Virtual power plant local knowledge base construction method based on large language model

Abstract

The invention provides a virtual power plant local knowledge base construction method based on a large language model, and relates to the technical field of power data application. According to the method and the device, format standardization processing is carried out on various complicated and massive data of the virtual power plant, so that key entity features can be conveniently identified and extracted in the later period, and the efficiency and the accuracy of constructing the knowledge base are improved. And then based on the data after standardization processing, entity extraction is carried out by utilizing a large language model to obtain an event object element set, so that the problem that all effective entities cannot be identified by a traditional entity extraction method due to the fact that whether dictionary information is comprehensive or not is solved, and meanwhile, a large amount of manpower and resource cost is saved. And then the event ontology knowledge model is constructed based on the event object element set for knowledge fusion storage to obtain a virtual power plant local knowledge base, so that the content accuracy of the knowledge base can be improved, the knowledge base construction efficiency can be improved, and the manpower and resource cost can be reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of power data application, and in particular to a method for constructing a local knowledge base of a virtual power plant based on a large language model. Background Art

[0002] As a new type of power resource combination, virtual power plants are widely used in new power systems. They refer to the integration of multi-source distributed energy equipment into a virtual power plant with centralized dispatching capabilities through network communication technology, and unified management and coordinated dispatch to improve the efficiency and reliability of power system operation while meeting customers' electricity needs.

[0003] Since the data types of virtual power plants may involve power equipment, basic business operation data, customer demand data, and related management systems, the data volume is huge and the types are diverse. In order to shorten the query response time and improve the efficiency and accuracy of professional knowledge query, it is particularly important to build a local knowledge base for virtual power plants.

[0004] However, in the process of building the local knowledge base of the virtual power plant, the method based on traditional entity relationship extraction is limited by the completeness of the dictionary. If the dictionary information is less or incomplete, the accuracy of named entity recognition will be reduced, which will cause erroneous content in the final knowledge base, which is not conducive to the output of query results and the use of daily data. Summary of the invention

[0005] The purpose of the present invention is to provide a method for constructing a local knowledge base of a virtual power plant based on a large language model to improve the above problems. In order to achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0006] This application provides a method for constructing a local knowledge base of a virtual power plant based on a large language model, including:

[0007] Acquire professional knowledge data in the electric power field, and format the professional knowledge data to obtain standardized data;

[0008] Extracting event key elements from each information element in the standardized data based on the large language model to obtain an event object element set, wherein the information element corresponds to one piece of professional knowledge data;

[0009] Constructing an event ontology knowledge model based on the event key elements in the event object element set;

[0010] The event object element set is input into the event ontology knowledge model for information fusion and storage to obtain a local knowledge base of the virtual power plant.

[0011] The beneficial effects of the present invention are:

[0012] In this application, the format of various types of complex and massive data of the virtual power plant is standardized to facilitate the identification and extraction of key entity features in the later stage, and to improve the efficiency and accuracy of building the knowledge base. Then, based on the standardized data, a large language model is used to extract entities to obtain a set of event object elements, which solves the problem that the traditional entity extraction method cannot identify all valid entities due to the constraint of whether the dictionary information is comprehensive, and also saves a lot of manpower and resource costs. Then, based on the set of event object elements, an event ontology knowledge model is constructed to carry out knowledge fusion storage, and a local knowledge base of the virtual power plant is obtained, which can not only improve the accuracy of the knowledge base content, but also improve the efficiency of knowledge base construction and reduce manpower and resource costs.

[0013] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or be understood by implementing the embodiments of the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0015] Figure 1 It is a flowchart of a method for constructing a local knowledge base of a virtual power plant based on a large language model in an embodiment of the present invention;

[0016] Figure 2 A system block diagram for constructing a local knowledge base of a virtual power plant based on a large language model in an embodiment of the present invention;

[0017] Figure 3 This is a structural block diagram of a method and device for constructing a local knowledge base of a virtual power plant based on a large language model in an embodiment of the present invention.

