LLM-based transformer substation patrol and maintenance intelligent dispatching method and device, medium and equipment

Abstract

The invention discloses an LLM-based transformer substation patrol and maintenance intelligent dispatching method and device, a medium and equipment. The LLM-based transformer substation patrol and maintenance intelligent dispatching method comprises the steps of extracting and generating a structured historical dispatching data set from to-be-processed historical dispatching data; expanding the historical dispatching data set based on a collaborative filtering algorithm to generate an expanded dispatching data set; automatically generating a plurality of new dispatching plan schemes in a predetermined period based on a power transformation inspection and maintenance production decision LLM model; and inputting the plurality of dispatching plan schemes into a decision supervision module for evaluation processing to obtain a final optimal dispatching plan scheme. According to the method, the expanded dispatching data set is generated through expansion of the historical dispatching data set based on the collaborative filtering algorithm, so that the acquisition cost of the data set during model training is reduced; and meanwhile, the decision supervision module is utilized to score the generated dispatching plan scheme, so that the final optimal dispatching plan scheme is obtained, and the intelligent level and the accuracy of the dispatching plan are improved.

Description

Technical Field

[0001] The present invention relates to the field of LLM technology, and in particular to a method, device, medium and equipment for intelligent dispatching of substation patrol and maintenance based on LLM. Background Art

[0002] LLM (Large Language Model) is an artificial intelligence technology based on deep learning, which aims to train large models that can process and generate natural language text. These models usually contain tens of billions or even hundreds of billions of parameters. They are pre-trained on large-scale text data and fine-tuned on specific tasks, so that they have powerful language understanding and generation capabilities. One of the main differences between LLM and small language models is that LLM has emergent capabilities. Emergent capabilities refer to capabilities that are prominent in large models but not obvious in small models, such as contextual learning and step-by-step reasoning. LLM can be used to build an intelligent question-answering system that gives accurate answers by understanding user questions and retrieving relevant information.

[0003] At present, the comprehensive management of substation production is still in the primitive manual review and dispatch stage. The patrol and maintenance tasks are manually arranged by experienced duty supervisors based on the periodic patrol and maintenance plan, problem maintenance list, power outage plan and weather conditions, which has the disadvantages of low efficiency and easy omissions. With the continuous increase in the number of sites under the jurisdiction of the patrol and maintenance center and the continuous decrease in the number of personnel, the overall workload of the patrol and maintenance center's daily production management has increased significantly, and the pressure on production safety has also increased simultaneously.

[0004] How to realize intelligent dispatching of substation patrol and maintenance through LLM technology is a technical problem that needs to be solved urgently. Summary of the invention

[0005] In view of the above problems, the present invention is proposed to provide a method, device, medium and equipment for intelligent dispatching of substation patrol and maintenance based on LLM, which overcomes the above problems or at least partially solves the above problems.

[0006] Other features and advantages of the present invention will become apparent from the following detailed description, or may be learned in part by practice of the present invention.

[0007] According to a first aspect of an embodiment of the present invention, a method for intelligent dispatching of substation maintenance based on LLM is provided, and the method for intelligent dispatching of substation maintenance based on LLM comprises:

[0008] Extracting and generating a structured historical work dispatch data set from the historical work dispatch data to be processed, wherein the historical work dispatch data is unstructured document data;

[0009] Generate an expanded work dispatch data set by expanding the historical work dispatch data set based on a collaborative filtering algorithm;

[0010] The basic LLM model is trained using the expanded work dispatching dataset and Lora technology to obtain the substation patrol and maintenance production decision LLM model for automatic intelligent work dispatching solution generation;

[0011] Based on the substation patrol and maintenance production decision LLM model, a plurality of new dispatching plans within a predetermined period are automatically generated according to the updated maintenance data;

[0012] The plurality of work dispatching plans are input into the decision-making supervision module for evaluation and processing to obtain a final optimal work dispatching plan.

[0013] In some embodiments of the present invention, extracting and generating a structured historical work dispatch data set from the historical work dispatch data to be processed includes:

[0014] Traversing the historical work dispatch data to be processed, and confirming the data attributes of each row by identifying the title bar of the historical work dispatch data;

[0015] Traverse the original data of each row corresponding to each data attribute to obtain the dispatch attribute data;

[0016] The structured historical work assignment data set is obtained by integrating the work assignment attribute data.

