Large natural language model-based power fault handling method, apparatus, and system, and computer program
By converting binary data from power monitoring systems into text and utilizing large natural language models for automated fault handling, the problems of low efficiency and reliability in power fault handling are solved, enabling rapid and accurate fault analysis and handling.
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- SHANGHAI SHANYUAN ELECTRONICS SCI & TECH CO LTD
- Filing Date
- 2025-10-30
- Publication Date
- 2026-07-23
Smart Images
Figure CN2025131126_23072026_PF_FP_ABST
Abstract
Description
Power Fault Handling Methods, Devices, Systems, and Computer Programs Based on Natural Language Models Technical Field
[0001] This disclosure relates to the field of power monitoring technology, and in particular to a power fault handling method, device, system and computer program based on a large natural language model. Background Technology
[0002] Power supply systems are the fundamental guarantee system for production in industrial and mining enterprises, and power monitoring systems are an important means to ensure the safe operation of power supply systems. A typical characteristic of power monitoring systems is the high density of monitored objects. Currently, they mainly rely on the professional skills and expertise of dispatchers to analyze and make decisions based on monitoring data. In the event of a power outage, the problem must be resolved and power restored quickly. If dispatchers make a misjudgment, serious consequences can result, not only failing to restore power quickly but also potentially escalating the accident and causing greater losses.
[0003] With the advancement of artificial intelligence applications across industries, the power sector has introduced AI technology, and large-scale natural language processing models (hereinafter referred to as "large models") are now available to provide solutions for handling power outages. In use, dispatchers ask questions based on the fault information, and the large model, based on pre-learned knowledge and the context of the question-and-answer environment, provides fault handling solutions to assist dispatchers in decision-making and execution.
[0004] However, the interaction with the large natural language model is achieved through natural language question and answer, which not only requires a long interaction time but also places high demands on the professionalism of the schedulers. There is still room for improvement in the efficiency and reliability of fault handling. Summary of the Invention
[0005] This invention provides a power fault handling method, device, system, and computer program based on a large natural language model, which addresses the problems of low efficiency and reliability in existing power fault handling methods.
[0006] To solve the above-mentioned technical problems, the present invention is implemented as follows:
[0007] On the one hand, this invention provides a power fault handling method based on a large natural language model, including:
[0008] Acquire real-time binary event data generated by the power monitoring system;
[0009] Convert the real-time event data within the preset time window into text materials;
[0010] Submit the text materials, corresponding questions and answer requirements to the large model, and receive the text report returned by the large model;
[0011] The content of the text report is formatted into code adapted to the power monitoring system for reading by the power monitoring system.
[0012] On the other hand, the present invention provides a power fault handling device based on a large natural language model, comprising:
[0013] The monitoring data acquisition module is used to acquire real-time binary event data generated by the power monitoring system.
[0014] The text conversion module is used to convert the real-time event data within a preset time window into text materials;
[0015] The large model interaction module is used to submit the text materials, corresponding questions and answer requirements to the large model, and receive the text report returned by the large model;
[0016] The text parsing module is used to format the content of the text report into code adapted to the power monitoring system for reading by the power monitoring system.
[0017] On the other hand, the present invention also provides a power fault handling system based on a large natural language model, characterized in that it includes:
[0018] Power monitoring systems are used to monitor the power supply network and generate real-time event data;
[0019] An intermediary agent is used to acquire real-time binary event data generated by the power monitoring system; convert the real-time event data within a preset time window into text materials; submit the text materials, corresponding questions and answer requirements to the large model, and receive a text report returned by the large model; format the content of the text report into code adapted to the power monitoring system for the power monitoring system to read.
