Reservoir modeling knowledge graph construction method and system

Abstract

The invention provides a reservoir modeling knowledge graph construction method and system, and belongs to the technical field of oil field exploration. The method comprises the steps of obtaining metadata and text information corresponding to each reservoir modeling literature based on each pre-collected reservoir modeling literature; constructing a dictionary containing modeling knowledge spatio-temporal information based on metadata and text information corresponding to each reservoir modeling document; key information of each reservoir modeling literature in the dictionary is extracted, and the extracted key information is stored in a structured database; and based on the structured database, constructing a knowledge graph to obtain a graph database. According to the method, the reservoir modeling literatures are automatically collected, meanwhile, the key information is extracted from all the reservoir modeling literatures through the information extraction technology, higher accuracy and comprehensiveness are achieved, the reservoir modeling literatures can be automatically classified and extracted, and the working efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of oilfield exploration technology, specifically to a method for constructing a reservoir modeling knowledge graph, a system for constructing a reservoir modeling knowledge graph, a machine-readable storage medium, and an electronic device. Background Technology

[0002] In today's information age, literature, as a vital carrier of knowledge, plays an increasingly important role. In the field of oil and gas reservoir modeling, literature is an indispensable resource. A wealth of theories, methods, technologies, and cutting-edge knowledge on oil and gas reservoir modeling can be found in various literature sources, including journal articles, monographs, proceedings, dissertations, technical reports, national and industry standards and specifications, as well as electronic documents available online or stored on various storage devices. These documents not only contain rich modeling knowledge but also provide data support for practical case studies.

[0003] In the current field of oil and gas reservoir modeling, although a wealth of literature is available for reference, several intractable problems remain. First, manually collecting and organizing literature is extremely time-consuming and labor-intensive, and it's difficult to guarantee the accuracy and comprehensiveness of the information. Second, existing literature management tools or search engines have limitations, such as inaccurate search results and the inability to automatically classify and extract literature. These problems make the utilization of literature and information extraction in modeling practice very difficult.

[0004] Therefore, it is urgent to solve the problem that existing reservoir modeling knowledge extraction algorithms are inefficient and inaccurate when faced with a large amount of literature data. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for constructing a reservoir modeling knowledge graph, so as to at least solve the problems of low efficiency and low accuracy of existing reservoir modeling knowledge extraction algorithms when faced with a large amount of literature data.

[0006] To achieve the above objectives, a first aspect of the present invention provides a method for constructing a reservoir modeling knowledge graph, comprising:

[0007] Based on the pre-collected reservoir modeling literature, the metadata and text information corresponding to each reservoir modeling literature are obtained;

[0008] Based on the metadata and text information corresponding to each reservoir modeling literature, a dictionary containing spatiotemporal information of modeling knowledge is constructed.

[0009] Key information from each reservoir modeling literature in the dictionary is extracted and stored in a structured database.

[0010] Based on structured databases, knowledge graphs are constructed to obtain graph databases.

[0011] Optionally, the text information corresponding to each reservoir modeling document includes first text information, second text information and / or text information extracted from PDF format files;

[0012] Based on the pre-collected literature on reservoir modeling, the text information corresponding to each literature on reservoir modeling is obtained, including:

[0013] Data cleaning and data extraction were performed on pre-collected reservoir modeling literature based on PDF format files, resulting in text information extracted from PDF format files, CAJ format files, and PDF documents that could not be directly extracted.

[0014] Based on OCR, text in PDF documents that cannot be directly extracted is extracted to obtain the first text information;

[0015] Based on a pre-built CAJ information extraction program, the CAJ format file is converted into a text file to obtain the second text information.

[0016] Optionally, the above-mentioned OCR-based method extracts text from PDF documents that cannot be directly extracted, obtaining the first text information, including:

[0017] For PDF documents that cannot be directly extracted, pagination is performed, and each page of the PDF is read as a corresponding image.

[0018] Based on the BGR2GRAY function, each image is converted into a corresponding grayscale image, and then each grayscale image is binarized using OpenCV.

[0019] Use the image_to_string package in the Pytesseract library to read the binarized image as a text image;

[0020] By combining the text and images corresponding to each page of the PDF, the first text information is obtained.

[0021] Optionally, the above-mentioned conversion of CAJ format files into text files based on a pre-built CAJ information extraction program includes:

[0022] The CAJ format files are used as conversion tasks, and the conversion tasks are broken down into multiple user interaction actions; each user interaction action corresponds to a CAJ format file in the conversion task, and each CAJ format file in the conversion task has an initial task icon.

[0023] Record the clicked icon and click method corresponding to each user interaction action;

[0024] If the current interactive interface shows whether the corresponding initial task icon exists in the conversion task, and if it does, then the corresponding user interaction action is executed according to the preset order and the corresponding click method to convert the CAJ format file corresponding to the initial task icon into a text file.

[0025] Optionally, the above key information includes keywords;

[0026] The above-mentioned key information extraction from the reservoir modeling literature in the dictionary includes:

[0027] Based on the jieba word segmentation library, keywords of each reservoir modeling document in the dictionary are extracted, and regular expressions are used to mark the keywords of each reservoir modeling document in the dictionary.

[0028] Optionally, the keywords extracted from various reservoir modeling documents in the dictionary based on the jieba word segmentation database include:

[0029] Based on a dictionary containing spatiotemporal information of modeling knowledge, a prefix dictionary of words in the dictionary is constructed;

[0030] The texts to be segmented for each reservoir modeling literature are sequentially cleaned and segmented into Chinese characters to obtain the individual characters corresponding to each text to be segmented.

[0031] Based on the prefix dictionary of words in the dictionary and the single characters corresponding to each text to be segmented, a directed acyclic graph corresponding to each text to be segmented is constructed, and the path with the highest probability corresponding to each directed acyclic graph is determined.

[0032] Based on the directed acyclic graph corresponding to each text to be segmented and the corresponding path with the highest probability, the set of word groups corresponding to each text to be segmented is obtained;

[0033] Read the Chinese dictionary and determine whether the set of word groups corresponding to each text to be segmented exists in the Chinese dictionary;

[0034] If the set of phrases exists in a Chinese dictionary, then the keywords of each reservoir modeling document are annotated according to the Chinese dictionary.