[0017] In some embodiments of the present invention, the generating an expanded work assignment data set by expanding the historical work assignment data set based on a collaborative filtering algorithm includes:

[0018] Reading employee information from the employee information data, and obtaining a structured employee information table based on the employee information;

[0019] The historical work dispatch data set is spliced ​​with the employee information table, and data is aggregated according to the time dimension to obtain a work dispatch data statistics table;

[0020] Based on the collaborative filtering algorithm, the work dispatch data statistics table is identified and processed to confirm the similarity of employees and evaluate the competent tasks of similar employees, and the competent task data of similar employees are generated;

[0021] Based on the competency data of the similar employees, the historical work assignment dataset is expanded with samples to obtain an expanded work assignment dataset.

[0022] In some embodiments of the present invention, the identification processing of the work dispatch data statistical table based on the collaborative filtering algorithm, confirming the similarity of employees and evaluating the competence tasks of similar employees, and generating the competence task data of similar employees includes:

[0023] Identify and process the work dispatch data statistics table to obtain target employee information and task execution information corresponding to the target employee;

[0024] Based on employee similarity, identify similar employees corresponding to the target employee, and search for a task set related to the task execution information of the similar employees and the target employee;

[0025] Confirm the target tasks in the task set that have not been performed by the similar employee, and generate competency task data of the similar employee according to the target tasks.

[0026] In some embodiments of the present invention, the use of the expanded work dispatching data set and the Lora technology to train the basic LLM model to obtain the substation patrol and maintenance production decision LLM model for automatic intelligent work dispatching solution generation includes:

[0027] Debugging the training environment of the basic LLM model, and loading the basic LLM model after the training environment debugging is completed;

[0028] Converting the maintenance information matched by the expanded work dispatch data set into structured maintenance data, and using the maintenance data and the expanded work dispatch data set as training sets for model training in the basic LLM model;

[0029] Establishing Lora layer parameter configuration and adding it to the basic LLM model for model training to obtain a model to be verified;

[0030] The trained model to be verified is evaluated based on the verification set to confirm the model accuracy. When the model accuracy meets the predetermined requirements, the model training is terminated to obtain the substation patrol and maintenance production decision LLM model.

[0031] In some embodiments of the present invention, the automatically generating a new dispatching plan within a predetermined period based on the substation patrol and maintenance production decision LLM model according to the updated maintenance data includes:

[0032] Build a reasoning environment and install the LLM model reasoning framework in the reasoning environment;

[0033] Loading the substation patrol and maintenance production decision LLM model into the LLM model reasoning framework;

[0034] The updated maintenance data is used as a motivation prompt for the LLM model reasoning framework, and a plurality of new work dispatching plans within a predetermined period are automatically generated through the LLM model reasoning framework.

[0035] In some embodiments of the present invention, the step of inputting the plurality of work dispatching plans into a decision-making supervision module for evaluation and processing to obtain a final optimal work dispatching plan includes:

[0036] Inputting the plurality of work dispatching plans into a decision-making supervision module for risk assessment;

[0037] The decision-making supervision module is used to score the parameter items of the dispatching plan, and the parameter items included in the risk assessment include the safety risk, labor cost, time cost of a single task assigned in the dispatching plan, and the number of dispatching tasks within a cycle;

[0038] The final optimal work dispatching plan is determined based on the scoring results.

[0039] According to a second aspect of an embodiment of the present invention, a substation patrol and maintenance intelligent dispatching device based on LLM is provided, and the substation patrol and maintenance intelligent dispatching device based on LLM includes:

[0040] A data conversion module, used to extract and generate a structured historical work dispatch data set from the historical work dispatch data to be processed, wherein the historical work dispatch data is unstructured document data;

[0041] A data expansion module, used to generate an expanded work assignment data set by expanding the historical work assignment data set based on a collaborative filtering algorithm;

[0042] Model training module, used to train the basic LLM model using the expanded work dispatching data set and Lora technology to obtain the substation patrol and maintenance production decision LLM model for automatic intelligent work dispatching solution generation;

[0043] A scheme generation module is used to automatically generate a plurality of new dispatching plan schemes within a predetermined period based on the substation patrol and maintenance production decision LLM model and the updated maintenance data;

[0044] The scoring module is used to input the multiple work dispatching plan schemes into the decision-making supervision module for evaluation and processing to obtain the final optimal work dispatching plan scheme.