[0018] In the figure: 710-acquisition module; 711-first unit; 712-second unit; 713-third unit; 714-fourth unit; 720-editing module; 721-acquisition sub-unit; 722-classification unit; 7221-preprocessing unit; 7222-conversion unit; 7223-comparison unit; 7224-judgment unit; 7225-de-duplication unit; 7226-traversal unit; 723-calculation unit; 724-rewriting unit; 725-integration unit; 730-extraction module; 731-recognition unit; 732-generation unit; 733-event element indexing unit; 740-construction module; 750-fusion module; 800-virtual power plant local knowledge base construction method and device based on large language model; 801-processor; 802-memory; 803-multimedia component; 804-I / O interface; 805-communication component. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present invention.

[0020] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0021] Embodiment 1:

[0022] See also Figure 1 , Figure 1 This is a flowchart of a method for constructing a local knowledge base of a virtual power plant based on a large language model in an embodiment of the present invention. The figure shows that the method includes steps S1, S2, S3 and S4, wherein:

[0023] Step S1: Acquire professional knowledge data in the electric power field, and format the professional knowledge data to obtain standardized data.

[0024] It is understandable that the professional knowledge data in the field of electricity in this application includes but is not limited to engineering technical manuals and papers in the field of electricity, internal management regulations of power plants, power equipment operation data, power equipment maintenance manuals, books related to the field of electricity, etc. In order to build a knowledge base that covers a wide range and operates efficiently, a large amount of data is required as a foundation and support. However, the data generated by power plants in the daily management and operation process are of various types and have different file structures. Therefore, it is necessary to use intelligent recognition technology to standardize the text format of these data, and then realize standardized storage and management, which is also convenient for the later extraction of key feature information to build a knowledge base.

[0025] Furthermore, the above step S1 includes step S11, step S12, step S13 and step S14.

[0026] Step S11, obtaining each of the professional knowledge data and its corresponding data storage medium format.

[0027] It can be understood that the data storage medium format in this application refers to the format of the medium that carries the professional knowledge data, such as text-editable Word, non-text-editable PDF, etc.

[0028] Step S12: judging whether the data storage medium format is a text-editable electronic document based on the format of the data storage medium, and if so, saving it as the first data; if not, judging whether the data storage medium format is a picture format document;

[0029] Step S13: If it is a document in picture format, scan it and use automatic parsing technology and optical character recognition technology to quickly extract text and icon information, and save it as second data; if it is not a document in picture format, determine whether the data storage medium format is video data, and if so, use texture feature layer-by-layer incremental deep learning algorithm to identify text information and save it as third data;

[0030] Step S14: Obtain the standardized data based on the first data, the second data and the third data.

[0031] Through steps S11 to S14, all professional knowledge data are classified and converted into text according to the carrier format in which they are stored, so that all file formats become standardized data that can be edited in text.

[0032] After the standardization process is performed and before the key elements of the event are extracted from each information element in the standardized data, the process further includes steps S15, S16, S17, S18 and S19, wherein:

[0033] Step S15: Obtain the information channel label corresponding to each of the professional knowledge data, wherein the information channel label is a level label determined based on the difference in authority of the information acquisition channel.

[0034] It is understandable that in this step, the corresponding levels are customized based on the different professional recognition of the professional knowledge data information acquisition channels and the credibility and authority of the information by the technical personnel in the professional field. For example, the information channel label of the paper corresponds to the first-level label (very high credibility), the Baidu data of the power problem corresponds to the second-level label (average credibility), and the word-of-mouth data without relevant evidence corresponds to the third-level label (poor credibility).

[0035] Step S16: performing similar topic classification processing based on the topic information corresponding to all the standardized data and the information channel labels to obtain a similar topic data set.

[0036] It can be understood that in this step, the content of all standardized data is classified into the same topic based on the information channel label, which is convenient for subsequent rectification of the different content in the similar topic information, improving the accuracy and professionalism of the corpus, and obtaining high-quality corpus. The similar topic data set corresponds to multiple similar topic data, and each similar topic data corresponds to the standardized data of at least one topic.

[0037] The above step S16 includes step S161, step S162, step S163 and step S164, wherein:

[0038] Step S161: perform separator removal processing based on all the standardized data to obtain a preprocessed data set.