[0035] If the phrase set does not exist in the Chinese dictionary, then the Chinese and English in the phrase set are separated, and the probability of the phrase sequence corresponding to the separated phrase set is calculated based on the HMM model and Viterbi algorithm to obtain the optimal phrase sequence.

[0036] Optionally, the keywords extracted from various reservoir modeling documents in the dictionary based on the jieba word segmentation database include:

[0037] Obtain the text to be segmented corresponding to each reservoir modeling document in the dictionary;

[0038] The text to be segmented corresponding to each reservoir modeling document is used as the input parameter of the cut function of the jieba word segmentation library to obtain all the word segmentation results corresponding to each text to be segmented.

[0039] Based on a preset stop word list, all segmentation results corresponding to each text to be segmented are traversed, and stop word removal is performed on each segmentation result. Words not in the stop word list are saved as keywords.

[0040] Optionally, the spatiotemporal information of the above modeling knowledge includes at least geological structural regional information, administrative division information, and stratigraphic information of the corresponding age.

[0041] Optionally, the above-mentioned knowledge graph is constructed based on a structured database to obtain a graph database, including:

[0042] Construct a knowledge graph pattern layer containing spatiotemporal information of modeling knowledge; wherein, the knowledge graph pattern layer includes an ontology and multiple entities;

[0043] Establish the associations between each node object and each entity in the NEO4J database to construct a knowledge graph;

[0044] After disambiguation of the structured database, the key information from the disambiguated structured database is filled into the corresponding node objects of the NEO4J database after the knowledge graph is constructed, thus obtaining the graph database.

[0045] Optional rules for collecting literature on reservoir modeling include:

[0046] Using Python web crawlers and Javascript scripts, multiple reservoir modeling documents were obtained from target information websites, and the metadata of each reservoir modeling document was organized and supplemented.

[0047] Optionally, before retrieving multiple reservoir modeling documents from the target information website, the method further includes:

[0048] Based on data integration tools, we extracted reservoir modeling literature from various target information websites;

[0049] Compare the reservoir modeling literature from each target information website to identify at least one reservoir modeling literature group; wherein, the reservoir modeling literature group contains multiple reservoir modeling literatures belonging to the same literature.

[0050] For each reservoir modeling literature group, the differences between the reservoir modeling literatures in the reservoir modeling literature group are compared. Based on the comparison results, the reservoir modeling literatures in the reservoir modeling literature group are corrected and filled in sequentially to obtain the processed reservoir modeling literatures.

[0051] The processed reservoir modeling literature was used to replace the corresponding reservoir modeling literature in each target information website.

[0052] Optionally, after establishing the associations between each node object and each entity in the NEO4J database, the method further includes:

[0053] Based on the node objects and their corresponding associated entities, a set of graph data is constructed, and identity identification information is generated for each set of graph data.

[0054] The identity information includes:

[0055] Data source information, data association information, and keyword tag information.

[0056] Optionally, the above-mentioned graph data is constructed based on node objects and their corresponding associated entities, and identity information is generated for each set of graph data, including:

[0057] Read the data extraction method of the current map data, and backtrack the source data of the current map data based on the determined extraction method;

[0058] Determine the source data extraction library and generate corresponding data source information label fields based on the extraction library;

[0059] Data source information is generated based on the data source information tag field, and the source data is compressed and stored in the pre-built source database.

[0060] Based on the knowledge and spatiotemporal information of the current map data, relevant data indexing is performed, and the degree of correlation between each relevant data and the current map data is determined based on the cosine similarity rule;

[0061] Data association information for the current map data is generated based on the relevant data and the degree of association between the relevant data and the current map data.

[0062] Construct a corresponding graph data terminology based on the keywords of the current graph data;

[0063] Calculate the Jaccard similarity between each graph data word set and the pre-constructed graph data word set corresponding to the preset index target. Use all corresponding index targets with Jaccard similarity greater than the preset similarity threshold as the keyword tag information of the current graph data.

[0064] A second aspect of the present invention provides a reservoir modeling knowledge graph construction system, comprising:

[0065] The text information acquisition module is used to obtain the metadata and text information corresponding to each reservoir modeling document based on the pre-collected reservoir modeling documents;

[0066] The dictionary construction module is used to construct a dictionary containing spatiotemporal information of modeling knowledge based on the metadata and text information corresponding to each reservoir modeling literature;

[0067] The key information extraction module is used to extract key information from the reservoir modeling literature in the dictionary and store the extracted key information in a structured database.

[0068] The knowledge graph construction module is used to construct knowledge graphs based on structured databases, resulting in a graph database.

[0069] In a third aspect, the present invention provides a machine-readable storage medium storing instructions that, when executed by a processor, configure the processor to perform the reservoir modeling knowledge graph construction method described above.

[0070] In a fourth aspect, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the reservoir modeling knowledge graph construction method described above.

[0071] The above technical solution provides a method and system for constructing a knowledge graph for reservoir modeling. After collecting reservoir modeling literature through various methods, the system organizes and supplements the metadata corresponding to each literature, and simultaneously digitizes the information of each literature into textual information. Based on the metadata and textual information of each literature, a dictionary containing spatiotemporal information of modeling knowledge is constructed. This targeted dictionary construction method ensures that the extracted key information is more accurate and refined, effectively improving the knowledge extraction effect. Key information of each literature in the dictionary is extracted and stored in a structured database. Based on the structured database, a knowledge graph is constructed, resulting in a graph database. This method and system automatically collect reservoir modeling literature while using information extraction technology to extract key information from each literature, achieving higher accuracy and comprehensiveness. Furthermore, it can automatically classify and extract information from reservoir modeling literature, greatly improving work efficiency. Furthermore, based on automating information extraction, the extracted key information was further structured and organized, ultimately constructing a knowledge graph (i.e., a graph database) for the field of reservoir modeling. This approach not only integrates and abstracts information from various literature sources but also clearly presents the relationships between these information. This knowledge graph is stored in the form of a graph database, enabling convenient and rapid querying and analysis. Through continuous updating and improvement of this knowledge graph, a more comprehensive understanding of various theories, methods, technologies, and cutting-edge knowledge in the field of reservoir modeling can be achieved, providing more accurate support and guidance for practical modeling work.

