Knowledge mining technology-based efficient fractured-vuggy reservoir knowledge management system and method

Abstract

The invention provides an efficient fracture-cavity type oil reservoir knowledge management system and method based on the knowledge mining technology, and the system comprises a knowledge obtaining module, a knowledge storage module, an application auxiliary module, an application management module and a visual display module. Multi-type specialized vocabularies for setting a development process and a matching process are provided, and assistance is provided for a system to realize file processing and analysis; the application management module comprises a label annotation sub-module, a knowledge global processing sub-module, an operation retrieval sub-module and a prediction analysis sub-module, performs classification, clustering and prediction processing on the basis of knowledge with labels, and performs literature similarity analysis by adopting a twin BERT network similarity evaluation model with an attention pooling method to realize knowledge retrieval; and the knowledge quantity change trend of future demand time is effectively analyzed and predicted. By adopting the system, the problems of fragmentation of knowledge storage and tedious management mode in the prior art can be solved, and integrated storage and management of business knowledge are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of knowledge organization and management of oilfield development projects, and in particular to an efficient fracture-cavity oil reservoir knowledge management system and method based on knowledge mining technology. Background Art

[0002] After years of research, key technologies for fracture-vuggy reservoir development have made significant progress in reservoir description, geological modeling, numerical simulation, water and gas injection for enhanced oil recovery, and supporting engineering technologies. This vast amount of knowledge requires effective management using information technology. In particular, the use of knowledge mining technology goes beyond simple query and maintenance to enable higher-level processing and analysis of this data to derive insights into its overall characteristics and predict development trends. Knowledge mining is the nontrivial process of identifying valid, novel, potentially useful, and ultimately understandable patterns from a dataset. Integrating research methods, processes, and case studies across disciplines to achieve seamless connectivity between diverse technologies is crucial for effectively managing the knowledge behind key technologies for fracture-vuggy reservoir development.

[0003] However, as a new reservoir type, carbonate fracture-vuggy reservoirs have already yielded a wealth of research results. However, a lack of knowledge management methods has led to cumbersome knowledge management, fragmented knowledge storage, and inconsistent knowledge presentation. Some existing studies address the need for knowledge management technologies for fracture-vuggy reservoirs, but lack clear implementation guidance or suffer from significant application deficiencies. For example, the article "Constructing a Carbonate Reservoir Knowledge Base and Its Application in Geological Modeling of Fracture-vuggy Reservoirs," published in Petroleum Exploration and Development, explored knowledge management methods for fracture-vuggy reservoirs based on knowledge mining, but provided only a partial theoretical introduction without further research and implementation. Another article, "Discussion on the Application of Data Mining Technology in Oilfield Informatization Construction," published in Tianjin Chemical Industry, proposed that data mining technology could be used in oilfield production enterprises to accurately assess instability factors during the bidding phase and establish an oilfield data mining platform. While theoretically expounding and envisioning this approach, it lacked specific implementation. The paper "Construction of a Natural Gas Geological Information Sharing Platform Based on Dynamic Storage" published in Natural Gas Industry uses data mining methods to form an algorithm parameter "setting guide" and constructs a "method library dictionary", realizing the shared management of natural gas geological structured data and the dynamic storage and search of unstructured data, and realizing the integrated application of natural gas geological data mining algorithms and their parameter tuning management; however, it still has shortcomings such as the high degree of coupling of multi-library systems and the low visualization of the data mining process.

[0004] The information disclosed in the background technology section of the present invention is only intended to deepen the understanding of the general background technology of the present invention, and should not be regarded as an admission or any form of suggestion that the information constitutes the prior art known to those skilled in the art. Summary of the Invention

[0005] To solve the above problems, the present invention provides an efficient fracture-cavity oil reservoir knowledge management system and method based on knowledge mining technology. The system includes a knowledge acquisition module, a knowledge storage module, an application auxiliary module, an application management module and a visualization display module. The knowledge storage module includes a professional vocabulary library, which provides multiple types of professional vocabulary for setting development processes and supporting processes, and provides assistance for the system to implement file processing and analysis; the application management module includes a label annotation submodule, a knowledge global processing submodule, an operation retrieval submodule and a prediction analysis submodule, which performs classification, clustering and prediction processing based on knowledge with labels, and uses a twin BERT network similarity evaluation model with an attention pooling method to perform document similarity analysis to achieve knowledge retrieval, and effectively analyze and predict the change trend of knowledge volume in future demand time. The use of this system can overcome the problems of fragmented knowledge storage and cumbersome management methods in the existing technology, and realize the integrated storage and management of business knowledge. Preferably, in one embodiment, the system is built based on an intelligent computer system and applies a multi-layer network architecture service, including a data acquisition layer, a data storage layer, an application support layer and a core application layer:

[0006] The knowledge acquisition module is provided in the data acquisition layer, and acquires different types of fracture-cavity reservoir knowledge files according to the multi-source knowledge approach required by fracture-cavity reservoirs, including basic data knowledge, business data knowledge and existing database objects;

[0007] The knowledge storage module is provided in the data storage layer and is used to provide knowledge storage, knowledge management and auxiliary services to users based on knowledge files;

[0008] The application auxiliary module is set up in the application support layer to provide permission management, file management, model management and internal data management support for knowledge management;

[0009] An application management module is provided at the core application layer and is used to provide key technology management, knowledge resource management, team management, training management, and status progress management services based on knowledge documents in response to user intentions;

[0010] A visualization display module, which is used to visualize the output results in response to user control based on the interactive interface;

[0011] The knowledge storage module includes a professional vocabulary database, including multiple types of professional vocabulary for setting development processes and supporting processes, which is used to assist the system in implementing file processing and analysis;

[0012] The application management module includes a label annotation submodule, which is configured to analyze the frequency of professional terms in the fracture-cavity reservoir knowledge file based on the professional word library, set the vocabulary label of the file and annotate it;

[0013] The application management module includes a knowledge global processing submodule, which is configured to perform classification, clustering and prediction processing based on multiple fracture-cavity reservoir knowledge objects with labels to be processed, and convert the knowledge files into structured characteristic files;

[0014] The application management module is provided with an operation retrieval submodule, including a knowledge retrieval engine and a sorting processing unit. The knowledge retrieval engine uses a twin BERT network similarity evaluation model with an attention pooling method to perform document similarity analysis to implement knowledge retrieval services. The sorting processing unit is used to sort the retrieved knowledge objects according to a relevance sorting algorithm for intelligent knowledge display.

[0015] The application management module includes a prediction analysis submodule, which is used to perform statistical analysis algorithms and modeling techniques based on existing knowledge data to predict the trend of knowledge volume changes in future demand time.

[0016] Optionally, in one embodiment, the set development process includes seismic imaging and prediction, reservoir description, geological modeling, numerical simulation, water injection and gas injection processes.

[0017] Furthermore, in one embodiment, the label annotation submodule is configured to read the text content, and if there are pictures, use optical character recognition technology to identify the picture information, combine it with the text content to form file content, and then perform word segmentation on the file content according to the professional vocabulary. Based on this, the frequency of professional vocabulary is statistically calculated, and words with high frequency are selected as vocabulary tags for the text and annotated.