[0039] It is understandable that in this step, the standardized data is traversed to see whether there is a preset delimiter, and if so, it is deleted. The preset delimiters include a period, a comma, a special symbol, and the like.

[0040] Step S162: Based on each preprocessed data in the preprocessed data set, convert it into a corresponding hash value using the SimHash algorithm, and convert the hash value into a binary string to obtain a corresponding binary string set.

[0041] Step S163, performing pairwise comparisons on all binary strings in the binary string set to determine difference values, wherein the difference value is the total number of digits corresponding to different digits at the same position in two binary strings;

[0042] Step S164: based on the difference value and the information channel label, delete or modify the corresponding standardized data in the similar topic data, traverse all the retained binary strings, and obtain a similar topic data set.

[0043] The above step S164 includes step S1641, step S1642 and step S1643, wherein:

[0044] Step S1641, based on the difference value, determine whether it is less than the difference threshold. If the difference value is less than the difference threshold, determine whether the corresponding information channel label is greater than or equal to the preset level; if the difference value is greater than or equal to the difference threshold, classify the corresponding binary string into the similar topic data corresponding to different topic categories and retain them.

[0045] Step S1642: If the information channel label is greater than or equal to the preset level, the corresponding two binary strings will be divided into the similar topic data corresponding to the same topic category for retention; if the information channel label is less than the preset level, the corresponding two binary strings will be deleted one by one and then divided into the corresponding similar topic data for retention.

[0046] Step S1643: Based on the difference threshold and the information channel label, traverse all the retained binary strings to obtain a similar topic data set.

[0047] In step S161 to step S164, based on comprehensive considerations of the authority of the data acquisition channel and the document similarity, duplicate documents with higher similarity are selectively deleted, mainly deleting duplicate documents with average credibility, while retaining duplicate documents with higher credibility to facilitate the subsequent rectification and integration of these differential contents and expand the breadth and depth of the knowledge base's professional knowledge.

[0048] Step S17, based on each similar subject data in the similar topic data set, sentence segmentation is performed, and similarity between sentences is calculated to obtain a first set and a second set, wherein the first set is all sentences whose sentence similarity values ​​are greater than or equal to a preset similarity threshold, and the second set is all sentences whose sentence similarity values ​​are less than the preset similarity threshold.

[0049] It is understandable that in this step, similarity calculation operations are performed on similar subject data of the same subject content. First, the content in the same similar subject data is divided into sentences based on the separator comma, and then the similarity between the sentences is calculated to determine a first set of sentences with a certain similarity and a second set of complete sentences.

[0050] Step S18: rewrite the information based on the first set and the information channel label to obtain a rewritten text.

[0051] It can be understood that in this step, based on the information channel label to which the professional knowledge data corresponding to each sentence in the first set belongs, it is determined whether it is greater than or equal to the preset level. If the information channel label is greater than or equal to the preset level, the difference text is determined based on each sentence, and then the first prompt word is written based on the difference text, and the first prompt word is input into the large language model to generate the rewritten text, wherein the first prompt word is the input information for sorting and rewriting the difference text, and the difference text is the difference in the text obtained by matching the sentences one by one. If the information channel label is less than the preset level, manual verification is performed and the rewritten text is obtained based on the correct answer.

[0052] Step S19: obtaining the high-quality standardized data based on the second set and the rewritten text.

[0053] Step S2: extracting event key elements from each information element in the standardized data based on the large language model to obtain an event object element set, wherein the information element corresponds to one professional knowledge data.

[0054] It can be understood that in this step, the method of entity extraction using neural networks is subject to the constraints of whether the dictionary information is comprehensive, and it may not be possible to effectively identify entities in the document. In addition, building a dictionary requires a lot of manpower and effort. This application solves the above problems based on a large language model, and uses a large language model to extract key elements of events to obtain a set of event object elements.

[0055] The above step S2 includes step S21, step S22 and step S23, wherein:

[0056] Step S21: extract entity objects from all the standardized data corresponding to the similar topic data set based on the large language model to obtain an entity set.