[0072] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0073] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings:

[0074] Figure 1 This is a flowchart of a reservoir modeling knowledge graph construction method provided by one embodiment of the present invention;

[0075] Figure 2 This is a block diagram of a reservoir modeling knowledge graph construction system provided in one embodiment of the present invention;

[0076] Figure 3 This is a flowchart of another method for constructing a reservoir modeling knowledge graph, provided by one embodiment of the present invention;

[0077] Figure 4 This is a type percentage chart of reservoir modeling literature provided by one embodiment of the present invention;

[0078] Figure 5 This is a flowchart of an embodiment of the present invention for data cleaning and text extraction of reservoir modeling literature;

[0079] Figure 6 This is a schematic diagram illustrating the extraction of text from an image using OCR, according to one embodiment of the present invention.

[0080] Figure 7 This is a flowchart illustrating the implementation of batch extraction of CAJ format data according to one embodiment of the present invention.

[0081] Figure 8 This is a flowchart of extracting key information from various reservoir modeling documents in a dictionary, provided by one embodiment of the present invention.

[0082] Figure 9 This is a schematic diagram of the design of a knowledge graph pattern layer containing spatiotemporal information of modeling knowledge, provided by one embodiment of the present invention.

[0083] Figure 10 This is a flowchart illustrating the implementation of knowledge graph construction using a structured database, according to one embodiment of the present invention. Detailed Implementation

[0084] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0085] Figure 1 This is a flowchart of a reservoir modeling knowledge graph construction method provided by one embodiment of the present invention. Figure 3 This is a flowchart of another method for constructing a reservoir modeling knowledge graph, provided by one embodiment of the present invention. For example... Figure 1 and Figure 3 As shown, this invention provides a method for constructing a reservoir modeling knowledge graph, including:

[0086] S110: Based on the pre-collected reservoir modeling literature, obtain the metadata and text information corresponding to each reservoir modeling literature;

[0087] Specifically, we collected reservoir modeling literature through various methods, and organized and supplemented the metadata of the collected reservoir modeling literature.

[0088] For example, reservoir modeling-related literature (i.e., reservoir modeling literature) is collected through various methods, and the metadata of the collected reservoir modeling literature is organized and supplemented. In this embodiment, all domestic literature related to reservoir modeling from 1996 to 2022 was collected, totaling 2668 metadata entries, including 1185 journal articles, 1163 master's theses, 183 doctoral dissertations, 133 conference papers, 2 newspaper articles, 2 special issues, and 2105 appendices. The collected literature data accounts for a significant portion of the total. Figure 4 As shown, Figure 4 This is a type percentage diagram of reservoir modeling literature collected according to one embodiment of the present invention; the literature attachments are used to store the full-text attachments corresponding to the literature and serve as the carrier for storing the actual content of the literature. Attachment formats include PDF, CAJ, etc. The content covers research on various genetic bodies of oil and gas reservoirs, outcrop interpretation research, and modeling method research.

[0089] Please refer to Figure 5 , Figure 5 This is a flowchart illustrating the implementation of data cleaning and text extraction for reservoir modeling literature according to one embodiment of the present invention. In some embodiments of this invention, the text information corresponding to each reservoir modeling literature includes first text information, second text information, and / or text information extracted from PDF format files. Based on pre-collected reservoir modeling literature, obtaining the text information corresponding to each reservoir modeling literature includes: performing data cleaning and data extraction on pre-collected reservoir modeling literature based on PDF format files to obtain text information extracted from PDF format files, CAJ format files, and PDF documents that cannot be directly extracted; extracting text from the PDF documents that cannot be directly extracted using OCR to obtain the first text information; and converting the CAJ format files into text files using a pre-set CAJ information extraction program to obtain the second text information.

[0090] PDF files can be extracted using the Python third-party library PDFMiner. Based on the PDFMiner implementation, text can be extracted from PDF documents that can be directly extracted. For implementations that can directly extract PDF documents, this method implements a method of parsing PDF files using PDFMiner. A "Lazy Parsing" strategy is employed to efficiently extract text from complex and large-volume PDF files. Data extraction and parsing are based on two core classes of PDFMiner: PDFParser and PDFDocument. PDFParser is specifically responsible for extracting data from PDF files, while PDFDocument is used to store this data. Simultaneously, PDFPageInterpreter processes page content, and PDFDevice converts the processed information into the required format. The text data is saved using PDFResourceManager, resulting in the extracted text information from the PDF file.

[0091] Specifically, the pre-collected reservoir modeling literature was first cleaned and extracted using the Python third-party library PDFMiner. This yielded text information extracted by PDFMiner, CAJ format files, and PDF documents that could not be directly extracted. For PDF documents that could not be parsed by PDFMiner, OCR technology was used to extract text data, resulting in the first type of text information. For CAJ format files that could not be directly opened, a method involving continuous detection and simulated key clicks (i.e., a pre-built CAJ information extraction program) was written to convert them into text data in batches, yielding the second type of text information. This achieved the goal of textualizing information from various reservoir modeling literature using PDFMiner, OCR, and the pre-built CAJ information extraction program.

[0092] In some implementations of this embodiment, the above-mentioned method of extracting text from PDF documents that cannot be directly extracted using OCR to obtain first text information includes: paginating the PDF documents that cannot be directly extracted, and reading each page of the PDF as a corresponding image; converting each image into a corresponding grayscale image using the BGR2GRAY function, and performing binarization processing on each grayscale image using OpenCV; reading the binarized image as a text image using the image_to_string package of the Pytesseract library; and combining the text images corresponding to each page of the PDF to obtain the first text information.

[0093] Specifically, for PDF documents where text cannot be directly extracted using Python scripts, an OCR (Optical Character Recognition) method is used to extract text from PDFs whose content is presented as images. First, the target PDF document (i.e., the PDF document that cannot be directly extracted) is paginated, and each page is read as an image. For each image, the BGR2GRAY function is used to convert it to a grayscale image, and then OpenCV is used to binarize the image, taking the grayscale range (0-127) as 0 and the grayscale range (128-255) as 255. The binarized image is then used to formally read as a text image using the `image_to_string` package of the Pytesseract library. By collecting data from all pages, the text in the target PDF document can be extracted, thus solving the problem that some PDF text cannot be directly extracted using Python scripts. For example... Figure 6 As shown, Figure 6 This is a schematic diagram illustrating the use of OCR to extract text from an image, according to one embodiment of the present invention. Figure 6 The demonstration features a PDF image of the collected data, showing the process of using OCR technology to recognize and extract text from top to bottom.