[0045] According to a third aspect of an embodiment of the present invention, a computer-readable storage medium is provided, in which computer program instructions are stored. The computer program instructions are loaded and executed by a processor to implement the operations performed by any of the methods described above.

[0046] According to a fourth aspect of an embodiment of the present invention, there is provided an electronic device, comprising a processor and a memory, wherein the memory stores computer program instructions that can be executed by the processor, and when the processor executes the computer program instructions, it implements instructions of any of the methods described above.

[0047] The technical solution provided in the embodiments of the present invention has at least the following technical effects or advantages:

[0048] The embodiment of the present invention provides a method, device, medium and equipment for intelligent dispatching of substation patrol and maintenance based on LLM. The method generates an expanded dispatching data set by expanding the historical dispatching data set based on a collaborative filtering algorithm, and can expand the data set to 10 to 50 times the previous one, thereby preparing for the subsequent training of the LLM model based on Lora technology, which is used to improve the intelligent dispatching prediction accuracy of the Lora training LLM model, and also reduces the acquisition cost of the data set during model training; at the same time, based on the substation patrol and maintenance production decision LLM model, a plurality of new dispatching plans within a predetermined period are automatically generated, and then the decision supervision module is used to score the generated dispatching plans, so as to obtain the final optimal dispatching plan, which effectively overcomes the illusion problem and randomness of the existing LLM model in making intelligent dispatching plan decisions, and improves the intelligence level and the accuracy of the dispatching plan.

[0049] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 A schematic diagram of a process flow of a substation patrol and maintenance intelligent dispatching method based on LLM provided in an embodiment of the present invention;

[0052] Figure 2 A reference diagram for expanding the work dispatch dataset;

[0053] Figure 3 A reference diagram of the scoring results of the decision supervision module;

[0054] Figure 4 A schematic diagram of the principle structure of a substation patrol and maintenance intelligent dispatching device based on LLM provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0055] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings.

[0056] Various structural schematic diagrams according to embodiments of the present disclosure are shown in the accompanying drawings. These figures are not drawn to scale, and some details are magnified and some details may be omitted for the purpose of clear expression. The shapes of various regions and layers shown in the figures and the relative sizes and positional relationships therebetween are only exemplary, and may deviate in practice due to manufacturing tolerances or technical limitations, and those skilled in the art may further design regions / layers with different shapes, sizes, and relative positions according to actual needs.

[0057] In the context of the present disclosure, when a layer / element is referred to as being "on" another layer / element, the layer / element may be directly on the other layer / element, or there may be an intervening layer / element therebetween. Additionally, if a layer / element is "on" another layer / element in one orientation, the layer / element may be "below" the other layer / element when the orientation is reversed. In the context of the present disclosure, similar or identical components may be denoted by the same or similar reference numerals.

[0058] In order to better understand the above-mentioned technical scheme, the above-mentioned technical scheme will be described in detail below in combination with specific implementation methods. It should be understood that the embodiments of the present disclosure and the specific features in the embodiments are detailed descriptions of the technical scheme of the present invention, rather than limitations on the technical scheme of the present invention. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0059] Figure 1 is a flow chart of a method for intelligent dispatching substation maintenance based on LLM provided by an embodiment of the present invention, such as Figure 1 As shown, the LLM-based substation patrol and maintenance intelligent dispatching method includes the following steps:

[0060] S1. Extract and generate a structured historical work dispatch data set from the historical work dispatch data to be processed;

[0061] In an embodiment of the present invention, the historical work dispatch data is unstructured document data. For example, the historical work dispatch data may be an Excel document or a data table in another format.

[0062] An embodiment of the present invention extracts and generates a structured historical work dispatch data set from historical work dispatch data to be processed, including: traversing the historical work dispatch data to be processed, confirming the data attributes of each row by identifying the title bar of the historical work dispatch data; traversing the original data of the row corresponding to each data attribute to obtain the work dispatch attribute data; and obtaining the structured historical work dispatch data set after integrating the work dispatch attribute data.