[0057] It is understandable that in this step, the text relation extraction algorithm based on the large language model and semantic enhancement extracts entity objects for each similar topic data respectively. The text relation extraction algorithm model based on the large language model and semantic enhancement in this application includes 4 parts: encoder, relation extractor, entity extractor and word filtering module. Among them, the large language model is used as an encoder to convert text sentences into feature vectors, and the subject and object and the entities related to the two are marked by feature vectorization mainly through context understanding. Then the feature vector is input into the relation extractor to enhance the degree of association between entity pairs and enhance the information sharing between relations and entities using the self-attention mechanism. Then, the feature vector output by the large language model is used to extract the repeated entities in the relation extractor, and then the word filtering module is used to filter out the entities with high correlation between the subject and the object, and the entity set is traversed through the similar topic data set to obtain the entity set.

[0058] Step S22: construct a second prompt word based on each entity object in the entity set, and input the second prompt word into the large language model respectively to obtain an entity relationship data set; wherein the second prompt word is a prompt word corresponding to the entity relationship found from the similar topic data set.

[0059] It can be understood that in this step, the second prompt word is written based on the entity pair, and the relationship event extraction task between entities is converted into a multi-round question-answering question, aiming to decompose the complex information extraction task into multiple simple subtasks to improve efficiency. For example, a standardized data is AAABBCCC, DDGGGGGHHHH. Among them, AAA and CCC are entity pairs, and the second prompt word is written as "the original text is AAABBCCC, DDGGGGGHHHH, what is the relationship between AAA and CCC".

[0060] Step S23: Obtain an event object element set based on the corresponding entity set and entity relationship data set.

[0061] It can be understood that in this step, the corresponding entity pairs and their entity relationships are integrated to form an event object element set.

[0062] Step S3: construct an event ontology knowledge model based on the event key elements in the event object element set.

[0063] It can be understood that in this step, an entity-event table is constructed based on each event object element. The table is used to store the relationship information between events and entities, indicating which entities are involved in each event. Then, the event ontology knowledge model is constructed by connecting entities such as topic keywords as nodes and entity relationships as edges.

[0064] Step S4: input the event object element set into the event ontology knowledge model for information fusion storage to obtain a local knowledge base of the virtual power plant.

[0065] It can be understood that in this step, based on the entity-event table, it is input into the event ontology knowledge model for information fusion and storage to obtain the local knowledge base of the virtual power plant.

[0066] In this application, the format of various types of complex and massive data of the virtual power plant is standardized to facilitate the identification and extraction of key entity features in the later stage, and to improve the efficiency and accuracy of building the knowledge base. Then, based on the standardized data, a large language model is used to extract entities to obtain a set of event object elements, which solves the problem that the traditional entity extraction method cannot identify all valid entities due to the constraint of whether the dictionary information is comprehensive, and also saves a lot of manpower and resource costs. Then, based on the set of event object elements, an event ontology knowledge model is constructed to carry out knowledge fusion storage, and a local knowledge base of the virtual power plant is obtained, which can not only improve the accuracy of the knowledge base content, but also improve the efficiency of knowledge base construction and reduce manpower and resource costs.

[0067] Embodiment 2:

[0068] like Figure 2 As shown, this embodiment provides a block diagram of a virtual power plant local knowledge base construction system based on a large language model, including a virtual power plant local knowledge base construction system based on a large language model, including an acquisition module 710, an extraction module 730, a construction module 740 and a fusion module 750, wherein:

[0069] Acquisition module 710: used to acquire professional knowledge data in the electric power field, and format the professional knowledge data to obtain standardized data;

[0070] Optionally, the acquisition module 710 includes a first unit 711, a second unit 712, a third unit 713 and a fourth unit 714, wherein:

[0071] The first unit 711 is used to obtain each of the professional knowledge data and its corresponding data storage medium format;

[0072] The second unit 712 is used to determine whether the data storage medium format is a text-editable electronic document based on the data storage medium format, and if it is a text-editable electronic document, save it as the first data; if it is not a text-editable electronic document, determine whether the data storage medium format is a picture format document;

[0073] The third unit 713 is used for scanning and using automatic parsing technology and optical character recognition technology to quickly extract text and icon information if the document is in picture format, and saving it as second data; if it is not a picture format document, determining whether the data storage medium format is video data, and if so, using a texture feature layer-by-layer incremental deep learning algorithm to recognize text information, and saving it as third data;

[0074] The fourth unit 714 is configured to obtain the standardized data based on the first data, the second data and the third data.