[0094] In some embodiments of this example, the above-mentioned conversion of CAJ format files into text files based on a preset CAJ information extraction program includes: taking the CAJ format files as conversion tasks, and decomposing the conversion tasks into multiple user interaction actions; wherein each user interaction action corresponds to each CAJ format file in the conversion task, and each CAJ format file in the conversion task corresponds to an initial task icon; recording the click icon and click method corresponding to each user interaction action; searching whether the initial task icon corresponding to the conversion task exists in the current interactive interface; if it exists, executing the corresponding user interaction action according to a preset order and the corresponding click method to convert the CAJ format file corresponding to the initial task icon into a text file.

[0095] The user interface can be a Windows interface.

[0096] Specifically, for CAJ format files, since this format lacks open-source batch processing tools, and existing readers only support exporting single documents as text, a Windows-based automated execution program (i.e., a pre-built CAJ information extraction program) was developed based on pyautogui and OpenCV to solve the problem of converting large amounts of CAJ text. Its basic idea is as follows: Figure 7 As shown, Figure 7This is a flowchart illustrating the implementation of a batch extraction process for CAJ format data according to one embodiment of the present invention: ① The conversion task for CAJ format files is broken down into several user interaction actions; ② The icons clicked and the clicking methods for each user interaction action are recorded; ③ After the Windows operation automation execution program is executed, the program automatically searches for the initial task icon in the task sequence of the conversion task in the current Windows interface. If it exists, the subsequent operations are executed. During execution, the Windows operation automation execution program can also determine whether to execute the next operation based on specific conditions, thereby avoiding the possibility of the program executing incorrect operations. Through the data extraction of this Windows operation automation execution program, all CAJ format files in the literature are converted into TXT files to facilitate subsequent unified data extraction.

[0097] S120: Construct a dictionary containing spatiotemporal information of modeling knowledge based on the metadata and text information corresponding to each reservoir modeling literature;

[0098] Specifically, a targeted dictionary for the spatiotemporal information of modeling knowledge is constructed to meet the specific characteristics of the information to be extracted (i.e., the key information in step S130), thereby satisfying the needs of subsequent word segmentation and part-of-speech tagging. The spatiotemporal information of modeling knowledge includes at least spatial information such as geological structural regional information and administrative division information, as well as temporal information such as the chronological stratigraphy. This targeted dictionary construction method ensures that the extracted key information is more accurate and refined, effectively improving the knowledge extraction effect. Due to the specificity of the information to be extracted, ordinary general-purpose dictionaries cannot meet the needs of word segmentation and part-of-speech tagging; therefore, a targeted dictionary for the spatiotemporal information of modeling knowledge needs to be constructed specifically for processing the spatiotemporal information of modeling knowledge. This dictionary includes spatial and temporal information. Spatial information includes basins and administrative divisions, while temporal information includes the chronological stratigraphy. A total of 1679 related terms are collected, with each spatiotemporal information term coming from relevant professional entries in Chapter Two, such as basins, administrative divisions, and stratigraphic units.

[0099] S130: Extract key information from each reservoir modeling literature in the dictionary and store the extracted key information in a structured database;

[0100] In some implementations of this embodiment, the aforementioned key information includes keywords; the extraction of key information from each reservoir modeling document in the dictionary includes: extracting keywords from each reservoir modeling document in the dictionary based on the jieba word segmentation library, and using regular expressions to mark the keywords of each reservoir modeling document in the dictionary.

[0101] Specifically, after the dictionary is constructed, text segmentation and information extraction can be performed. Jieba segmentation, as a widely used segmentation tool, demonstrates high performance in both efficiency and effectiveness. Keywords from various reservoir modeling documents are extracted using the Python jieba segmentation library, and then regular expressions are used to identify keywords present in the dictionary. Using jieba segmentation to extract spatiotemporal information from the constructed document text data enables rapid extraction and persistent storage in a structured database.

[0102] Please refer to Figure 8 , Figure 8 This is a flowchart illustrating the extraction of key information from reservoir modeling literature in a dictionary, provided by one embodiment of the present invention. In some embodiments of this invention, the extraction of keywords from reservoir modeling literature in the dictionary based on the jieba word segmentation database includes:

[0103] Based on a dictionary containing spatiotemporal information of modeling knowledge, a prefix dictionary of words in the dictionary is constructed;

[0104] The texts to be segmented for each reservoir modeling literature are sequentially cleaned and segmented into Chinese characters to obtain the individual characters corresponding to each text to be segmented.

[0105] Specifically, the text to be segmented for each reservoir modeling document is first cleaned, corrected, filtered, and transformed to improve data quality and usability. Then, the cleaned text is segmented into Chinese characters according to clauses (e.g., regular expressions can be used, with punctuation marks or non-Chinese characters as delimiters), thus obtaining the individual characters corresponding to each segmented text.

[0106] Based on the prefix dictionary of words in the dictionary and the single characters corresponding to each text to be segmented, a directed acyclic graph corresponding to each text to be segmented is constructed, and the path with the highest probability corresponding to each directed acyclic graph is determined.

[0107] Specifically, based on the prefix dictionary of words in the dictionary and the single characters corresponding to each text to be segmented, a directed acyclic graph (DAG) is constructed for each text to be segmented. Then, a path with the highest probability is found on the DAG using dynamic programming. The text is then segmented based on the path with the highest probability corresponding to each DAG to obtain a set of word groups corresponding to each text to be segmented.

[0108] Based on the directed acyclic graph corresponding to each text to be segmented and the corresponding path with the highest probability, the set of word groups corresponding to each text to be segmented is obtained;

[0109] Read the Chinese dictionary and determine whether the set of phrases corresponding to each text to be segmented exists in the Chinese dictionary; if the set of phrases exists in the Chinese dictionary, then mark the keywords of each reservoir modeling literature according to the Chinese dictionary; if the set of phrases does not exist in the Chinese dictionary, then separate the Chinese and English in the set of phrases, and calculate the probability of the phrase sequence corresponding to the separated set of phrases based on the HMM model and Viterbi algorithm to obtain the optimal phrase sequence.