[0020] The technical solution of this invention interacts with a large natural language model through an intermediary agent. It converts real-time fault data from the power monitoring system, combined with real-time data and relevant information such as power grid structure, into natural language that the large model can understand. This data is automatically submitted to the large model, which then provides solutions to the relevant problems. The intermediary agent then parses and categorizes these solutions before providing them to dispatchers for decision support. Applying this technical solution significantly improves the accuracy of fault analysis and handling, reduces human error, and enhances the reliability and stability of the power supply system. Attached Figure Description
[0021] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0022] Figure 1 is a schematic diagram of an application scenario of the power failure handling solution provided in the embodiments of this disclosure;
[0023] Figure 2 is a flowchart of a power fault handling method provided in an embodiment of this disclosure;
[0024] Figure 3 is a schematic diagram of data flow in an embodiment of this disclosure;
[0025] Figure 4 is a structural block diagram of the power fault handling device provided in the embodiment of this disclosure;
[0026] Figure 5 is a structural block diagram of the power fault handling system provided in the embodiments of this disclosure. Detailed Implementation
[0027] To enable those skilled in the art to better understand the technical solutions of this disclosure, the technical solutions of this disclosure will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this disclosure. Obviously, the described embodiments are merely some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this disclosure. Furthermore, for clarity, parts unrelated to the described exemplary embodiments have been omitted from the drawings.
[0028] In this specification, it should be understood that terms such as "comprising" or "having" are intended to indicate the presence of features, figures, steps, behaviors, components, portions, or combinations thereof disclosed in this disclosure, and are not intended to exclude the possibility of one or more other features, figures, steps, behaviors, components, portions, or combinations thereof being present or added. It should also be noted that, unless otherwise specified, embodiments and features within embodiments of this disclosure can be combined with each other.
[0029] Figure 1 is a schematic diagram of the application scenario of the power fault handling solution provided in the embodiments of this disclosure.
[0030] As shown in Figure 1, this application scenario is a coal mine power supply system. The ground monitoring platform communicates with the power distribution devices and electrical equipment in the underground substation through the coal mine power supply system ring network, forming a power monitoring system. The ground monitoring platform grasps the real-time status of the power grid through the status of various switches, telemetry data, and remote signaling data, and performs remote monitoring. When a power supply failure occurs, the large model performs fault analysis through the monitoring information provided by the intermediate agent and provides corresponding fault handling solutions.
[0031] The intermediate agent in Figure 1 carries the power fault handling method, apparatus, or computer program provided in this embodiment of the disclosure, runs on the ground, and communicates with the ground monitoring platform and the large model. Preferably, the intermediate agent and the monitoring platform's operation and maintenance program run on the same application terminal or server. The large model, as an independent unit, can run on a separate server, allowing multiple such intermediate agents to interact with it.
[0032] The following describes an embodiment of the power fault handling method provided in this disclosure, with reference to Figure 2.
[0033] Figure 2 is a flowchart of a power fault handling method provided in an embodiment of this disclosure. From a programming perspective, the entity executing the process can be a program mounted on an application server or application terminal. It is understood that this method can be executed by any device, equipment, platform, or cluster of devices with computing and processing capabilities.
[0034] As shown in Figure 2, the method in this embodiment includes operations S210 to S240.
[0035] S210: Acquire binary real-time event data generated by the power monitoring system.
[0036] Existing power monitoring systems generate binary data from monitoring the power grid, including switch status, remote signaling, telemetry, and electricity consumption data. The information displayed on the visual operation and maintenance program is obtained after corresponding parsing and conversion.
[0037] The intermediate agent can read real-time binary event data generated by the power monitoring system via shared memory. This shared memory approach enables rapid data interaction with the power monitoring system, reducing processing latency for real-time data. In the event of a power outage, the intermediate agent can immediately access the corresponding real-time fault data.
[0038] S220: Convert real-time event data within a preset time window into text material.
[0039] For real-time event data, a sliding time window algorithm is used, which processes only the data within the time window at a time. Specifically, this may include the following steps: filtering and normalizing the real-time event data; starting the sliding time window algorithm on the preprocessed real-time event data; and performing data textification processing on the real-time event data within the time window to obtain the corresponding text materials.