[0075] Extraction module 730: used to extract event key elements from each information element in the standardized data based on the large language model to obtain an event object element set, wherein the information element corresponds to one piece of professional knowledge data;

[0076] Construction module 740: used to construct an event ontology knowledge model based on the event key elements in the event object element set;

[0077] Fusion module 750: used to input the event object element set into the event ontology knowledge model for information fusion storage to obtain a local knowledge base of the virtual power plant.

[0078] Furthermore, in other embodiments, the system for constructing a local knowledge base of a virtual power plant based on a large language model further includes an editing module 720, wherein the editing module 720 includes an acquisition subunit 721, a classification unit 722, a calculation unit 723, a rewriting unit 724 and an integration unit 725, wherein:

[0079] The acquisition subunit 721 is used to acquire the information channel label corresponding to each of the professional knowledge data, wherein the information channel label is a level label determined based on the difference in authority of the information acquisition channel;

[0080] Classification unit 722: used for performing similar topic classification processing based on the topic information corresponding to all the standardized data and the information channel labels to obtain a similar topic data set;

[0081] Optionally, in other embodiments, the classification unit 722 includes a preprocessing unit 7221, a conversion unit 7222, a comparison unit 7223, a judgment unit 7224, a deduplication unit 7225 and a traversal unit 7226, wherein:

[0082] Preprocessing unit 7221: used to remove separators based on all the standardized data to obtain a preprocessed data set;

[0083] The conversion unit 7222 is used to convert each preprocessed data in the preprocessed data set into a corresponding hash value using the SimHash algorithm, and convert the hash value into a binary string to obtain a corresponding binary string set;

[0084] Comparison unit 7223: used to compare all binary strings in the binary string set in pairs to determine difference values, where the difference value is the total bit value corresponding to different digits at the same position in two binary strings;

[0085] The judging unit 7224 is used to judge whether the difference value is less than a difference threshold based on the difference value, and if the difference value is less than the difference threshold, judge whether the corresponding information channel label is greater than or equal to a preset level; if the difference value is greater than or equal to the difference threshold, classify the corresponding binary string into the similar topic data corresponding to different topic categories and retain them;

[0086] Deduplication unit 7225: used for, if the information channel label is greater than or equal to the preset level, classifying the corresponding two binary strings into the similar subject data corresponding to the same subject category for retention; if the information channel label is less than the preset level, deleting one of the corresponding two binary strings and then classifying them into the corresponding similar subject data for retention;

[0087] The traversal unit 7226 is used to traverse all the retained binary strings based on the difference threshold and the information channel label to obtain a similar topic data set.

[0088] Calculation unit 723: used to perform sentence segmentation based on each similar subject data in the similar subject data set, and perform similarity calculation between sentences to obtain a first set and a second set, wherein the first set is all sentences whose sentence similarity values ​​are greater than or equal to a preset similarity threshold, and the second set is all sentences whose sentence similarity values ​​are less than the preset similarity threshold;

[0089] A rewriting unit 724 is configured to rewrite information based on the first set and the information channel label to obtain a rewritten text;

[0090] Integration unit 725: used for obtaining the high-quality standardized data based on the second set and the rewritten text.

[0091] Furthermore, in other embodiments, the extraction module 730 includes an identification unit 731, a generation unit 732 and an event element indexing unit 733, wherein:

[0092] Identification unit 731: used for extracting entity objects based on all the standardized data corresponding to the same similar subject data to obtain an entity set;

[0093] Generating unit 732: used to construct a second prompt word based on each entity object in the entity set, and input the second prompt word into the large language model respectively to obtain an entity relationship data set; wherein the second prompt word is a prompt word corresponding to the entity relationship found from the similar topic data set;

[0094] Event element indexing unit 733: used to obtain an event object element set based on the corresponding entity set and entity relationship data set.