[0018] Preferably, in one embodiment, the global knowledge processing submodule is configured to use the vocabulary tags of the file as file features, convert the file feature text into text vector form through the input layer of the BERT model, and then merge the word vector, text vector, and position vector in the input layer through the word embedding layer to obtain the word embedding representation of each word, further extract the features through the encoding layer of the attention mechanism based on the word embedding representation, and finally further extract the features through the attention pooling layer to obtain the final document and citation features of the file, forming a structured feature file.

[0019] In an optional embodiment, the knowledge retrieval engine introduces the idea of ​​attention into the twin BERT network similarity evaluation model, uses both whole sentence feature representation and single character feature representation to perform attention pooling, completes the extraction of important features, and realizes document similarity calculation.

[0020] In one embodiment, the knowledge retrieval engine implements document similarity calculation using a twin BERT network similarity evaluation model with an attention pooling method according to the following operations:

[0021] The similarity between the document and each cited document is calculated using two similarity evaluation functions: inner product similarity and cosine similarity.

[0022] Calculate the proportion of similarity of all citations pointed to by the document to the total;

[0023] The calculated proportion is used to replace the transfer matrix of the average distribution method, and the PageRank algorithm is iteratively executed until the set stability requirements are met. The calculation result is output as the document similarity result to realize knowledge document retrieval.

[0024] Furthermore, in one embodiment, the knowledge search engine calculates the similarity between the document and each cited document by the following operations:

[0025] Compare the overall features and individual character features of the BERT word embedding layer and the BertEncoder layer, calculate the attention weight, and then determine the attention pooling feature extraction formula based on it;

[0026] The output results of the attention pooling layer of the two twin sub-networks are obtained respectively, and the cosine similarity of the two output result vectors is calculated as the similarity between the document and the cited document.

[0027] Preferably, in one embodiment, the knowledge search engine calculates the proportion of the similarity of all citations pointed to by the document to the total similarity by the following operation:

[0028] A random walk model is defined on a directed graph constructed based on documents and citations, and the proportion of the similarity of all citations pointed to by the documents is calculated using the basic transfer matrix of the random walk model and the completely random transfer matrix.

[0029] Specifically, in an optional embodiment, the knowledge retrieval engine determines the following attention pooling feature extraction formula:

[0030]

[0031] Among them, the attention weight α i =Softmax(sim(x i ·e i )), x i 、e i ∈R d , R d represents the set of all d-dimensional real vectors, d is the dimension of the word vector, S represents the output of the attention pooling layer, x i represents the overall features of the ith citation document after the BERT word embedding layer and the BertEncoder layer, e irepresents the single character feature of the ith citation document after the BERT word embedding layer and the BertEncoder layer; N represents the number of citation documents; R k Represents a set of k-dimensional real vectors.

[0032] In one embodiment, the knowledge search engine calculates the overall proportion of similarities of all citations pointed to by the document according to the following formula:

[0033]

[0034] Among them, W i S is the proportion of the similarity of the i-th reference of the document to the total similarity of the references i is the similarity between the document and the i-th reference, and N is the number of references corresponding to the document.

[0035] Preferably, in one embodiment, the knowledge retrieval engine iteratively executes the PageRank algorithm according to the following formula:

[0036]

[0037] Among them, 1 is an n-dimensional vector with all components equal to 1, t is the number of iterations of the algorithm, R t is the PR value after the tth iteration, d is the damping coefficient, and n is the number of nodes in the directed graph. The stability condition is set to satisfy the difference between the PR value obtained in the previous iteration and the PR value of this iteration is less than the preset threshold. The result is considered stable and the calculation result is output at this time.

[0038] Based on the application of the method described in any one or more of the above embodiments, the present invention further provides an efficient fracture-vuggy reservoir knowledge management method based on knowledge mining technology, which is applied to the system described in any one or more of the above embodiments, and includes:

[0039] The knowledge acquisition module is used to obtain different types of fracture-cavity reservoir knowledge files, including basic data knowledge, business data knowledge and existing database objects, based on the multi-source knowledge approach required by fracture-cavity reservoirs.

[0040] The knowledge storage module provides users with knowledge storage, knowledge management and auxiliary services based on knowledge files. It uses a preset professional vocabulary to assist the system in file processing and analysis. The professional vocabulary includes various professional terms for setting development processes and supporting processes.

[0041] Use application auxiliary modules to provide authority management, file management, model management and internal data management support for knowledge management;

[0042] Provide key technology management, knowledge resource management, team management, training management, and status progress management services based on knowledge documents in response to user intent through the application management module;

[0043] Using a visualization display module to visualize the output results in response to user control based on an interactive interface;

[0044] The process of providing key technology management and knowledge resource management services through the application management module includes:

[0045] Analyzing the frequency of professional terms in the fracture-cavity reservoir knowledge file based on the professional term database, setting vocabulary tags for the file and annotating it;

[0046] Based on the multiple labeled fracture-cavity reservoir knowledge objects to be processed, classification, clustering and prediction processing are performed to convert the knowledge files into structured characteristic files;

[0047] Use the twin BERT network similarity evaluation model with attention pooling method to perform document similarity analysis to realize knowledge retrieval service; sort the retrieved knowledge objects according to the relevance ranking algorithm;

[0048] Based on the statistical analysis algorithms and modeling techniques of existing knowledge data, the trend of changes in knowledge volume in future demand time is predicted.

[0049] Compared with the closest prior art, the present invention also has the following beneficial effects:

[0050] The present invention provides an efficient fracture-cavity oil reservoir knowledge management system and method based on knowledge mining technology. The system includes a knowledge acquisition module, a knowledge storage module, an application assistance module, an application management module, and a visualization display module. The knowledge storage module includes a professional vocabulary library, which provides multiple types of professional vocabulary for setting development processes and supporting processes, and assists the system in implementing file processing and analysis. By establishing a key professional vocabulary library with characteristics of fracture-cavity oil reservoirs, it helps to acquire and manage dispersed knowledge of fracture-cavity oil reservoirs in a targeted manner.

[0051] The application management module features a tag annotation submodule, which uses automatic text tagging technology and introduces optical character recognition technology to improve distinctive tag annotations. The global knowledge processing submodule performs classification, clustering, and prediction based on tagged knowledge, converting knowledge files into structured, feature-rich files. This fully characterizes the characteristics of each knowledge object, facilitating subsequent knowledge analysis, management, and retrieval operations, improving operational efficiency and accuracy.

[0052] The computational retrieval submodule uses a twin BERT network similarity evaluation model with an attention pooling method to perform document similarity analysis to achieve knowledge retrieval, overcoming the defect of the traditional PageRank algorithm that cannot reasonably assign weights. It estimates the similarity between documents and citations, and then uses the similarity calculated by the model as the weight for assigning the PR value of the document, efficiently achieving high-quality retrieval that meets user intentions. The system is used to comprehensively explore the implicit knowledge of fracture-cavity oil reservoirs, realize the integrated management of business knowledge, the storage and management of experiential knowledge, and achieve the tangibility of intangible knowledge, the integration of fragmented knowledge, and the process-based development of dispersed knowledge, thereby promoting the development of key technologies and R&D work for fracture-cavity oil reservoirs.