[0040] The sliding time window size can be set to 6-8 seconds, with a maximum of 15 seconds, and the sliding step is one window. The sliding window setting ensures timely detection of fault events.
[0041] Real-time event data undergoes filtering and normalization preprocessing, including removing unnecessary information while retaining switch protection activation and tripping information. Unnecessary information includes waveform recording activation and setting modification information. The filtered real-time event data is then normalized to mask differences between different protectors. Since many protectors in the power grid come from different manufacturers, their output and transmission data vary; normalization eliminates problems caused by these differences.
[0042] Data textification of real-time event data within a time window can include: clearing sliding time windows that do not contain events; for each event in a sliding time window that does contain events, converting the corresponding binary event data into text and adding auxiliary data parsing. Auxiliary data includes substation name, switch name, switch type, event name, event type, event category, and occurrence time. Switch type includes: incoming line switch, outgoing line switch, tie switch, bus tie switch, etc.; event type includes: short circuit protection, overcurrent protection, undervoltage protection, differential protection, intelligent backup, etc.; event category includes: protection warning signal, fault protection trip, etc. In other words, according to the power system's preset data specifications, data at the corresponding locations is extracted and values are assigned to the fields of the auxiliary data. During the textification process, known network topology relationships and substation structure information can also be used to obtain text content that is packaged into text materials after textification.
[0043] Furthermore, the data textification processing of real-time event data within the time window may include: textification of the basic information of each event within the time window, and the processed text content includes the substation name, switch name, switch type, event name, event type, event category, and occurrence time.
[0044] Furthermore, the data textification processing for real-time events within the time window can also include: textifying the hierarchical relationships of each switch involved in each event. The processed text content includes the substation name, switch name, the name of the switch's parent switch, and the name of the switch's child switch. The intermediate agent provides the large model with the text containing the hierarchical relationships of each switch involved. For example, the parent of the high-voltage switch #3 of the central substation is the high-voltage switch #1 of the central substation; the parent of the high-voltage switch #1 of the central substation is the high-voltage switch #12 of the ground substation, etc.
[0045] Furthermore, the data textification of real-time event data within the time window can also include: textifying the structural information of each substation involved in each event. The processed text content includes the substation name, the number and names of incoming line switches, and the name of the bus tie switch. To provide more detailed information for large-scale model analysis, the intermediate agent provides the large-scale model with the structural information of each substation involved, such as: the central substation has two incoming lines, #1 and #11, and one bus tie switch, #10, etc.
[0046] S230: Submit the text materials, corresponding questions and answer requirements to the large model, and receive the text report returned by the large model.
[0047] Provide the aforementioned textual materials to the large model. Based on this, pose corresponding questions and provide the expected answers, then wait for the large model's response. For example, you could ask: "Based on the above information, please provide a fault analysis report for this fault, including the cause of the fault, fault location, fault isolation plan, and power restoration plan." After completing the fault analysis, the large model will provide a text report in the form of a fault analysis report.
[0048] S240: Format the content of the text report into code adapted for the power monitoring system so that the power monitoring system can read it.
[0049] The intermediate agent analyzes the text report of the fault analysis returned by the large model, and formats and categorizes the report, specifically including:
[0050] Based on the text report, analyze the cause of the fault and give the cause of the fault, such as: short circuit fault, phase A tripping;
[0051] Based on the text report, the fault location is analyzed and the fault point information that caused the fault is given, for example: the line below No. 3 high-voltage switch of the central substation;
[0052] Based on the text report, the fault isolation scheme is analyzed, and it is given which switches cannot be closed in order to isolate the fault, for example: the No. 3 high-voltage switch of the central substation.
[0053] Based on the text report, the power restoration plan is analyzed, and the switches that should be closed in sequence to restore power are given, for example: the No. 1 high-voltage switch of the central substation is closed; the No. 2 high-voltage switch of the central substation is closed; the No. 4 high-voltage switch of the central substation is closed.