[0095] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0096] Embodiment 3:

[0097] Corresponding to the above method embodiment, this embodiment also provides a method device 800 for constructing a local knowledge base of a virtual power plant based on a large language model. The method device 800 for constructing a local knowledge base of a virtual power plant based on a large language model described below and the method for constructing a local knowledge base of a virtual power plant based on a large language model described above can be referenced to each other.

[0098] Figure 3 8 is a block diagram of a method and device 800 for constructing a local knowledge base of a virtual power plant based on a large language model according to an exemplary embodiment. Figure 3 As shown, the method and device 800 for constructing a local knowledge base of a virtual power plant based on a large language model may include: a processor 801 and a memory 802. The method and device 800 for constructing a local knowledge base of a virtual power plant based on a large language model may also include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0099] The processor 801 is used to control the overall operation of the method and device 800 for constructing a local knowledge base of a virtual power plant based on a large language model, so as to complete all or part of the steps in the method and device 800 for constructing a local knowledge base of a virtual power plant based on a large language model. The memory 802 is used to store various types of data to support the operation of the method and device 800 for constructing a local knowledge base of a virtual power plant based on a large language model. For example, these data may include instructions for any application or method operating on the method and device 800 for constructing a local knowledge base of a virtual power plant based on a large language model, as well as application-related data, such as contact data, messages sent and received, pictures, audio, video, etc. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone, which is used to receive external audio signals. The received audio signal may be further stored in the memory 802 or sent via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 804 provides an interface between the processor 801 and other interface modules, and the above-mentioned other interface modules can be keyboards, mice, buttons, etc. These buttons can be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the virtual power plant local knowledge base construction method device 800 based on a large language model and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G or 4G, or a combination of them or several thereof, so the corresponding communication component 805 can include: Wi-Fi module, Bluetooth module, NFC module.

[0100] In an exemplary embodiment, the method and device 800 for constructing a local knowledge base of a virtual power plant based on a large language model can be implemented by one or more application-specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to execute the above-mentioned method for constructing a local knowledge base of a virtual power plant based on a large language model.

[0101] In another exemplary embodiment, a computer storage medium including program instructions is also provided, and when the program instructions are executed by a processor, the steps of the above-mentioned method for constructing a local knowledge base of a virtual power plant based on a large language model are implemented. For example, the computer storage medium can be the above-mentioned memory 802 including program instructions, and the above-mentioned program instructions can be executed by the processor 801 of the device 800 for constructing a local knowledge base of a virtual power plant based on a large language model to complete the above-mentioned method for constructing a local knowledge base of a virtual power plant based on a large language model.

[0102] Embodiment 4:

[0103] Corresponding to the above method embodiment, a storage medium is also provided in this embodiment. The storage medium described below and the method for constructing a local knowledge base of a virtual power plant based on a large language model described above can refer to each other.

[0104] A storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for constructing a local knowledge base of a virtual power plant based on a large language model in the above-mentioned method embodiment.

[0105] The storage medium may specifically be a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or other storage medium that can store program codes.

[0106] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0107] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field of the present invention can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for constructing a local knowledge base of a virtual power plant based on a large language model, characterized in that: include: Acquire professional knowledge data in the electric power field, and format the professional knowledge data to obtain standardized data; Extracting event key elements from each information element in the standardized data based on the large language model to obtain an event object element set, wherein the information element corresponds to one piece of professional knowledge data; Constructing an event ontology knowledge model based on the event key elements in the event object element set; The event object element set is input into the event ontology knowledge model for information fusion and storage to obtain a local knowledge base of the virtual power plant.

2. The method for constructing a local knowledge base of a virtual power plant based on a large language model according to claim 1, characterized in that: Obtaining professional knowledge data in the electric power field and formatting the professional knowledge data to obtain standardized data including: Obtain each of the professional knowledge data and its corresponding data storage medium format; Determining whether the data storage medium format is a text-editable electronic document based on the format of the data storage medium, and if it is a text-editable electronic document, saving it as the first data; if it is not a text-editable electronic document, determining whether the data storage medium format is a picture format document; If it is a document in picture format, scan it and use automatic parsing technology and optical character recognition technology to quickly extract text and icon information, and save it as the second data; if it is not a document in picture format, determine whether the data storage medium format is video data, and if so, use texture feature layer-by-layer incremental deep learning algorithm to identify text information and save it as the third data; The standardized data is obtained based on the first data, the second data and the third data.