[0063] The historical work dispatch data is a collection of information on the execution of tasks corresponding to each employee. The embodiment of the present invention obtains the historical work dispatch data to be processed, and confirms the data attributes of each row by identifying the title bar of the historical work dispatch data. The data attributes are used to mark the original data. The data attributes may be, for example, date, employee name, inspection substation name, work task name, work task details, and work task precautions, etc. After confirming the data attributes of each row, the original data of the row corresponding to each data attribute is traversed respectively to obtain the work dispatch attribute data, and the work dispatch attribute data is integrated to obtain the structured historical work dispatch data set. The historical work dispatch data set is shown in the following Table 1 for reference.

[0064] Table 1 Historical work assignment dataset

[0065]

[0066] S2. Generate an expanded work assignment data set by expanding the historical work assignment data set based on a collaborative filtering algorithm;

[0067] The embodiment of the present invention generates an expanded work dispatch data set by expanding the historical work dispatch data set based on a collaborative filtering algorithm, including: reading employee information in employee information data, and obtaining a structured employee information table based on the employee information; splicing the historical work dispatch data set with the employee information table, and aggregating data according to the time dimension to obtain a work dispatch data statistics table; identifying and processing the work dispatch data statistics table based on a collaborative filtering algorithm, confirming the similarity of employees and evaluating the competence of similar employees, and generating competence data of similar employees; based on the competence data of similar employees, performing sample expansion on the historical work dispatch data set to obtain an expanded work dispatch data set.

[0068] First, the embodiment of the present invention reads the employee information in the employee information data, and obtains a structured employee information table according to the employee information. The employee information table is shown in the following Table 2 for reference:

[0069] Table 2 Employee information table

[0070] Employee Name age job title Length of service Zhang San 24 primary 2 Li Si 30 intermediate 5

[0071] Then, the historical work dispatch data set is spliced ​​with the employee information table, and data is aggregated according to the time dimension to obtain a work dispatch data statistics table, which is shown in the following Table 3 for reference:

[0072] Table 3 Statistics of work dispatch data

[0073]

[0074] After obtaining the work dispatch data statistics table, an embodiment of the present invention identifies and processes the work dispatch data statistics table based on a collaborative filtering algorithm, confirms the similarity of employees and evaluates the competence tasks of similar employees, and generates competence task data of similar employees, including: identifying and processing the work dispatch data statistics table to obtain target employee information and task execution information corresponding to the target employee; based on employee similarity, confirming similar employees corresponding to the target employee, and searching for task sets related to the similar employees and the task execution information of the target employee; confirming the target tasks in the task set that have not been performed by the similar employees, and generating competence task data for similar employees based on the target tasks.

[0075] Specifically, the collaborative filtering algorithm can be implemented, for example, using the ALS package in the mllib library of the spark platform (not limited to the spark platform). The collaborative filtering algorithm discovers the behavioral rules of the substation patrol and maintenance employees in performing patrol and maintenance tasks through the historical behavior data of the substation patrol and maintenance employees (such as having performed patrol and maintenance tasks a at site A), and performs identification processing on the dispatch data statistics table. Based on the obtained target employee information and the task execution information corresponding to the target employee, the similarity between employees and the task similarity are confirmed. The relationship between employees is calculated based on the execution frequency of the same patrol and maintenance tasks by different employees, and patrol and maintenance tasks are recommended between employees with high similarity. For example, employees Zhang San and Li Si have both performed patrol and maintenance tasks a at site A, patrol and maintenance tasks b at site B, and patrol and maintenance tasks c at site C, and the execution frequency is 3 times a month. Then Zhang San and Li Si belong to the same type of employees with high similarity. The patrol and maintenance tasks d at site D that Zhang San has performed can also be assigned to Li Si to perform.

[0076] Regarding the calculation and confirmation of employee similarity, for example, it can be calculated using the Pearson correlation coefficient:

[0077]

[0078] In the formula, X represents the vector of employee 1 (such as Zhang San), and Y represents the vector of employee 2 (such as Li Si). represents the mean value of employee 1 vector, Represents the mean of employee 2 vectors.