[0110] In HMM part-of-speech tagging, the Viterbi algorithm used is a dynamic programming method that searches for the optimal tagging sequence for a given sentence. If the current node is on the optimal path, then regardless of the subsequent paths of the current node, the source path of the current node must be optimal. The solution for the optimal path can be iterative. After word segmentation, based on the defined dictionary, spatiotemporal information can be quickly annotated. Based on the annotated information, various types of information can be quickly obtained, thereby completing the extraction of spatiotemporal information.

[0111] Specifically, for phrase sets not included in Chinese dictionaries, an HMM (Hidden Markov Model) model is used, and the Viterbi algorithm is employed to find the most likely hidden state sequence as the optimal phrase sequence. Thus, by using a prefix dictionary based on words in the defined dictionary, rapid annotation of spatiotemporal information is achieved. Based on this annotation information, various required information can be quickly obtained, completing the extraction of spatiotemporal information. The fields of the domain dictionary corresponding to each document title are stored in CSV format.

[0112] In some implementations of this embodiment, the above-mentioned extraction of keywords from each reservoir modeling document in the dictionary based on the jieba word segmentation library includes: obtaining the text to be segmented corresponding to each reservoir modeling document in the dictionary; using the text to be segmented corresponding to each reservoir modeling document as the input parameter of the cut function of the jieba word segmentation library to obtain all the segmentation results corresponding to each text to be segmented; based on a preset stop word list, traversing all the segmentation results corresponding to each text to be segmented, performing stop word removal processing on each segmentation result, and saving the words not in the stop word list as keywords.

[0113] Specifically, the process involves loading the text to be segmented using Python's file reading functions and storing the text content in a string variable. Then, the jieba word segmentation library is used. The `cut` function within the jieba library is called, taking the text to be segmented as input. The `cut` function returns a generator containing all segmentation results for that text. This generator is then iterated through to obtain each segmentation result. Stop word removal is performed on the segmented results. A stop word list can be defined, and the results are iterated through, saving words not on the stop word list as keywords. Finally, the extracted keywords are saved, either in a text file or in a database, for subsequent processing and analysis.

[0114] The above implementation first requires loading a file containing text content. Using Python's file operations, the `open()` and `read()` functions can easily read the text file content into a string variable. Next, the jieba word segmentation library is used to process this text. By calling jieba's `cut()` function and passing the read string as input, this function returns a generator containing the word segmentation results, which are then retrieved one by one. To improve the quality and relevance of the text analysis, stop word removal is necessary. Stop word removal can be achieved by creating a list containing common stop words. When processing the word segmentation results, it is only necessary to check whether each word exists in the stop word list; if it is not in the list, it is considered a valid keyword.

[0115] S140: Based on the structured database, construct a knowledge graph to obtain a graph database.

[0116] Specifically, this method collects reservoir modeling literature through various methods, organizes and completes the metadata corresponding to each literature, and simultaneously textualizes the information of each literature to obtain its corresponding textual information. Based on the metadata and textual information of each literature, a dictionary containing spatiotemporal information of modeling knowledge is constructed. This targeted dictionary construction method ensures that the extracted key information is more accurate and refined, effectively improving the knowledge extraction effect. Key information from each literature in the dictionary is extracted and stored in a structured database. Based on the structured database, a knowledge graph is constructed to obtain a graph database. This method automatically collects reservoir modeling literature while using information extraction technology to extract key information from each literature, achieving higher accuracy and comprehensiveness. It can also automatically classify and extract information from reservoir modeling literature, greatly improving work efficiency. Furthermore, based on automating information extraction, the extracted key information is further structured, ultimately constructing a knowledge graph (i.e., a graph database) in the field of reservoir modeling. This approach not only integrates and abstracts information from various literature sources but also clearly presents the relationships between these pieces of information. This knowledge graph is stored in the form of a graph database, enabling convenient and rapid querying and analysis. Through continuous updating and improvement of this knowledge graph, a more comprehensive understanding of various theories, methods, technologies, and cutting-edge knowledge in the field of reservoir modeling can be achieved, providing more accurate support and guidance for practical modeling work.

[0117] In the aforementioned implementation process, this method provides an automated approach to literature collection and information extraction. It can automatically collect literature and extract key information from various sources (including journals, monographs, papers, technical reports, standards, and specifications), thereby significantly improving the efficiency and accuracy of reservoir modeling. Furthermore, compared to existing literature management tools and search engines, this method offers higher accuracy and comprehensiveness, enabling automatic classification and extraction of literature to better meet the professional needs of the reservoir modeling field. Simultaneously, by combining automated information extraction with knowledge graph construction, it integrates and abstracts information from various literature sources, clearly presenting the relationships between information, thus providing more comprehensive and in-depth knowledge support for reservoir modeling. Therefore, this method has broad application prospects in the field of oilfield exploration and development, especially in reservoir modeling. It can significantly improve work efficiency, accuracy, and knowledge integration capabilities, and is of great significance for optimizing oilfield exploration and development strategies and improving oilfield development results. Leveraging a spatiotemporally relevant domain dictionary, this method provides an automated approach to better extract relevant literature from massive datasets and to structure, store, and display it. This method can be used for knowledge extraction tasks that extract relevant geological and spatiotemporal information from a large number of reservoir modeling-related literature, improving the efficiency of domain knowledge extraction and structured knowledge base construction.

[0118] In some embodiments of this example, the above-mentioned construction of a knowledge graph based on a structured database to obtain a graph database includes: constructing a knowledge graph pattern layer containing spatiotemporal information of modeling knowledge; wherein, the knowledge graph pattern layer includes an ontology and multiple entities; establishing associations between each node object of the NEO4J database and each entity to construct the knowledge graph; after disambiguating the structured database, filling the key information in the disambiguated structured database into the node objects corresponding to the NEO4J database after the knowledge graph has been constructed, thereby obtaining the graph database.