[0053] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purposes and other advantages of the present invention can be realized and obtained through the structures particularly pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0055] Figure 1 This is a schematic diagram of the structural principle of an efficient fracture-cavity oil reservoir knowledge management system based on knowledge mining technology provided by one embodiment of the present invention;

[0056] Figure 2 This is a schematic diagram of the overall system architecture distribution of an efficient fracture-cavity oil reservoir knowledge management system based on knowledge mining technology provided by an embodiment of the present invention;

[0057] Figure 3 This is a schematic diagram of the knowledge combing principle of the knowledge storage module in the efficient fracture-cavity oil reservoir knowledge management system based on the knowledge mining technology provided by one embodiment of the present invention;

[0058] Figure 4 This is a schematic diagram of a file label annotation process of an efficient fracture-cavity oil reservoir knowledge management system based on knowledge mining technology provided by an embodiment of the present invention;

[0059] Figure 5 This is a schematic diagram of the knowledge retrieval PageRank algorithm principle of an efficient fracture-cavity oil reservoir knowledge management system based on knowledge mining technology provided by another embodiment of the present invention;

[0060] Figure 6 This is a schematic diagram of the overall functional summary architecture of an efficient fracture-cavity oil reservoir knowledge management system based on knowledge mining technology provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The following will describe in detail the implementation methods of the present invention in conjunction with the accompanying drawings and embodiments, so that practitioners of the present invention can fully understand how the present invention applies technical means to solve technical problems and achieve the implementation process of technical effects, and can implement the present invention in accordance with the above implementation process. It should be noted that as long as no conflict exists, the various embodiments and various features of each embodiment in the present invention can be combined with each other, and the resulting technical solutions are all within the scope of protection of the present invention.

[0062] Although the flowcharts depict the operations as sequential processes, many of the operations can be performed in parallel, concurrently, or simultaneously. The order of the operations can be rearranged. A process can be terminated when its operations are completed, but can also have additional steps not included in the figures. A process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0063] Computer devices include user devices and network devices. User devices or clients include, but are not limited to, computers, smartphones, and PDAs (Personal Digital Assistants). Network devices include, but are not limited to, a single network server, a server group consisting of multiple network servers, or a cloud computing-based cloud consisting of a large number of computers or network servers. Computer devices can operate independently to implement the present invention, or they can connect to a network and interact with other computer devices in the network to implement the present invention. The network in which the computer device resides includes, but is not limited to, the Internet, wide area networks, metropolitan area networks, local area networks, and VPN networks.

[0064] The terms "first," "second," and the like may be used herein to describe various elements, but these elements should not be limited by these terms, and these terms are used merely to distinguish one element from another. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items. When an element is referred to as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element, or intervening elements may be present.

[0065] The terms used herein are intended only to describe specific embodiments and are not intended to limit exemplary embodiments. Unless the context clearly indicates otherwise, the singular forms "a", "an", "an item" used herein are also intended to include the plural. It should also be understood that the terms "comprise" and / or "include" used herein specify the presence of stated features, integers, steps, operations, units and / or components, and do not preclude the presence or addition of one or more other features, integers, steps, operations, units, components and / or combinations thereof.

[0066] It is very important for oil companies to carry out knowledge management. Some foreign oil companies have achieved good results by implementing knowledge management systems. For example:

[0067] British oil company BP uses T-type management within its enterprise to establish virtual work teams. Schlumberger has implemented various practice community forums and document areas on its intranet, enabling employees from all locations to collaborate and achieve excellent results. Mobil Oil has established an information / knowledge sharing platform to fully share and apply information, knowledge, and experience within the company. C&C, a US company, provides upstream oil and gas clients with an expert knowledge base system for exploration and development analogy decision-making and related professional services. This system is staffed by senior European and American petroleum geoscientists, specializing in reservoir analogy research. C&C's core technologies include the Global Large Oil and Gas Field Analogy Decision Expert System (DAS Expert System) and the Optimization Knowledge Base for Exploration and Development Decisions in Various Oil and Gas Fields (EPS Knowledge Base).

[0068] To improve oil recovery, a knowledge management platform for key technologies in fracture-vuggy reservoirs has been established, integrating data, knowledge, and business processes. This platform controls business processes, stores and manages empirical knowledge, and promotes communication, collaboration, and knowledge sharing across projects, making intangible knowledge tangible, integrating fragmented knowledge, and streamlining distributed knowledge. Knowledge engineering technologies are used to integrate reservoir description, geological modeling, and numerical simulation. Search engine-driven, minimalist applications and intelligent presentation enable the representation, storage, management, mining, and display of tacit knowledge. Cutting-edge technologies such as artificial intelligence are used to integrate research methods, processes, and case studies across disciplines, achieving seamless connectivity between different technologies. This effectively manages key technical knowledge in fracture-vuggy reservoir development.

[0069] After years of research, key technologies for fracture-vuggy reservoir development have made significant progress in reservoir description, geological modeling, numerical simulation, water and gas injection for enhanced oil recovery, and supporting engineering technologies. This vast amount of knowledge requires effective management using information technology. In particular, knowledge mining technology goes beyond simple query and maintenance to enable higher-level processing and analysis of this data, revealing its overall characteristics and predicting development trends. Knowledge mining is the nontrivial process of identifying valid, novel, potentially useful, and ultimately understandable patterns from a dataset. Using knowledge mining technology, we can deeply analyze the scientific research results and technologies developed over the years from major special projects led by the Petroleum Exploration and Development Research Institute, establishing a knowledge management platform for key technologies for fracture-vuggy reservoir development. This will facilitate the exchange and sharing of information, technology, and knowledge across various research topics. The research process for key fracture-vuggy reservoir development technologies, including training, processes, demonstrations, and novelty searches, will be integrated with project technical methods, including geophysical prediction, geological modeling, numerical simulation, reservoir engineering, and reservoir technology for fracture-vuggy reservoirs, as well as typical project cases and technical achievements, including patents, software, C&C, and technical benchmarking, to form an integrated, tangible, and process-based knowledge representation and management approach. This will lay the foundation for improving the research and application level of major special projects, the development of core and distinctive technologies, and personnel training.

[0070] Some existing research addresses the need for knowledge management technologies for fracture-vuggy reservoirs, but lacks clear implementation guidance or exhibits significant application deficiencies. For example, the article "Constructing a Carbonate Reservoir Knowledge Base and Its Application in Geological Modeling of Fracture-vuggy Reservoirs," published in Petroleum Exploration and Development, attempts a knowledge management approach for fracture-vuggy reservoirs based on knowledge mining. However, it provides only a theoretical introduction and lacks further research and implementation. The article "Discussion on the Application of Data Mining Technology in Oilfield Informatization Construction," published in Tianjin Chemical Industry, proposes that data mining technology can be applied to accurately assess instability factors during the bidding phase and to establish an oilfield data mining platform within oilfield production enterprises. While providing theoretical explanations and ideas, it lacks specific implementation. The article "Constructing a Natural Gas Geological Information Sharing Platform Based on Dynamic Storage," published in Natural Gas Industry, employs data mining methods to develop a "setting guide" for algorithm parameters and constructs a "method library dictionary," enabling the shared management of structured natural gas geological data and the dynamic storage and search of unstructured data. It also implements the integrated application of natural gas geological data mining algorithms and their parameter tuning management. However, it still suffers from limitations such as a high degree of coupling between multiple database systems and low visualization of the data mining process.