[0079] By calculating the similarity between employees, patrol and maintenance tasks are recommended among employees with high similarity.

[0080]

[0081] Where W u,s is the similarity between employee u and employee s, R s,p is the frequency of employee s participating in task p, and finally R is calculated u,pIt indicates the frequency with which employee U may participate in task p.

[0082] For example, the similarity calculated based on Wang Wu and Zhao Liu in Table 3 is 0.85, indicating that the similarity between the two is high. Then calculate R u,p It is also 6, indicating that Zhao Liu is also competent for the D site_task d job. In this way, the competent task data corresponding to "Zhao Liu can also arrange the D site_task d job" can be constructed.

[0083] By generating competent task data of similar employees through the collaborative filtering algorithm, the data set can be expanded to 10 to 50 times the previous one, and the K level can be expanded to 10K to 100K data levels, so as to prepare for the subsequent training of the LLM model based on Lora technology and improve the accuracy of intelligent dispatch prediction of the LLM model trained by Lora. Figure 2 As shown, Figure 2 A reference diagram for expanding the work dispatch dataset.

[0084] S3. Use the expanded work dispatching data set and Lora technology to train the basic LLM model to obtain the substation patrol and maintenance production decision LLM model for automatic intelligent work dispatching solution generation;

[0085] The embodiment of the present invention uses an expanded work dispatching data set and Lora technology to train the basic LLM model, and obtains a substation patrol and maintenance production decision LLM model for automatic intelligent work dispatching solution generation, including: debugging the training environment of the basic LLM model, loading the basic LLM model after the training environment debugging is completed; converting the maintenance information matching the expanded work dispatching data set into structured maintenance data, and using the maintenance data and the expanded work dispatching data set as training sets for model training in the basic LLM model; establishing Lora layer parameter configuration, and adding it to the basic LLM model for model training to obtain a model to be verified; evaluating the trained model to be verified based on the verification set to confirm the model accuracy, and when the model accuracy meets the predetermined requirements, ending the model training to obtain the substation patrol and maintenance production decision LLM model. The verification set can be directly obtained from an existing database.

[0086] Before loading the basic LLM model, the embodiment of the present invention needs to build a training environment for the operation of the basic LLM model, by installing Pytorch, Hugging Face Transformers, perf (not limited to) and other libraries on the LLM model training server, and recording the basic LLM model. The basic LLM model is, for example, an open source LLM large model, such as Llama3 (not limited to this model).

[0087] The maintenance data includes but is not limited to periodic inspection maintenance plan, problem maintenance list, power outage plan and weather conditions, etc.

[0088] Since the basic LLM model that has not been fine-tuned has not learned the substation patrol and maintenance data, it is unable to make automatic patrol and maintenance dispatch decisions, or even if it can output a patrol and maintenance dispatch plan, the accuracy is very low. The embodiment of the present invention trains the basic LLM model by using an expanded dispatch data set and Lora technology to obtain a substation patrol and maintenance production decision LLM model for generating automatic intelligent dispatch plans.

[0089] S4. Based on the substation patrol and maintenance production decision LLM model, automatically generate multiple new dispatching plans within a predetermined period according to the updated maintenance data;

[0090] The embodiment of the present invention is based on the substation patrol and maintenance production decision LLM model, and automatically generates a new dispatch plan within a predetermined period according to the updated maintenance data, including: building an inference environment, installing the LLM model inference framework in the inference environment; loading the substation patrol and maintenance production decision LLM model into the LLM model inference framework; using the updated maintenance data as the incentive prompt of the LLM model inference framework, and automatically generating multiple new dispatch plans within the predetermined period through the LLM model inference framework.

[0091] The number N of generated work dispatching plans is ≥3; the predetermined period may be, for example, one week, one month, etc., which may be selected according to actual needs, and the embodiment of the present invention has no limitation on this.

[0092] S5. Input the plurality of work dispatching plans into the decision-making supervision module for evaluation and processing to obtain a final optimal work dispatching plan.

