, analysis of key faults, trend predictions, etc. Pay attention to events that occur frequently and have a significant impact on the safe operation of the power grid. Alarm content: {alarm information from 8:00 yesterday to 8:00 today}; Appendix - Daily alarm analysis report for the past 7 days: {daily reports generated in the previous 7 days}."

[0063] The generated daily report includes an overall overview, key fault analysis, trend forecasts, etc., and is then stored in a database for querying and push notifications.

[0064] 6. Message Push Module

[0065] The message push module takes the fault descriptions and handling suggestions output by the message generation module, as well as the alarm daily reports generated by the daily report generation module, and sends them in real time to the message sending program on the intranet via an automated script (such as the Python playwright library) for dispatchers to view quickly on their mobile terminal applications (such as the State Grid Corporation of China's dedicated office APP).

[0066] This system can generate and send relevant information in a short time (e.g., within 1 minute), meeting the real-time requirements of power grid monitoring and assisting dispatchers in handling faults in a timely manner.

[0067] 7. Query and Display Module

[0068] The query and display module mainly refers to the front-end interface deployed on the dispatcher's workstation. This interface reads data generated by the message generation module and the daily report generation module from the database, and displays alarm details, time, confirmation status, and other information, as well as the daily alarm report, to the user in real time.

[0069] Users can change the confirmation status of alarms using buttons on the interface, and can query the corresponding alarm content or daily reports by time period, which facilitates subsequent tracing and analysis.

[0070] This invention has been pre-developed in the laboratory, and its performance meets the expectations of power grid dispatchers based on existing large-scale modeling technology. Test results show that the system can accurately identify typical faults such as distribution network switch tripping, generates reasonable handling suggestions, and the daily report content meets requirements. The processing delay meets the real-time requirements of power grid monitoring, and it can perform power grid alarm analysis and handling, demonstrating promising prospects for field application.

[0071] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Those skilled in the art can make various improvements and modifications to the invention without departing from its principles, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A workflow implementation method for handling power grid monitoring alarm information based on a large model, characterized in that, Includes the following steps: Data acquisition steps: Real-time acquisition of alarm information generated by the power dispatching system, structured parsing, and storage in the database; Business rule filtering steps: Read alarm information from the database, filter and comprehensively analyze the alarm information according to the preset business rules, and identify the alarm information corresponding to the events that meet the rules; Power grid equipment knowledge base retrieval steps: Match the equipment information corresponding to the events identified in the business rule filtering step with the pre-built power grid equipment knowledge base to retrieve the corresponding equipment parameter information; Message generation step: Input the alarm information filtered by the business rule filtering step and the equipment parameter information obtained by the power grid equipment knowledge base retrieval step into the big language model to generate a natural language fault description and handling suggestions; Daily report generation steps: Collect all alarm information within the set time period after the data collection steps, classify and aggregate it, combine it with historical daily report data, and call the big data model to generate a comprehensive alarm daily report; Message push step: Push the fault description and handling suggestions generated in the message generation step, as well as the alarm summary daily report generated in the daily report generation step, to the dispatcher's terminal; The query and display process involves reading the data generated by the message generation step and the daily report generation step from the database, displaying it through the front-end interface, and allowing users to confirm and query the alarm status.

2. The method according to claim 1, characterized in that, In the data acquisition step, the alarm file written by the power dispatching system is obtained through the file listening service, and the key fields in the alarm information are parsed, extracted, and then stored in the database.

3. The method according to claim 1, characterized in that, In the business rule filtering step, alarm information within a set time window is filtered and specific fault modes are identified based on a combination of rules set according to one or more conditions in the device type, alarm type, and time window.

4. The method according to claim 1, characterized in that, The power grid equipment knowledge base retrieval steps include: The specifications and parameters of the power grid equipment are vectorized in advance and stored in a vector database. During runtime, the device information corresponding to the alarm information obtained from the business rule filtering step is vectorized, and relevant device information is retrieved from the vector database; The search results are reordered to filter out the device parameter information with the highest matching degree.

5. The method according to claim 1, characterized in that, In the message generation step, prompt words containing alarm information and device parameter information are sent to the large language model to generate a natural language description of the fault and operation suggestions for the fault.

6. The method according to claim 1, characterized in that, The daily report generation steps include: All alarm information within a set time period is categorized, aggregated, and statistically analyzed according to device type or alarm type; By using historical daily data as context and interacting with a large language model, a natural language alarm daily report is generated, which includes at least one of the following: overall overview, key fault analysis, and trend prediction.

7. The method according to claim 1, characterized in that, In the message push step, the generated fault description, handling suggestions and alarm daily reports are pushed to the mobile terminal application used by dispatchers, and the push delay meets the real-time requirements.

8. The method according to claim 1, characterized in that, In the query and display step, the front-end interface displays alarm details, time, confirmation status and alarm daily report in real time, and supports users to modify the confirmation status and retrieve historical data by time period.

9. A workflow system for processing power grid monitoring alarm information based on a large model to implement the method described in any one of claims 1 to 8, characterized in that, include: The data acquisition module is used to collect alarm information generated by the power dispatching system in real time, perform structured parsing, and store it in the database; The business rule filtering module is used to read alarm information from the database, filter and comprehensively analyze the alarm information according to preset business rules, and identify the alarm information corresponding to events that meet the rules. The power grid equipment knowledge base module is used to match the equipment information corresponding to the events identified in the business rule filtering step with the pre-built power grid equipment knowledge base to retrieve the corresponding equipment parameter information. The message generation module is used to input the alarm information filtered by the business rule filtering step and the equipment parameter information obtained by the power grid equipment knowledge base retrieval step into the big language model to generate natural language fault descriptions and handling suggestions. The daily report generation module is used to collect all alarm information within a set time period after the data collection steps, classify and aggregate it, combine it with historical daily report data, and call the big data model to generate a comprehensive alarm daily report. The message push module is used to push the fault description and handling suggestions generated in the message generation step, as well as the alarm summary daily report generated in the daily report generation step, to the dispatcher's terminal. The query and display module is used to read the data generated by the message generation step and the daily report generation step from the database, display it through the front-end interface, and support users to confirm and query the alarm status.

Images