[0071] There are no successful precedents for fracture-vuggy reservoir development abroad. Through scientific and technological research, a series of fracture-vuggy reservoir development technologies have been developed, supporting the development of large fracture-vuggy reservoirs worldwide. However, as a new reservoir type, carbonate fracture-vuggy reservoirs have already yielded a wealth of research results. However, the lack of knowledge management methods has led to cumbersome knowledge management, fragmented knowledge storage, and inconsistent knowledge display methods. The purpose of this invention is to provide an efficient fracture-vuggy reservoir knowledge management system based on knowledge mining technology, enabling core knowledge extraction, prediction, and association, discovering the knowledge underlying oilfield development, and effectively supporting the development and research of fracture-vuggy carbonate reservoirs.

[0072] To solve the above problems, the present invention provides an efficient fracture-cavity oil reservoir knowledge management system and method based on knowledge mining technology, which mines, screens, processes, organizes and aggregates multi-dimensional and multi-source knowledge resources, and sorts, classifies and organizes the mined, screened and processed knowledge so that users can easily find, browse and use it; realizes centralized management and sharing of knowledge, thereby enhancing its value; aims to solve the problems of cumbersome existing fracture-cavity oil reservoir knowledge management methods, fragmented knowledge storage, and inconsistent knowledge display methods, realizes core knowledge extraction, prediction and association functions, and solves the problem of scattered knowledge storage in fracture-cavity oil reservoirs.

[0073] The following describes in detail the structural components, connection methods, and functional principles of the system according to the embodiment of the present invention based on the accompanying drawings. Although the logical order of each operation is shown in the process of describing the system structure and operating principle, in some cases, the operations shown or described may be performed in a different order than here.

[0074] Example 1:

[0075] Figure 1 The schematic diagram of the structure principle of the efficient fracture-cavity reservoir knowledge management system based on knowledge mining technology provided by the first embodiment of the present invention is shown. Figure 1 It can be seen that the system is built on an intelligent computer system and uses a multi-layer network architecture service, including a data acquisition layer, a data storage layer, an application support layer, and a core application layer:

[0076] The system includes a knowledge acquisition module, which is set in the data acquisition layer and acquires different types of fracture-cavity reservoir knowledge files according to the multi-source knowledge approach required by fracture-cavity reservoirs, including basic data knowledge, business data knowledge and existing database objects;

[0077] The knowledge storage module is provided in the data storage layer and is used to provide knowledge storage, knowledge management and auxiliary services to users based on knowledge files;

[0078] The application auxiliary module is set up in the application support layer to provide permission management, file management, model management and internal data management support for knowledge management;

[0079] An application management module is provided at the core application layer and is used to provide key technology management, knowledge resource management, team management, training management, and status progress management services based on knowledge documents in response to user intentions;

[0080] The visualization display module is used to visualize the output results in response to user control based on the interactive interface.

[0081] The knowledge storage module includes a professional vocabulary database, including multiple types of professional vocabulary for setting development processes and supporting processes, which is used to assist the system in implementing file processing and analysis;

[0082] The application management module includes a label annotation submodule, which is configured to analyze the frequency of professional terms in the fracture-cavity reservoir knowledge file based on the professional word library, set the vocabulary label of the file and annotate it;

[0083] The application management module includes a knowledge global processing submodule, which is configured to perform classification, clustering and prediction processing based on multiple fracture-cavity reservoir knowledge objects with labels to be processed, and convert the knowledge files into structured characteristic files;

[0084] The application management module is provided with an operation retrieval submodule, including a knowledge retrieval engine and a sorting processing unit. The knowledge retrieval engine uses a twin BERT network similarity evaluation model with an attention pooling method to perform document similarity analysis to implement knowledge retrieval services. The sorting processing unit is used to sort the retrieved knowledge objects according to a relevance sorting algorithm for intelligent knowledge display.

[0085] The application management module includes a prediction analysis submodule, which is used to perform statistical analysis algorithms and modeling techniques based on existing knowledge data to predict the trend of knowledge changes in future demand time. Figure 2 shown.

[0086] The efficient fracture-vuggy reservoir knowledge management system based on knowledge mining technology in the embodiments of the present invention provides users with a knowledge management platform that integrates key technical knowledge of fracture-vuggy reservoirs. This platform enables the integrated management of business knowledge, the storage and management of experiential knowledge, and promotes communication, collaboration, and knowledge sharing across projects, thereby tangibleizing intangible knowledge, integrating fragmented knowledge, and streamlining distributed knowledge. This invention can support the subsequent integrated process management of fracture-vuggy reservoirs, using knowledge management technology to support application analysis and management such as reservoir description, geological modeling, and numerical simulation.

[0087] Knowledge management of key technologies for fracture-vuggy reservoir development serves as a bridge between user information needs and information resources. System design is a core issue in information resource development, requiring full consideration of factors such as the application environment, resource quantity and type, disciplinary characteristics, user types, and the existing knowledge management system foundation. The fundamental principle of knowledge management system design is to maximize the description of data resources and content based on an analysis of existing resources and user needs, describing and expressing them as comprehensively and accurately as possible to meet user needs. This includes incorporating more relevant attributes, keywords, classification tags, and other information, as well as providing richer content and diverse expression methods to effectively deliver data resources that meet user needs.

[0088] The efficient fracture-cavity oil reservoir knowledge management system based on knowledge mining technology provided by an embodiment of the present invention includes a knowledge acquisition module, which is set in the data acquisition layer and obtains different types of fracture-cavity oil reservoir knowledge files according to the multi-source knowledge approach required by the fracture-cavity oil reservoir, including basic data knowledge, business data knowledge and existing database objects.

[0089] The knowledge acquisition module has a metadata management substructure, which provides the data acquisition foundation for the entire system, including metadata definition and management functions. Based on the requirements of the initial system construction of the present invention, the metadata definition of fracture-vuggy reservoir knowledge is clarified, and the meaning and standards of related knowledge resources are defined.

[0090] At the same time, the metadata management substructure realizes the management function for metadata, including: (1) the design and formulation of data structure specifications in the stages of data collection, uploading, aggregation, analysis and display; (2) the systematic sorting and display of fracture-vuggy reservoir knowledge according to a specific structure, so that the knowledge data in the system can be interpreted, analyzed and managed throughout the process.

[0091] The knowledge acquisition module includes a data upload substructure, which is used to implement the planning and uploading of knowledge data. Specifically, it includes the following operations:

[0092] (1) Planning knowledge data based on the definition and standards of the metadata management substructure, allowing each department to upload fine-grained knowledge based on needs and flexibly adjust upload tasks based on actual conditions; in embodiments of the present invention, the process of uploading, submitting, or storing data or knowledge in smaller, more detailed units. In contrast to coarse-grained upload, fine-grained upload allows users to split data into smaller parts for more flexible management, search, and processing.