[0093] The embodiment of the present invention inputs multiple work dispatching plans into a decision-making supervision module for evaluation and processing, and obtains a final optimal work dispatching plan, including: inputting multiple work dispatching plans into a decision-making supervision module for risk assessment; scoring the parameter items of the work dispatching plan through the decision-making supervision module, wherein the parameter items included in the risk assessment include the safety risk, labor cost, time cost of a single task assigned in the work dispatching plan, and the number of dispatching tasks within a cycle; and confirming the final optimal work dispatching plan based on the scoring results.

[0094] Specifically, in the embodiment of the present invention, the decision supervision module obtains the scoring result through the following formula:

[0095]

[0096] In the formula, S is the total score of the work dispatch plan. The lower the score, the better the plan, and the higher the score, the worse the plan.

[0097] Ts indicates the level of safety risk of a single task assigned in the work dispatch plan. A safety risk of 0 means that the task is risk-free, and a safety risk greater than 0 means that there is a certain risk. For example, employee Wang Wu is a senior engineer, and the task D_site_task d is complex and requires a senior engineer to operate. The safety risk is close to 0, but if it is assigned to junior engineer Zhang San, there will be a certain safety risk.

[0098] Tc represents the labor cost of the work dispatch plan. The lower the labor cost of the overall work dispatch plan, the better the plan, and vice versa.

[0099] Tt represents the time cost of the work dispatch plan. The shorter the overall work dispatch plan takes, the better the plan is, otherwise the plan is worse.

[0100] α, β, and γ are weight coefficients corresponding to various costs, which are summarized by technical staff based on historical work dispatch data or selected based on actual application requirements.

[0101] N represents the number of dispatched tasks within the dispatch plan cycle.

[0102] For example, in the embodiment of the present invention, the N work dispatching plans output in step S4 are input into the decision-making supervision module for evaluation processing, and the decision-making supervision module scores the parameter items of the N work dispatching plans respectively, and obtains the scoring result by comprehensive scoring. The scoring result is as follows: Figure 3 As shown in the reference, the final optimal dispatching plan is confirmed according to the scoring result. By setting the decision supervision module, the embodiment of the present invention can overcome the illusion problem and randomness of the LLM model in making intelligent dispatching plan decisions, and improve the intelligence level and the accuracy of the dispatching plan.

[0103] The LLM-based intelligent dispatching method for substation maintenance described in the embodiment of the present invention has the following advantages compared with the prior art:

[0104] 1. Based on the collaborative filtering algorithm, the historical work dispatch data set is expanded to generate an expanded work dispatch data set, which can expand the data set to 10 to 50 times the previous one, so as to prepare for the subsequent training of the LLM model based on the Lora technology, and to improve the intelligent work dispatch prediction accuracy of the Lora training LLM model, while also reducing the data set acquisition cost during model training;

[0105] 2. Based on the substation patrol and maintenance production decision-making LLM model, multiple new dispatching plans within a predetermined period are automatically generated, and then the generated dispatching plans are scored using the decision supervision module to obtain the final optimal dispatching plan, which effectively overcomes the illusion problem and randomness of the existing LLM model for intelligent dispatching plan decision-making, and improves the intelligence level and accuracy of the dispatching plan.

[0106] Based on the above embodiment, as the above Figure 1 The present invention provides an embodiment of a substation patrol maintenance intelligent dispatching device based on LLM, and the device embodiment is similar to Figure 1 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices, see Figure 4 As shown, the LLM-based substation patrol and maintenance intelligent dispatching device includes:

[0107] The data conversion module 100 is used to extract and generate a structured historical work dispatch data set from the historical work dispatch data to be processed, wherein the historical work dispatch data is unstructured document data;

[0108] A data expansion module 200, configured to generate an expanded work assignment data set by expanding the historical work assignment data set based on a collaborative filtering algorithm;

[0109] The model training module 300 is used to train the basic LLM model using the expanded work dispatching data set and Lora technology to obtain the substation patrol and maintenance production decision LLM model for automatic intelligent work dispatching solution generation;

[0110] A plan generation module 400 is used to automatically generate a plurality of new dispatching plan plans within a predetermined period based on the substation patrol and maintenance production decision LLM model and the updated maintenance data;

[0111] The scoring module 500 is used to input the multiple work dispatching plans into the decision-making supervision module for evaluation and processing to obtain the final optimal work dispatching plan.