[0054] Therefore, the intermediate agent extracts the text content of the fault cause, fault location, fault isolation plan and power restoration plan based on the text report; and formats the extracted text content into code adapted to the power monitoring system and writes it into shared memory.
[0055] The results provided to the power monitoring system include the following: fault cause, indicating the cause of the fault; fault location, providing information on the fault point that caused the fault; fault isolation plan, indicating which switches should not be closed in order to isolate the fault; and power restoration plan, indicating which switches should be closed in sequence to restore power supply.
[0056] After the intermediate agent writes the above formatting and classification results into shared memory, the power monitoring system automatically reads them and displays them in categories on the operation and maintenance platform, providing auxiliary decision support for dispatchers to identify fault points, formulate fault isolation plans and power restoration plans.
[0057] Figure 3 is a schematic diagram of data flow in the method of the above embodiments.
[0058] As shown in Figure 3, the intermediate agent obtains binary real-time event data from the power monitoring system. After filtering and normalization, the real-time event data enters a sliding time window. The intermediate agent performs text processing on the real-time event data within each time window and packages it into text materials by combining topology information and substation structure information. This text material, along with the proposed questions and answer requirements, is submitted to the large model, requesting the large model to perform fault analysis. The large model responds by returning a text report of the fault analysis. The intermediate agent parses and categorizes this report and provides it to the power monitoring system. The entire process is automated and requires no manual intervention. When a power supply fault occurs, the intermediate agent can detect the fault event through the sliding time window and automatically use the large model to promptly output a fault handling plan.
[0059] According to the power fault handling method in this embodiment, by interacting with a natural language processing (NLP) model through an intermediate agent, real-time fault data from the power monitoring system, combined with relevant information such as power grid topology and substation structure, can be converted into natural language and automatically submitted to the NLP model. The NLP model then provides solutions to the relevant problems, which are further parsed by the intermediate agent and provided to dispatchers for decision support. Applying the technical solution of this invention can significantly improve the accuracy of fault analysis and handling, reduce human error, and improve the reliability and stability of the power supply system.
[0060] The above are embodiments of the fault handling method provided in this disclosure. Correspondingly, this disclosure also provides an embodiment of a power fault handling device 400 based on a large natural language model.
[0061] Figure 4 is a structural block diagram of the power fault handling device provided in the embodiments of this disclosure.
[0062] As shown in Figure 4, the power fault handling device 400 includes a monitoring data acquisition module 410, a text conversion module 420, a large model interaction module 430, and a text parsing module 440. The power fault handling device 400 can be implemented through software, hardware, or a combination of both.
[0063] The monitoring data acquisition module 410 is used to acquire binary real-time event data generated by the power monitoring system.
[0064] The text conversion module 420 is used to convert the real-time event data within a preset time window into text materials.
[0065] The large model interaction module 430 is used to submit text materials, corresponding questions and answer requirements to the large model, and receive text reports returned by the large model.
[0066] The text parsing module 440 is used to format the content of the text report into code adapted to the power monitoring system so that the power monitoring system can read it.
[0067] The fault handling device according to this embodiment interacts with a large natural language model through an intermediate agent, reducing reliance on manual intervention and improving the efficiency and accuracy of fault handling.
[0068] This disclosure also provides an embodiment of a power fault handling system 500 based on a large natural language model. Figure 5 is a structural block diagram of the power fault handling system.
[0069] As shown in Figure 5, the power fault handling system 500 includes a power monitoring system 510 and an intermediate agent 520.
[0070] The power monitoring system 510 is used to monitor the power supply network and generate real-time event data;
[0071] The intermediate agent 520 is used to: acquire binary real-time event data generated by the power monitoring system; convert real-time event data within a preset time window into text materials; submit the text materials, corresponding questions and answer requirements to the large model, and receive the text report returned by the large model; format the content of the text report into code adapted to the power monitoring system so that the power monitoring system 510 can read it.