[0093] (2) The knowledge of various fracture-vuggy reservoirs can be collected through online reporting by departments, import and load of historical data, and other data collection services. The online reporting function mainly uses online forms or interfaces to input and submit data; the import and load of historical data mainly converts existing knowledge data into a format that can be recognized by the system and imports it into the system. The import and load methods include batch import, file upload, API interface, etc.; data collection mainly realizes the function of collecting data from various sources.

[0094] (3) The data collected from different paths are aggregated and imported into the database. In addition, metadata management personnel are set up in this substructure and given corresponding authority control to meet the needs of adding, deleting, modifying and checking metadata.

[0095] The knowledge of fracture-cavity reservoirs involves patents, software works, technical documents and other types of knowledge about fracture-cavity reservoirs. The file types of the knowledge content include .doc, .docx, .pdf, .pptx and other types.

[0096] The knowledge acquisition module also includes a data integration and processing substructure, which is used to convert, load and process various types of data in the metadata database and knowledge database formed by the above metadata management substructure and data upload substructure, and upload them to the data warehouse of the knowledge storage module.

[0097] The knowledge storage module is responsible for integrating and storing fracture-vuggy reservoir knowledge from the data acquisition layer according to database design. Through the metadata management substructure, knowledge is divided, labeled, and stored according to key technologies, latest progress reports, and knowledge resources. Key technologies are organized into a structured hierarchy of primary technologies and secondary technologies within them. Latest progress reports are divided by time period, and knowledge resources store and perform statistical analysis on external fracture-vuggy reservoir knowledge. The system also sets up a user permission table, assigning different permissions to different users to view and modify system platform content.

[0098] Based on the embodiment of the present invention, each business department of the fracture-vuggy oil reservoir uploads the corresponding knowledge according to the prescribed data structure (implemented by the data upload substructure). The system platform converts, processes, and integrates the acquired data through the data integration and processing substructure, and finally transmits it to the data storage layer to complete the construction of the system database.

[0099] The acquired knowledge content can be flexibly managed in the knowledge storage module based on demand, and can be categorized and stored in different formats according to user settings. For example, it can be divided primarily based on the knowledge information resources of the multi-level optimization operation schedule cycle. Knowledge from the initial optimization operation schedule cycle and the early optimization operation schedule cycle includes research on fracture-vuggy reservoir identification and description technology, research on three-dimensional geological modeling technology for fracture-vuggy carbonate reservoirs, research on technology for improving development effectiveness of fracture-vuggy carbonate reservoirs, research on key technologies for the development of complex fractured carbonate reservoirs, complex media numerical simulation technology and software, and research on efficient acid fracturing technology for fracture-vuggy carbonate reservoirs. The six categories of content include project self-evaluation report, project technical summary report, project achievement report multimedia file, task contract, task contract multimedia text, acceptance application, achievement review opinion, contract review opinion, special feasibility study report, project research report, task acceptance report, and mid-term inspection report; the knowledge of the later optimization operation progress cycle includes seismic identification and prediction technology of different fracture-cavity reservoirs, detailed description and geological modeling technology of fracture-cavity oil reservoirs, fracture-cavity oil reservoir improved water drive to increase recovery rate technology, fracture-cavity oil reservoir gas injection to increase recovery rate technology, fracture-cavity oil reservoir plugging and targeted acid fracturing process technology. The five categories of content include achievement certification materials, task contract, special feasibility study report, task contract multimedia text, task acceptance report, mid-term inspection report, and mid-term inspection multimedia. The knowledge combing principle diagram of the knowledge storage module of the system of the embodiment of the present invention is as follows: Figure 3 shown.

[0100] The knowledge storage module includes a professional vocabulary database, including multiple types of professional vocabulary for setting development processes and supporting processes, which is used to assist the system in implementing file processing and analysis;

[0101] The system sets up an application auxiliary module at the application support layer to provide permission management, file management, model management and internal data management support for knowledge management.

[0102] Among them, permission management is associated with the user permission table in the data storage layer, and performs fine-grained management of the permissions of system users, including role permission definition, user permission allocation, permission verification mechanism, etc.; document management is the structured management of knowledge documents in the system; model management is the management of algorithmic models such as classification, clustering, and identification in the system; internal data management is the management of important information in the system, such as reservoir data, experimental data, production data, internal documents and reports on reservoir research, etc.

[0103] The system's application management module is set up in the core application layer, which is used to respond to user intentions and provide key technology management, knowledge resource management, team management, training management and status progress management services based on knowledge files;

[0104] The application management module of the embodiment of the present invention includes seven functions: key technology management, user management, knowledge resource management, standard and specification management, latest progress management, training courseware management, and other information management.

[0105] Key technology management includes basic information management, special achievement management, tangible product management and sub-technology management; user management includes user basic information management and authority allocation management; knowledge resource management includes basic knowledge management and human point research and analysis; standards and specifications management includes basic knowledge management of standards and specifications; latest progress management includes real-time progress management and base station report management; training courseware management includes basic information management of courseware; other information management includes annotation and description information management, technical team management and basic personnel information management.

[0106] The application management module includes a label annotation submodule, which is configured to analyze the frequency of professional terms in the fracture-cavity reservoir knowledge file based on the professional word library, set the vocabulary label of the file and annotate it;

[0107] The set development process includes seismic imaging and prediction, reservoir description, geological modeling, numerical simulation, water injection and gas injection technology.

[0108] The label annotation submodule is configured to read the text content, and if there is an image file, use optical character recognition technology to identify the image information, combine it with the text content to form the file content, and then perform word segmentation on the file content according to the professional vocabulary. Based on this, the frequency of professional vocabulary is counted, and words with high frequency are selected as vocabulary tags for the text and annotated.

[0109] Specifically, in an optional embodiment, the embodiment of the present invention adopts text automatic labeling technology, uses statistical thinking to sort the frequency of occurrence of professional vocabulary in the text, and selects vocabulary with high frequency as labels for the text.

[0110] If the content in the document is text, it is directly read. If the image content in the document exists, optical character recognition (OCR) technology is used to recognize it. After obtaining the content in the document, the content is segmented according to the established vocabulary, and then the word frequency statistics of the segmentation results are performed to achieve the overall process of label annotation. Figure 4 shown.

[0111] In practical applications, this invention establishes a specialized vocabulary for fracture-cavity reservoirs through contributions from domain experts, literature research, and domain text data mining. This specialized vocabulary is designed to enhance the system's text processing and analysis capabilities. It contains specialized terminology and domain vocabulary related to reservoirs and can be designed as a dataset or a list file within a database.

[0112] The embodiment of the present invention establishes a key professional vocabulary database with the characteristics of fracture-vuggy oil reservoirs, adopts text automatic labeling technology, introduces optical character recognition technology, improves characteristic label annotations, and classifies, clusters, and performs prediction operations on fracture-vuggy oil reservoir knowledge, realizing the integrated management of business knowledge and the storage and management of experiential knowledge, promoting communication, collaboration, and knowledge sharing among various subjects, and achieving the tangibility of intangible knowledge, the integration of fragmented knowledge, and the process-based operation of dispersed knowledge.