3. The method for constructing a local knowledge base of a virtual power plant based on a large language model according to claim 1, characterized in that: Before extracting event key elements from each information element in the standardized data based on the large language model, the method further includes: Obtaining an information channel label corresponding to each of the professional knowledge data, wherein the information channel label is a level label determined based on the difference in authority of the information acquisition channel; Based on the subject information corresponding to all the standardized data and the information channel labels, similar subject classification processing is performed to obtain a similar subject data set; Sentences are divided based on each similar subject data in the similar subject data set, and similarity between the sentences is calculated to obtain a first set and a second set, wherein the first set is all sentences whose sentence similarity values ​​are greater than or equal to a preset similarity threshold, and the second set is all sentences whose sentence similarity values ​​are less than the preset similarity threshold; Rewrite the information based on the first set and the information channel label to obtain a rewritten text; Based on the second set and the rewritten text, high-quality standardized data is obtained.

4. The method for constructing a local knowledge base of a virtual power plant based on a large language model according to claim 3 is characterized in that: Based on the subject information corresponding to all the standardized data and the information channel labels, similar subject classification processing is performed to obtain a similar subject data set including: Based on all the standardized data, delimiters are removed respectively to obtain a preprocessed data set; Based on each preprocessed data in the preprocessed data set, convert it into a corresponding hash value using the SimHash algorithm, and convert the hash value into a binary string to obtain a corresponding binary string set; Based on all binary strings in the binary string set, pairwise comparison is performed to determine a difference value, wherein the difference value is a total bit value corresponding to different digits at the same position in the two binary strings; Based on the difference value and the information channel label, the corresponding standardized data in the similar topic data is deleted and modified, and all the retained binary strings are traversed to obtain a similar topic data set.

5. The method for constructing a local knowledge base of a virtual power plant based on a large language model according to claim 4 is characterized in that: Based on the difference value and the information channel label, the standardized data corresponding to the similar topic data is deleted and modified, and all the retained binary strings are traversed to obtain a similar topic data set, including: Based on the difference value, it is determined whether it is less than a difference threshold. If the difference value is less than the difference threshold, it is determined whether the corresponding information channel label is greater than or equal to a preset level; if the difference value is greater than or equal to the difference threshold, the corresponding binary string is divided into the similar topic data corresponding to different topic categories and retained; If the information channel label is greater than or equal to the preset level, the corresponding two binary strings are divided into the similar topic data corresponding to the same topic category and retained; if the information channel label is less than the preset level, the corresponding two binary strings are selected and deleted and then divided into the corresponding similar topic data and retained; Based on the difference threshold and the information channel label, all the retained binary strings are traversed to obtain a similar topic data set.

6. The method for constructing a local knowledge base of a virtual power plant based on a large language model according to claim 4 is characterized in that: Based on the large language model, each information element in the standardized data is respectively extracted with event key elements, and the event object element set obtained includes: Extracting entity objects based on all the standardized data corresponding to the same similar subject data to obtain an entity set; Construct a second prompt word based on each entity object in the entity set, and input the second prompt words into the large language model respectively to obtain an entity relationship data set; wherein the second prompt word is a prompt word corresponding to the entity relationship found from the similar topic data set; Based on the corresponding entity set and entity relationship data set, an event object element set is obtained.

7. The method for constructing a local knowledge base of a virtual power plant based on a large language model according to claim 3, characterized in that: Rewriting information based on the first set and the information channel label to obtain a rewritten text includes: Determining whether the information channel label to which the professional knowledge data corresponding to each sentence in the first set belongs is greater than or equal to a preset level; If the information channel tag is greater than or equal to the preset level, determining difference texts based on each sentence, then writing a first prompt word based on the difference texts, and inputting the first prompt word into the large language model to generate a rewritten text, wherein the first prompt word is input information for arranging and rewriting the difference texts, and the difference texts are text differences obtained by matching sentences one by one; If the information channel tag is less than the preset level, a manual check is performed and a rewritten text is obtained based on the correct answer.

Images