[0119] Specifically, the relationships of the knowledge graph schema layer are constructed. Taking the reservoir modeling knowledge base as the target, a knowledge graph schema layer system tailored to spatiotemporal characteristics is designed. For the document storage in the reservoir modeling knowledge base, a knowledge graph schema layer system is constructed as shown in Figure 9. Figure 9This is a schematic diagram of the design of a knowledge graph model layer containing spatiotemporal information of modeling knowledge, provided by one embodiment of the present invention. Taking the reservoir modeling knowledge base domain as the ontology, it sequentially includes document entities, chronostratigraphic entities, geological structural region entities (such as basin structural regions), and administrative division entities. Document entities include document name attributes, document source attributes, author attributes, publication journal attributes, and document abstract attributes. Chronostratigraphic entities include: cosmological entities, boundary entities, system entities, stratum entities, and stage entities. Administrative division entities include: national entities, provincial entities, and county / city entities. A Python preprocessor is written to process and analyze the structured database. The specific process is as follows: Figure 10 This is a flowchart illustrating the implementation of knowledge graph construction using a structured database, as provided in one embodiment of the present invention. Figure 10 As shown, the structured database is disambiguated by referencing different provinces, cities, and counties. For cases where the references are the same but different, the shortest provincial, municipal, and county keywords are compared to link the document keywords to the national code, provincial code, and county / city code. Administrative division codes are used to replace administrative division names, thus ensuring consistency in referencing. The PY2NEO Graph class is used to connect to the NEO4J database. Node labels are created in the NEO4J database for documents, administrative divisions, geological structural regions (such as basin structural regions), and chronostratigraphic units. These node labels will be used to identify and classify different types of nodes when importing data.

[0120] Within the PY2NEO framework, the `create` method of the `Graph` class is used to construct nodes. Data structure initialization is achieved by explicitly defining node labels and properties. Specifically, document entity nodes are configured with attributes such as document name, source, author, publication journal, and abstract; country entity nodes define the country name, capital city name, and country code; province entity nodes include the province name, country of origin, and province code; county / city entity nodes include the county / city name, province, and county / city code; and chronostratigraphic entity nodes encompass the chronostratigraphic name, superordinate chronostratigraphic strata, and the start and end dates of the study. CSV format data is read from Python's CSV module and injected into the node entities. Subsequently, relationship entities corresponding to the relevant node entities are constructed, and the `Graph`'s `Relationship` method is used to define the relationship type, start node, and end node, thus forming the graph structure. During the build process, nodes and relationships are imported into the NEO4J database in batches, and indexes or constraints are considered to optimize query efficiency or ensure data uniqueness.

[0121] In the above implementation process, this method performs data cleaning and extraction based on collected literature, constructs a knowledge graph pattern layer relationship, and designs a knowledge graph pattern layer targeting spatiotemporal characteristics with the reservoir modeling knowledge base as the goal. This layer is used for storing literature in the reservoir modeling knowledge base, such as... Figure 9 The diagram illustrates the construction of a schema layer system for a knowledge graph. The Graph class in py2neo is used to connect to the NEO4J database to build a knowledge graph related to reservoir modeling. The knowledge graph triples extracted through this reservoir modeling knowledge graph construction method are stored in the NEO4J database. This method improves the efficiency of structured data extraction and construction processes in large-scale data extraction tasks within the reservoir modeling field, and speeds up the entire process of extracting and displaying reservoir modeling literature data.

[0122] In some implementations of this embodiment, the collection rules for each reservoir modeling literature include: obtaining multiple reservoir modeling literatures from target information websites based on Python crawlers and Javascript scripts, and organizing and supplementing the metadata of each reservoir modeling literature.

[0123] The target information website can include various sources (including journals, monographs, papers, technical reports, standards and specifications, etc.).

[0124] In some embodiments of this example, before retrieving multiple reservoir modeling documents from the target information website, the method further includes:

[0125] Based on data integration tools, we extracted reservoir modeling literature from various target information websites;

[0126] Compare the reservoir modeling literature from each target information website to identify at least one reservoir modeling literature group; wherein, the reservoir modeling literature group contains multiple reservoir modeling literatures belonging to the same literature.

[0127] For each reservoir modeling literature group, the differences between the reservoir modeling literatures in the reservoir modeling literature group are compared. Based on the comparison results, the reservoir modeling literatures in the reservoir modeling literature group are corrected and filled in sequentially to obtain the processed reservoir modeling literatures.

[0128] The processed reservoir modeling literature was used to replace the corresponding reservoir modeling literature in each target information website.

[0129] Specifically, data integration tools are used to connect and facilitate data flow between different target information websites, achieving data integration and sharing. This allows for the comparison of reservoir modeling literature across various target information websites, identifying multiple reservoir modeling literature groups. Differences between the literature within each group are then compared, and the literature in each group is sequentially corrected and filled. The corrected and filled literature is then used to replace the corresponding reservoir modeling literature on each target information website. This ensures the completeness and accuracy of the reservoir modeling literature across all target information websites.

[0130] In some embodiments of this example, after establishing the association between each node object and each entity in the NEO4J database, the method further includes: constructing a set of graph data based on the node object and the corresponding associated entity, and generating identity information for each set of graph data; wherein, the identity information includes: data source information, data association information, and keyword tag information.

[0131] Specifically, data source information can characterize the source channel of the data. Data association information can characterize information about other reservoir modeling literature associated with the corresponding map data (for example, the data association information of map data A can be information about reservoir modeling literature related to the content or field of the reservoir modeling literature contained in map data A). Keyword tag information can indicate that when a user uses the map database, if the keywords entered by the user match the keyword tag information of the map data, the map data can be returned to the user. Thus, identity identification information can further facilitate users' retrieval of map data in the map database, thereby further improving the retrieval efficiency of the map database and making it more convenient for users.

[0132] In some implementations of this embodiment, the above-mentioned construction of a set of graph data based on node objects and their corresponding associated entities, and the generation of identity information for each set of graph data, includes:

[0133] Read the data extraction method of the current map data, and backtrack the source data of the current map data based on the determined extraction method;

[0134] Determine the source data extraction library and generate corresponding data source information label fields based on the extraction library;

[0135] Data source information is generated based on the data source information tag field, and the source data is compressed and stored in the pre-built source database.

[0136] Specifically, based on the current data extraction method of the map data, source data backtracking is performed to determine the extraction library of the source data corresponding to the current map data, thereby generating a data source information label field. Then, the data source information label field is combined to obtain the data source information of the current map data, and the compressed source data is stored in the pre-built source database. The compressed source data not only represents the source channel of the map data, but also reduces the memory occupied in the source database, thereby improving the data loading efficiency of the source database.