[0112] The LLM-based substation patrol and maintenance intelligent dispatching device described in the embodiment of the present invention can execute the LLM-based substation patrol and maintenance intelligent dispatching method provided in the above embodiment. The LLM-based substation patrol and maintenance intelligent dispatching device has the corresponding functional steps and beneficial effects of the LLM-based substation patrol and maintenance intelligent dispatching method described in the above embodiment. Please refer to the above-mentioned embodiment of the LLM-based substation patrol and maintenance intelligent dispatching method for details. The embodiment of the present invention will not be repeated here.

[0113] The embodiment of the present invention also provides an electronic device, which may include a processor and a memory, wherein the processor and the memory may be connected via a bus or other means. The processor may be a central processing unit (CPU). The processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components and other chips, or a combination of the above-mentioned various types of chips. The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs, non-transient computer executable programs and modules, such as the program instructions / modules corresponding to the intelligent dispatching method for substation patrol maintenance based on LLM in the embodiment of the present invention. The processor executes various functional applications and data processing of the processor by running the non-transient software programs, instructions and modules stored in the memory, that is, the intelligent dispatching method for substation patrol maintenance based on LLM in the above method embodiment is realized.

[0114] The memory may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required by at least one function; the data storage area may store data created by the processor, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. The one or more modules are stored in the memory, and when executed by the processor, the intelligent dispatching method for substation patrol and maintenance based on LLM in the above method embodiment is executed. The specific details of the above electronic equipment can be understood by corresponding to the corresponding descriptions and effects in the above method embodiment, and will not be repeated here. Those skilled in the art can understand that all or part of the processes in the above embodiment method can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and the program, when executed, may include the processes of the embodiments of the above methods. Among them, the storage medium can be a read-only memory (ROM), a random access memory (RAM), a flash memory (Flash Memory), a hard disk drive (HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above types of memory.

[0115] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0116] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting the intention that the claimed invention requires more features than those expressly recited in each claim. Rather, as reflected in the claims, inventive aspects lie in less than all of the features of the individual embodiments previously disclosed. Therefore, the claims that follow the detailed description are hereby expressly incorporated into the detailed description, with each claim itself serving as a separate embodiment of the present invention.

[0117] It should be noted that the above-mentioned embodiments illustrate rather than limit the invention and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims.

Claims

1. A method for intelligent dispatching of substation maintenance based on LLM, characterized in that: The LLM-based substation patrol and maintenance intelligent dispatching method includes: Extracting and generating a structured historical work dispatch data set from the historical work dispatch data to be processed, wherein the historical work dispatch data is unstructured document data; Generate an expanded work dispatch data set by expanding the historical work dispatch data set based on a collaborative filtering algorithm; The basic LLM model is trained using the expanded work dispatching dataset and Lora technology to obtain the substation patrol and maintenance production decision LLM model for automatic intelligent work dispatching solution generation; Based on the substation patrol and maintenance production decision LLM model, a plurality of new dispatching plans within a predetermined period are automatically generated according to the updated maintenance data; The plurality of work dispatching plans are input into the decision-making supervision module for evaluation and processing to obtain a final optimal work dispatching plan.

2. The LLM-based intelligent dispatching method for substation patrol and maintenance according to claim 1 is characterized in that: The step of extracting and generating a structured historical work dispatch data set from the historical work dispatch data to be processed includes: Traversing the historical work dispatch data to be processed, and confirming the data attributes of each row by identifying the title bar of the historical work dispatch data; Traverse the original data of each row corresponding to each data attribute to obtain the dispatch attribute data; The structured historical work assignment data set is obtained by integrating the work assignment attribute data.

3. The LLM-based intelligent dispatching method for substation patrol and maintenance according to claim 1 is characterized in that: The generating an expanded work assignment data set by expanding the historical work assignment data set based on the collaborative filtering algorithm includes: Reading employee information from the employee information data, and obtaining a structured employee information table based on the employee information; The historical work dispatch data set is spliced ​​with the employee information table, and data is aggregated according to the time dimension to obtain a work dispatch data statistics table; Based on the collaborative filtering algorithm, the work dispatch data statistics table is identified and processed to confirm the similarity of employees and evaluate the competent tasks of similar employees, and the competent task data of similar employees are generated; Based on the competency data of the similar employees, the historical work assignment dataset is expanded with samples to obtain an expanded work assignment dataset.