[0113] The application management module includes a knowledge global processing submodule, which is configured to perform classification, clustering and prediction processing based on multiple fracture-cavity reservoir knowledge objects with labels to be processed, and convert the knowledge files into structured characteristic files;

[0114] The global knowledge processing submodule is configured to use the vocabulary tag of the file as the file feature, and convert the file feature text into text vector form through the input layer of the BERT model. Then, through the word embedding layer, the word vector, text vector, and position vector in the input layer are merged to obtain the word embedding representation of each word. Based on the word embedding representation, the vocabulary tag features are further extracted through the encoding layer of the attention mechanism. Finally, the vocabulary tag features are further extracted through the attention pooling layer to obtain the final document and citation features of the file, forming a structured characteristic file.

[0115] After completing the label annotation of the fracture-cavity reservoirs independently, the grey prediction, Holt-Winters and Prophet methods were used to classify, cluster and predict the knowledge of fracture-cavity reservoirs.

[0116] Based on this, tags are used as document keywords to represent the core themes of documents, and tag annotations can be directly used as document features. The specific process for processing document text features is as follows: the features are first passed through the input layer of the BERT model to convert the document feature text into a vector form. Then, the word vectors, text vectors, and position vectors in the input layer are merged through the word embedding layer to obtain a word embedding representation for each word. The word embedding representation is further extracted from the encoding layer of the attention mechanism. Finally, the features are further extracted through the attention pooling layer to obtain the final document and citation features. This method overcomes the high-dimensional sparsity problem of text data and transforms document information into a more structured and usable feature representation, which helps to better explore the core themes and relationships in the document, improving model performance and interpretation capabilities.

[0117] The efficient fracture-cavity reservoir knowledge management system based on knowledge mining technology provided by the embodiment of the present invention displays existing problems, research methods, research results, research effects, application examples and key maps based on keywords through minimalist application and intelligent presentation dominated by search engines, forming a standardized integrated display.

[0118] Therefore, the application management module is provided with an operation retrieval submodule, including a knowledge retrieval engine and a sorting processing unit. The knowledge retrieval engine uses a twin BERT network similarity evaluation model with an attention pooling method to perform document similarity analysis to realize knowledge retrieval services. The sorting processing unit is used to sort the retrieved knowledge objects according to a relevance sorting algorithm for intelligent knowledge display.

[0119] Search engines search for relevant literature using keywords, then sort the results using a relevance ranking algorithm. These rankings are then displayed to users based on relevance, helping them quickly find the most relevant research literature for the keywords they entered. Among literature search relevance ranking algorithms, the PageRank algorithm is the most widely used.

[0120] When the traditional PageRank algorithm is applied to citation network analysis, it uses an average weight distribution method as the current document PR value, that is, the relevance of the document to the cited document is considered to be the same, and the current PR value of the document cannot be reasonably distributed to the citation. Therefore, finding a more reasonable and reliable way to distribute the PR value of the document is the main goal of making the algorithm more suitable for citation network analysis. To address this problem, the present invention adopts a twin BERT network similarity evaluation model with an attention pooling method to estimate the similarity between the document and the citation, and then uses the similarity calculated by the model as the weight for assigning the PR value of the document.

[0121] In a preferred embodiment, the present invention designs the knowledge retrieval engine to introduce the idea of ​​attention into the twin BERT network similarity evaluation model, uses both whole sentence feature representation and single character feature representation to perform attention pooling, completes the extraction of important features, and realizes document similarity calculation.

[0122] The knowledge retrieval engine uses the twin BERT network similarity evaluation model with attention pooling method to calculate document similarity according to the following operations:

[0123] The similarity between the document and each cited document is calculated using two similarity evaluation functions: inner product similarity and cosine similarity.

[0124] Calculate the proportion of similarity of all citations pointed to by the document to the total;

[0125] The calculated proportion is used to replace the transfer matrix of the average distribution method, and the PageRank algorithm is iteratively executed until the set stability requirements are met. The calculation result is output as the document similarity result to realize knowledge document retrieval.

[0126] The twin BERT network similarity evaluation model of the embodiment of the present invention introduces the idea of ​​attention into the BERT network model, and uses two methods of sentence feature representation (CLS: Classification) and single character feature representation (token) to perform attention pooling to complete the extraction of important features; the principle diagram of the algorithm is shown in the figure Figure 5 As shown, in a preferred embodiment, the knowledge retrieval engine calculates the similarity between the document and each cited document according to the following operations:

[0127] Compare the overall features and individual character features of the BERT word embedding layer and the BertEncoder layer, calculate the attention weight, and then determine the attention pooling feature extraction formula based on it;

[0128] The output results of the attention pooling layer of the two twin sub-networks are obtained respectively, and the cosine similarity of the two output result vectors is calculated as the similarity between the document and the cited document.

[0129] Assume that the total number of citations contained in the document is N. After the document features pass through the BERT word embedding layer and the BertEncoder layer, the overall feature representation becomes X CLS =[x1, x2, ..., x i ,…,x N ], the representation of a single character feature in the literature becomes E token =[e1, e2, ..., e i ,…,e N ], where x i 、e i ∈R d, d is the dimension of word vector. CLS and E token The comparison is performed to calculate the attention weight. The higher the similarity between the overall feature representation and each local representation, the greater the attention weight assigned to the local representation. Each local representation refers to the representation of each individual character or word in the document after being processed by the BERT model.

[0130] The attention weight is calculated as follows:

[0131] α i =Softmax(sim(x i ·e i )), x i 、e i ∈R d , (1)

[0132] Where, α i is the attention weight, the function sim is used to measure the similarity between its two inputs, R d represents a set of vectors in a Euclidean space; R represents the set of real numbers, d represents the dimension of the vector, and R d represents the set of all d-dimensional real vectors;

[0133] In the embodiment of the present invention, the BERT network model is set to use two similarity evaluation functions, inner product similarity and cosine similarity, for calculation. The final attention pooling feature extraction formula is:

[0134]

[0135] Among them, S represents the output result of the attention pooling layer, x i represents the overall features of the ith citation document after the BERT word embedding layer and the BertEncoder layer, e i represents the single character feature of the ith citation document after the BERT word embedding layer and the BertEncoder layer; N represents the number of citation documents; R k Represents a set of k-dimensional real vectors.

[0136] For the document and citation features obtained using the model, the cosine similarity between the two is calculated, and the result is used as the similarity between the two documents. That is, the cosine similarity of the two vectors u and v output by the attention pooling layer of the two twin subnets is calculated as the similarity between the document and the citing document:

[0137]

[0138] Where S i Indicates the similarity between the document and the i-th cited document, u i,v i They represent the output results of the attention pooling layers of the two twin subnetworks when the i-th cited document participates in the calculation, and N represents the number of cited documents.