[0137] Based on the knowledge and spatiotemporal information of the current map data, relevant data indexing is performed, and the degree of correlation between each relevant data and the current map data is determined based on the cosine similarity rule;

[0138] Data association information for the current map data is generated based on the relevant data and the degree of association between the relevant data and the current map data.

[0139] Specifically, firstly, relevant data indexing is performed based on the knowledge spatiotemporal information of the current atlas data. Then, according to the cosine similarity rule, the degree of association between each relevant data and the current atlas data is determined. Subsequently, the relevant data are arranged according to the degree of association between each relevant data and the current atlas data. This allows the data association information of the atlas data to not only represent the information of other reservoir modeling literature related to the corresponding atlas data, but also to arrange the relevant data from high to low according to the degree of association between each relevant data and the current atlas data, thus intuitively showing the user the degree of association between each relevant data and the current atlas data.

[0140] Construct a corresponding graph data terminology based on the keywords of the current graph data;

[0141] Calculate the Jaccard similarity between each graph data word set and the pre-constructed graph data word set corresponding to the preset index target. Use all corresponding index targets with Jaccard similarity greater than the preset similarity threshold as the keyword tag information of the current graph data.

[0142] Specifically, based on the pre-constructed terminology set of the current graph data using keywords, the keyword tag information of the current graph data is determined by calculating the Jaccard similarity between each terminology set and the pre-constructed terminology set corresponding to the preset index target. Thus, the current graph data can be accurately indexed using its keyword tag information.

[0143] Figure 2 This is a block diagram of a reservoir modeling knowledge graph construction system provided in one embodiment of the present invention. Figure 2As shown, an embodiment of the present invention provides a reservoir modeling knowledge graph construction system, comprising:

[0144] The text information acquisition module is used to obtain the metadata and text information corresponding to each reservoir modeling document based on the pre-collected reservoir modeling documents;

[0145] The dictionary construction module is used to construct a dictionary containing spatiotemporal information of modeling knowledge based on the metadata and text information corresponding to each reservoir modeling literature;

[0146] The key information extraction module is used to extract key information from the reservoir modeling literature in the dictionary and store the extracted key information in a structured database.

[0147] The knowledge graph construction module is used to construct knowledge graphs based on structured databases, resulting in a graph database.

[0148] Specifically, after collecting reservoir modeling literature through various methods, the system organizes and completes the metadata corresponding to each literature, and simultaneously digitizes the information in each literature to obtain its corresponding textual information. Based on the metadata and textual information of each literature, a dictionary containing spatiotemporal information of modeling knowledge is constructed. This targeted dictionary construction method ensures that the extracted key information is more accurate and refined, effectively improving the knowledge extraction effect. Key information from each literature in the dictionary is extracted and stored in a structured database. Based on this structured database, a knowledge graph is constructed, resulting in a graph database. This system automatically collects reservoir modeling literature while using information extraction technology to extract key information from each literature, achieving higher accuracy and comprehensiveness. It can also automatically classify and extract information from reservoir modeling literature, greatly improving work efficiency. Furthermore, based on automating information extraction, the extracted key information is further structured, ultimately constructing a knowledge graph (i.e., a graph database) in the field of reservoir modeling. This approach not only integrates and abstracts information from various literature sources but also clearly presents the relationships between these pieces of information. This knowledge graph is stored in the form of a graph database, enabling convenient and rapid querying and analysis. Through continuous updating and improvement of this knowledge graph, a more comprehensive understanding of various theories, methods, technologies, and cutting-edge knowledge in the field of reservoir modeling can be achieved, providing more accurate support and guidance for practical modeling work.

[0149] The optional embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above embodiments. Within the scope of the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention.

[0150] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not describe the various possible combinations separately.

[0151] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a microcontroller, chip, or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0152] Furthermore, various different implementations of the present invention can be combined arbitrarily, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed in the present invention.

Claims

1. A method for constructing a knowledge graph for reservoir modeling, characterized in that, include: Based on the pre-collected reservoir modeling literature, the metadata and text information corresponding to each reservoir modeling literature are obtained; Based on the metadata and text information corresponding to each reservoir modeling literature, a dictionary containing spatiotemporal information of modeling knowledge is constructed. Key information from each reservoir modeling literature in the dictionary is extracted and stored in a structured database. Based on structured databases, knowledge graphs are constructed to obtain graph databases.

2. The reservoir modeling knowledge graph construction method according to claim 1, characterized in that, The text information corresponding to each reservoir modeling document includes first text information, second text information and / or text information extracted from PDF format files; Based on the pre-collected literature on reservoir modeling, the text information corresponding to each literature on reservoir modeling is obtained, including: Data cleaning and data extraction were performed on pre-collected reservoir modeling literature based on PDF format files, resulting in text information extracted from PDF format files, CAJ format files, and PDF documents that could not be directly extracted. Based on OCR, text in PDF documents that cannot be directly extracted is extracted to obtain the first text information; Based on a pre-built CAJ information extraction program, the CAJ format file is converted into a text file to obtain the second text information.

3. The method for constructing a reservoir modeling knowledge graph according to claim 2, characterized in that, The OCR-based method extracts text from PDF documents that cannot be directly extracted, obtaining the first text information, including: For PDF documents that cannot be directly extracted, pagination is performed, and each page of the PDF is read as a corresponding image. Based on the BGR2GRAY function, each image is converted into a corresponding grayscale image, and then each grayscale image is binarized using OpenCV. Use the image_to_string package in the Pytesseract library to read the binarized image as a text image; By combining the text and images corresponding to each page of the PDF, the first text information is obtained.

4. The reservoir modeling knowledge graph construction method according to claim 2, characterized in that, The conversion of CAJ format files into text files based on a pre-set CAJ information extraction program includes: The CAJ format file is used as a conversion task, and the conversion task is decomposed into multiple user interaction actions; wherein, each user interaction action corresponds to each CAJ format file in the conversion task, and each CAJ format file in the conversion task corresponds to an initial task icon. Record the clicked icon and click method corresponding to each user interaction action; If the current interactive interface shows whether the corresponding initial task icon exists in the conversion task, and if it does, then the corresponding user interaction action is executed according to the preset order and the corresponding click method to convert the CAJ format file corresponding to the initial task icon into a text file.