[0072] The power monitoring system 510 and the intermediate agent 520 can be installed on the same terminal device or server and can share memory data. By exchanging data through shared memory, data access efficiency and fault handling efficiency can be improved.
[0073] Another aspect of this disclosure provides a computer program that, when executed by a processor, causes the processor to perform the various methods described above.
[0074] The foregoing has described specific embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0075] The units or modules described in the embodiments of this disclosure can be implemented in software or programmable hardware. The described units or modules can also be located in a processor, and the names of these units or modules do not necessarily constitute a limitation on the unit or module itself.
[0076] The various embodiments in this disclosure are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device, system, and computer program embodiments are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.
[0077] The apparatus, system, and computer program and method provided in the embodiments of this disclosure are corresponding to each other. Therefore, the apparatus, system, and computer program also have similar beneficial technical effects as the corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, they will not be repeated here.
[0078] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features disclosed in this disclosure that have similar functions.
Claims
1. A power fault handling method based on a large natural language model, characterized in that, include: Acquire real-time binary event data generated by the power monitoring system; Convert the real-time event data within the preset time window into text materials; Submit the text materials, corresponding questions and answer requirements to the large model, and receive the text report returned by the large model; The content of the text report is formatted into code adapted to the power monitoring system for reading by the power monitoring system.
2. The method according to claim 1, characterized in that, The acquisition of binary real-time event data generated by the power monitoring system includes: Real-time binary event data generated by the power monitoring system is read via shared memory.
3. The method according to claim 1, characterized in that, The step of converting the real-time event data within a preset time window into text material includes: The real-time event data is filtered and normalized preprocessed. A sliding time window algorithm is applied to the preprocessed real-time event data. The real-time event data within the time window is processed into text data to obtain the corresponding text materials.
4. The method according to claim 3, characterized in that, The process of converting real-time event data within the time window into textual data includes: The basic information of each event within the time window is processed into text. The processed text content includes the substation name, switch name, switch type, event name, event type, event category, and occurrence time.
5. The method according to claim 4, characterized in that, Also includes: The hierarchical relationship of each switch involved in each event is processed into text. The processed text content includes the substation name, switch name, the name of the switch's parent switch, and the name of the switch's child switch.
6. The method according to claim 4, characterized in that, Also includes: The structural information of each substation involved in each event is processed into text, and the processed text content includes the substation name, the number and name of incoming line switches, and the name of the bus tie switch.
7. The method according to claim 1, characterized in that, The code that formats the content of the text report to be compatible with the power monitoring system includes: Based on the text report, extract the text content of the fault cause, fault location, fault isolation plan and power restoration plan; The extracted text content is formatted into an encoding suitable for the power monitoring system and written into shared memory.
8. The method according to claim 3, characterized in that, The filtering and normalization preprocessing of the real-time event data includes: Unnecessary event information is removed from the real-time event data, while switch protection activation information and switch tripping information are retained. The filtered real-time event data is then normalized to mask the differences between different protectors.
9. A power fault handling device based on a large natural language model, characterized in that, include: The monitoring data acquisition module is used to acquire real-time binary event data generated by the power monitoring system. The text conversion module is used to convert the real-time event data within a preset time window into text materials; The large model interaction module is used to submit the text materials, corresponding questions and answer requirements to the large model, and receive the text report returned by the large model; The text parsing module is used to format the content of the text report into code adapted to the power monitoring system for reading by the power monitoring system.
10. A power fault handling system based on a large natural language model, characterized in that, include: Power monitoring systems are used to monitor the power supply network and generate real-time event data; An intermediary agent is used to obtain real-time binary event data generated by the power monitoring system; The real-time event data within the preset time window is converted into text material; the text material, corresponding questions and answer requirements are submitted to the large model, and a text report returned by the large model is received; the content of the text report is formatted into code adapted to the power monitoring system for the power monitoring system to read.
11. A computer program that, when executed by a processor, causes the processor to perform the method as described in any one of claims 1 to 8.