[0139] Optionally, in one embodiment, the knowledge search engine calculates the proportion of the similarity of all citations pointed to by the document to the total similarity by the following operation:

[0140] A random walk model is defined on a directed graph constructed based on documents and citations, and the proportion of the similarity of all citations pointed to by the documents is calculated using the basic transfer matrix of the random walk model and the completely random transfer matrix.

[0141] In practical application, let's construct a directed graph with n nodes based on the literature and references. We then define a general random walk model, or first-order Markov chain, on this directed graph. The transition matrix for this model consists of two components: the basic transition matrix M of the directed graph, which indicates that the probability of transitions from a node to all connected nodes is equal; and the completely random transition matrix, which indicates that the probability of transitions from any node to any other node is 1 / n.

[0142] In a preferred embodiment, the knowledge search engine calculates the overall proportion of similarities of all citations pointed to by the document according to the following formula:

[0143]

[0144] Among them, W i S is the proportion of the similarity of the i-th reference of the document to the total similarity of the references i is the similarity between the document and the i-th reference, and N is the number of references corresponding to the document.

[0145] Furthermore, we define the stationary distribution vector R as the general PageRank of the directed graph, and use the similarity ratio W calculated by formula (4) to replace the transfer matrix M originally calculated by the average distribution method in the algorithm. The calculation formula is as follows:

[0146]

[0147] Among them, 1 is an n-dimensional vector with all components equal to 1, t is the number of iterations of the algorithm, R t is the PR value after the tth iteration, d is the damping coefficient (0≤d≤1), and n is the number of nodes in the directed graph; the stability condition is set to satisfy the difference between the PR value obtained in the previous iteration and the PR value of this iteration is less than the preset threshold, the result is considered stable, and the calculation result is output at this time.

[0148] The document similarity ranking results obtained by the above method are more in line with the search needs of professionals. The ranking results are displayed to users according to relevance, helping them to quickly find the scientific research documents most relevant to the input keywords, which has better practicality.

[0149] By utilizing the above-mentioned settings of the system of the embodiment of the present invention, the potential knowledge mined through hot source analysis - project research plays a leading role in industry research, and the formation of literature retrieval and reports has achieved a development from manual to automatic, saving researchers' workload by more than 90%. The quality of research reports has been greatly improved, allowing researchers or managers to quickly understand the research status of the technology they own, so that the key technical knowledge of fracture-cavity oil reservoir development can be effectively managed.

[0150] The application management module includes a prediction analysis submodule, which is used to perform statistical analysis algorithms and modeling techniques based on existing knowledge data to predict the trend of knowledge volume changes in future demand time.

[0151] The application management module includes a related knowledge analysis submodule that analyzes the data in the knowledge base through associated text search technology, identifies related knowledge sets, helps users discover the correlation between knowledge, and provides responses and recommendations to user queries.

[0152] Managing knowledge about fracture-vuggy reservoirs also requires associating various types of knowledge. This paper uses associative text search technology to identify several articles with the highest similarity. The basic technical approach is as follows: First, all text in the text library is processed, and the content of each text is read and saved to a file. Each text is then segmented and a word vector is generated for each text. Assuming there are N1 texts, there will be N1 word vectors. Next, the cosine similarity between these N1 word vectors is calculated to generate a similarity matrix with dimension N1×N1.

[0153] Next, for a certain text, assuming its row index in the similarity matrix is ​​i, we traverse the i-th row of the matrix and find the column index corresponding to the second largest value. The text corresponding to this index value, i.e., text j, is the text with the highest similarity to text i (the largest value is the similarity between i and itself, which is 1). Correspondingly, the text k corresponding to the column index of the third largest value is the text with the second highest similarity to text i, and so on.

[0154] The efficient fracture-vuggy reservoir knowledge management system based on knowledge mining technology provided by the embodiments of the present invention can effectively implement core knowledge extraction, prediction, and association functions, solving the problems of cumbersome existing fracture-vuggy reservoir knowledge management methods, fragmented knowledge storage, and inconsistent knowledge display methods. Through knowledge engineering technology, reservoir description, geological modeling, and numerical simulation are integrated. Through search engine-driven minimalist applications and intelligent presentation, data mining technology is used to represent, store, manage, mine, and display implicit knowledge. This allows for the effective management of key technical knowledge for fracture-vuggy reservoir development and promotes the research and development of key technologies for fracture-vuggy reservoir development.

[0155] Figure 6 This is the overall functional architecture diagram of the system. The system includes five sub-modules: data import, data management, data analysis, key technology information display, and knowledge display. The data import and data management modules are responsible for implementing the functions of the data upload substructure and the integration and processing substructure. The data analysis module is responsible for implementing the functions of the analysis tool management substructure. The key technology information display and knowledge display modules are responsible for implementing the functions of the visualization interface substructure. Specific functions can be described as follows: the data import submodule includes the import and processing functions of data; the data management submodule includes the storage of data and the modification of information; the data analysis submodule includes the prediction of knowledge quantity, analysis of related knowledge, and label extraction functions; the key technology information display submodule includes the display of key technology information, sub-technology information, and the search function for key technology information; the knowledge display submodule includes the display of standards and specifications, the latest developments, knowledge resources, and training courseware.

[0156] The efficient fracture-cavity reservoir knowledge management system based on knowledge mining technology provided by the embodiment of the present invention achieves efficient management of fracture-cavity reservoir knowledge, discovers the knowledge hidden behind oilfield development, and effectively supports the development and research of fracture-cavity carbonate reservoirs. To this end, based on the characteristics of fracture-cavity reservoirs, such as multiple reservoir types, multiple fluid flows, and differentiated development methods, a key professional vocabulary database with fracture-cavity reservoir characteristics has been established, and the related knowledge derived from it is efficiently managed in many aspects. The system continuously improves the annotation of special tags and improves the existing PageRank data mining algorithm to mine the knowledge required by fracture-cavity reservoir researchers, achieving the integration of research methods, processes, and cases across different disciplines. Using knowledge push technology, hot business knowledge in the business is pushed to the corresponding push mode according to user needs, quickly providing researchers with a basic overview of existing technologies and artificial intelligence analysis results, promoting the research and development of key technologies for fracture-cavity reservoir development, and greatly reducing the workload of researchers. The actual application is expected to save researchers more than 90% of their workload.

[0157] In the efficient fracture-vuggy reservoir knowledge management system based on knowledge mining technology provided by the embodiment of the present invention, each module or unit structure can be operated independently or in combination according to actual knowledge management requirements and operation analysis requirements to achieve corresponding technical effects.

[0158] Example 2:

[0159] The above embodiments of the present invention describe the system in detail. Based on other aspects of the system described in any one or more of the above embodiments, the present invention also provides an efficient knowledge management method for fracture-vuggy reservoirs based on knowledge mining technology. This method is applicable to the efficient knowledge management system for fracture-vuggy reservoirs based on knowledge mining technology described in any one or more of the above embodiments. A specific embodiment is provided below for detailed description.

[0160] Specifically, the efficient fracture-vuggy reservoir knowledge management method based on knowledge mining technology provided in the embodiment of the present invention includes:

[0161] The knowledge acquisition module is used to obtain different types of fracture-cavity reservoir knowledge files, including basic data knowledge, business data knowledge and existing database objects, based on the multi-source knowledge approach required by fracture-cavity reservoirs.