5. The method for constructing a reservoir modeling knowledge graph according to claim 1, characterized in that, The key information includes keywords; The extraction of key information from each reservoir modeling literature in the dictionary includes: Based on the jieba word segmentation library, keywords of each reservoir modeling document in the dictionary are extracted, and regular expressions are used to mark the keywords of each reservoir modeling document in the dictionary.

6. The method for constructing a reservoir modeling knowledge graph according to claim 5, characterized in that, The keyword extraction from various reservoir modeling documents in the dictionary based on the jieba word segmentation library includes: Based on a dictionary containing spatiotemporal information of modeling knowledge, a prefix dictionary of words in the dictionary is constructed; The texts to be segmented for each reservoir modeling literature are sequentially cleaned and segmented into Chinese characters to obtain the individual characters corresponding to each text to be segmented. Based on the prefix dictionary of words in the dictionary and the single characters corresponding to each text to be segmented, a directed acyclic graph corresponding to each text to be segmented is constructed, and the path with the highest probability corresponding to each directed acyclic graph is determined. Based on the directed acyclic graph corresponding to each text to be segmented and the corresponding path with the highest probability, the set of word groups corresponding to each text to be segmented is obtained; Read the Chinese dictionary and determine whether the set of word groups corresponding to each text to be segmented exists in the Chinese dictionary; If the set of phrases exists in the Chinese dictionary, then the keywords of each reservoir modeling document are annotated according to the Chinese dictionary; If the phrase set does not exist in the Chinese dictionary, the Chinese and English words in the phrase set are separated, and the probability of the phrase sequence corresponding to the separated phrase set is calculated based on the HMM model and Viterbi algorithm to obtain the optimal phrase sequence.

7. The method for constructing a reservoir modeling knowledge graph according to claim 5, characterized in that, The keyword extraction from various reservoir modeling documents in the dictionary based on the jieba word segmentation library includes: Obtain the text to be segmented corresponding to each reservoir modeling document in the dictionary; The text to be segmented corresponding to each reservoir modeling document is used as the input parameter of the cut function of the jieba word segmentation library to obtain all the word segmentation results corresponding to each text to be segmented. Based on a preset stop word list, all segmentation results corresponding to each text to be segmented are traversed, and stop word removal is performed on each segmentation result. Words not in the stop word list are saved as keywords.

8. The method for constructing a reservoir modeling knowledge graph according to claim 1, characterized in that, The spatiotemporal information of the modeling knowledge includes at least geological structural regional information, administrative division information, and stratigraphic information of the corresponding age.

9. The method for constructing a reservoir modeling knowledge graph according to claim 1, characterized in that, The construction of a knowledge graph based on a structured database yields a graph database, including: Construct a knowledge graph pattern layer containing spatiotemporal information of modeling knowledge; wherein, the knowledge graph pattern layer includes an ontology and multiple entities; Establish the associations between each node object and each entity in the NEO4J database to construct a knowledge graph; After disambiguation of the structured database, the key information from the disambiguated structured database is filled into the corresponding node objects of the NEO4J database after the knowledge graph is constructed, thus obtaining the graph database.

10. The method for constructing a reservoir modeling knowledge graph according to claim 1, characterized in that, The collection rules for literature on reservoir modeling include: Using Python web crawlers and Javascript scripts, multiple reservoir modeling documents were obtained from target information websites, and the metadata of each reservoir modeling document was organized and supplemented.

11. The method for constructing a reservoir modeling knowledge graph according to claim 10, characterized in that, Before retrieving multiple reservoir modeling documents from the target information website, the method further includes: Based on data integration tools, we extracted reservoir modeling literature from various target information websites; Compare the reservoir modeling literature from each target information website to identify at least one reservoir modeling literature group; wherein, the reservoir modeling literature group contains multiple reservoir modeling literatures belonging to the same literature. For each reservoir modeling literature group, the differences between the reservoir modeling literatures in the reservoir modeling literature group are compared. Based on the comparison results, the reservoir modeling literatures in the reservoir modeling literature group are corrected and filled in sequentially to obtain the processed reservoir modeling literatures. The processed reservoir modeling literature was used to replace the corresponding reservoir modeling literature in each target information website.

12. The method for constructing a reservoir modeling knowledge graph according to claim 9, characterized in that, After establishing the associations between each node object and each entity in the NEO4J database, the method further includes: Based on the node objects and their corresponding associated entities, a set of graph data is constructed, and identity identification information is generated for each set of graph data. The identity information includes: Data source information, data association information, and keyword tag information.

13. The method for constructing a reservoir modeling knowledge graph according to claim 12, characterized in that, The process involves constructing a set of graph data based on node objects and their corresponding associated entities, and generating identity information for each set of graph data, including: Read the data extraction method of the current map data, and backtrack the source data of the current map data based on the determined extraction method; Determine the extraction library for the source data, and generate corresponding data source information label fields based on the extraction library; Data source information is generated based on the data source information tag field, and the source data is compressed and stored in a pre-built source database. Based on the knowledge and spatiotemporal information of the current map data, relevant data indexing is performed, and the degree of correlation between each relevant data and the current map data is determined based on the cosine similarity rule; Data association information for the current map data is generated based on the relevant data and the degree of association between the relevant data and the current map data. Construct a corresponding graph data terminology based on the keywords of the current graph data; Calculate the Jaccard similarity between each graph data word set and the pre-constructed graph data word set corresponding to the preset index target. Use all corresponding index targets with Jaccard similarity greater than the preset similarity threshold as the keyword tag information of the current graph data.

14. A reservoir modeling knowledge graph construction system, characterized in that, include: The text information acquisition module is used to obtain the metadata and text information corresponding to each reservoir modeling document based on the pre-collected reservoir modeling documents; The dictionary construction module is used to construct a dictionary containing spatiotemporal information of modeling knowledge based on the metadata and text information corresponding to each reservoir modeling literature; The key information extraction module is used to extract key information from the reservoir modeling literature in the dictionary and store the extracted key information in a structured database. The knowledge graph construction module is used to construct knowledge graphs based on structured databases, resulting in a graph database.

15. A machine-readable storage medium storing instructions thereon, characterized in that, When executed by a processor, the instruction causes the processor to be configured to perform the reservoir modeling knowledge graph construction method as described in any one of claims 1 to 13.

16. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the reservoir modeling knowledge graph construction method according to any one of claims 1 to 13.

Images