4. The LLM-based intelligent dispatching method for substation patrol and maintenance according to claim 3 is characterized in that: The identification and processing of the work dispatch data statistics table based on the collaborative filtering algorithm, confirming the similarity of employees and evaluating the competent tasks of similar employees, and generating competent task data of similar employees include: Identify and process the work dispatch data statistics table to obtain target employee information and task execution information corresponding to the target employee; Based on employee similarity, identify similar employees corresponding to the target employee, and search for a task set related to the task execution information of the similar employees and the target employee; Confirm the target tasks in the task set that have not been performed by the similar employee, and generate competency task data of the similar employee according to the target tasks.

5. The LLM-based intelligent dispatching method for substation patrol and maintenance according to claim 1 is characterized in that: The expanded work dispatching data set and Lora technology are used to train the basic LLM model to obtain the substation patrol and maintenance production decision LLM model for automatic intelligent work dispatching solution generation, including: Debugging the training environment of the basic LLM model, and loading the basic LLM model after the training environment debugging is completed; Converting the maintenance information matched by the expanded work dispatch data set into structured maintenance data, and using the maintenance data and the expanded work dispatch data set as training sets for model training in the basic LLM model; Establishing Lora layer parameter configuration and adding it to the basic LLM model for model training to obtain a model to be verified; The trained model to be verified is evaluated based on the verification set to confirm the model accuracy. When the model accuracy meets the predetermined requirements, the model training is terminated to obtain the substation patrol and maintenance production decision LLM model.

6. The LLM-based intelligent dispatching method for substation patrol and maintenance according to claim 1 is characterized in that: The automatic generation of a new dispatching plan within a predetermined period based on the substation patrol and maintenance production decision LLM model according to the updated maintenance data includes: Build a reasoning environment and install the LLM model reasoning framework in the reasoning environment; Loading the substation patrol and maintenance production decision LLM model into the LLM model reasoning framework; The updated maintenance data is used as a motivation prompt for the LLM model reasoning framework, and a plurality of new work dispatching plans within a predetermined period are automatically generated through the LLM model reasoning framework.

7. The LLM-based intelligent dispatching method for substation patrol and maintenance according to claim 1 is characterized in that: The step of inputting the plurality of work dispatching plans into the decision-making supervision module for evaluation and processing to obtain the final optimal work dispatching plan includes: Inputting the plurality of work dispatching plans into a decision-making supervision module for risk assessment; The decision-making supervision module is used to score the parameter items of the dispatching plan, and the parameter items included in the risk assessment include the safety risk, labor cost, time cost of a single task assigned in the dispatching plan, and the number of dispatching tasks within a cycle; The final optimal work dispatching plan is determined based on the scoring results.

8. A substation patrol and maintenance intelligent dispatching device based on LLM, applied to the method according to any one of claims 1 to 7, characterized in that: The LLM-based substation patrol and maintenance intelligent dispatching device includes: A data conversion module, used to extract and generate a structured historical work dispatch data set from the historical work dispatch data to be processed, wherein the historical work dispatch data is unstructured document data; A data expansion module, used to generate an expanded work assignment data set by expanding the historical work assignment data set based on a collaborative filtering algorithm; Model training module, used to train the basic LLM model using the expanded work dispatching data set and Lora technology to obtain the substation patrol and maintenance production decision LLM model for automatic intelligent work dispatching solution generation; A scheme generation module is used to automatically generate a plurality of new dispatching plan schemes within a predetermined period based on the substation patrol and maintenance production decision LLM model and the updated maintenance data; The scoring module is used to input the multiple work dispatching plan schemes into the decision-making supervision module for evaluation and processing to obtain the final optimal work dispatching plan scheme.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, which are loaded and executed by a processor to implement the operations performed by the method according to any one of claims 1 to 7.

10. An electronic device comprising a processor and a memory, characterized in that: The memory stores computer program instructions that can be executed by the processor, and when the processor executes the computer program instructions, the processor implements instructions of any one of the methods described in claims 1-7.

Images