[0162] The knowledge storage module provides users with knowledge storage, knowledge management and auxiliary services based on knowledge files. It uses a preset professional vocabulary to assist the system in file processing and analysis. The professional vocabulary includes various professional terms for setting development processes and supporting processes.

[0163] Use application auxiliary modules to provide authority management, file management, model management and internal data management support for knowledge management;

[0164] Provide key technology management, knowledge resource management, team management, training management, and status progress management services based on knowledge documents in response to user intent through the application management module;

[0165] Using a visualization display module to visualize the output results in response to user control based on an interactive interface;

[0166] The process of providing key technology management and knowledge resource management services through the application management module includes:

[0167] Analyzing the frequency of professional terms in the fracture-cavity reservoir knowledge file based on the professional term database, setting vocabulary tags for the file and annotating it;

[0168] Based on the classification, clustering and prediction processing of multiple labeled fracture-cavity reservoir knowledge objects to be processed, the knowledge files are converted into structured feature files.

[0169] Use the twin BERT network similarity evaluation model with attention pooling method to perform document similarity analysis to realize knowledge retrieval service; sort the retrieved knowledge objects according to the relevance ranking algorithm;

[0170] Based on the statistical analysis algorithms and modeling techniques of existing knowledge data, the trend of changes in knowledge volume in future demand time is predicted.

[0171] The efficient fracture-cavity reservoir knowledge management system used is built on an intelligent computer system and uses a multi-layer network architecture service, including a data acquisition layer, a data storage layer, an application support layer, and a core application layer:

[0172] The knowledge acquisition module is provided in the data acquisition layer, and acquires different types of fracture-cavity reservoir knowledge files according to the multi-source knowledge approach required by fracture-cavity reservoirs, including basic data knowledge, business data knowledge and existing database objects;

[0173] The knowledge storage module is provided in the data storage layer and is used to provide knowledge storage, knowledge management and auxiliary services to users based on knowledge files;

[0174] The application auxiliary module is set up in the application support layer to provide permission management, file management, model management and internal data management support for knowledge management;

[0175] An application management module is provided at the core application layer and is used to provide key technology management, knowledge resource management, team management, training management, and status progress management services based on knowledge documents in response to user intentions;

[0176] A visualization display module, which is used to visualize the output results in response to user control based on the interactive interface;

[0177] The knowledge storage module includes a professional vocabulary database, including multiple types of professional vocabulary for setting development processes and supporting processes, which is used to assist the system in implementing file processing and analysis;

[0178] The application management module includes a label annotation submodule, which is configured to analyze the frequency of professional terms in the fracture-cavity reservoir knowledge file based on the professional word library, set the vocabulary label of the file and annotate it;

[0179] The application management module includes a knowledge global processing submodule, which is configured to perform classification, clustering and prediction processing based on multiple fracture-cavity reservoir knowledge objects with labels to be processed, and convert the knowledge files into structured characteristic files;

[0180] The application management module is provided with an operation retrieval submodule, including a knowledge retrieval engine and a sorting processing unit. The knowledge retrieval engine uses a twin BERT network similarity evaluation model with an attention pooling method to perform document similarity analysis to implement knowledge retrieval services. The sorting processing unit is used to sort the retrieved knowledge objects according to a relevance sorting algorithm for intelligent knowledge display.

[0181] The application management module includes a prediction analysis submodule, which is used to perform statistical analysis algorithms and modeling techniques based on existing knowledge data to predict the trend of knowledge volume changes in future demand time.

[0182] Optionally, in one embodiment, the set development process includes seismic imaging and prediction, reservoir description, geological modeling, numerical simulation, water injection and gas injection processes.

[0183] Furthermore, in one embodiment, the text content is read, and if there are pictures, optical character recognition technology is used to identify the picture information, which is combined with the text content to form the file content, and then the file content is segmented according to a professional vocabulary. Based on this, the frequency of professional words is counted, and words with high frequency are selected as vocabulary tags for the text and annotated.

[0184] Preferably, in one embodiment, the vocabulary tag of the file is used as the file feature, and the file feature text is converted into a text vector form through the input layer of the BERT model. Then, the word vector, text vector, and position vector in the input layer are merged through the word embedding layer to obtain the word embedding representation of each word. Based on the word embedding representation, the features are further extracted through the encoding layer of the attention mechanism, and finally, the features are further extracted through the attention pooling layer to obtain the final document and citation features of the file, forming a structured feature file.

[0185] In an optional embodiment, the knowledge retrieval engine is used to introduce the attention concept into the twin BERT network similarity evaluation model, and attention pooling is performed using both whole sentence feature representation and single character feature representation to complete the extraction of important features and realize document similarity calculation.

[0186] In one embodiment, a knowledge retrieval engine is used to implement document similarity calculation using a twin BERT network similarity evaluation model with an attention pooling method according to the following operations:

[0187] The similarity between the document and each cited document is calculated using two similarity evaluation functions: inner product similarity and cosine similarity.

[0188] Calculate the proportion of similarity of all citations pointed to by the document to the total;

[0189] The calculated proportion is used to replace the transfer matrix of the average distribution method, and the PageRank algorithm is iteratively executed until the set stability requirements are met. The calculation result is output as the document similarity result to realize knowledge document retrieval.

[0190] Furthermore, in one embodiment, the knowledge search engine is used to calculate the similarity between the document and each cited document by the following operations:

[0191] Compare the overall features and individual character features of the BERT word embedding layer and the BertEncoder layer, calculate the attention weight, and then determine the attention pooling feature extraction formula based on it;

[0192] The output results of the attention pooling layer of the two twin sub-networks are obtained respectively, and the cosine similarity of the two output result vectors is calculated as the similarity between the document and the cited document.

[0193] Preferably, in one embodiment, the knowledge search engine is used to calculate the proportion of the similarity of all citations pointed to by the document to the total similarity by the following operation:

[0194] A random walk model is defined on a directed graph constructed based on documents and citations, and the proportion of the similarity of all citations pointed to by the documents is calculated using the basic transfer matrix of the random walk model and the completely random transfer matrix.

[0195] Specifically, in an optional embodiment, the following attention pooling feature extraction formula is determined:

[0196]

[0197] Among them, the attention weight α i =Softmax(sim(x i ·e i )), x i 、e i ∈R d , R d represents the set of all d-dimensional real vectors, d is the dimension of the word vector, S represents the output of the attention pooling layer, x i represents the overall features of the ith citation document after the BERT word embedding layer and the BertEncoder layer, e i represents a single character feature of the i-th citation document after the BERT word embedding layer and the BertEncoder layer; N represents the number of citation documents;

[0198] In one embodiment, the proportion of similarity of all citations pointed to by the document to the total is calculated as follows:

[0199]

[0200] Among them, W i S is the proportion of the similarity of the i-th reference of the document to the total similarity of the references i is the similarity between the document and the i-th reference, and N is the number of references corresponding to the document.

[0201] Preferably, in one embodiment, the PageRank algorithm is iteratively executed as follows:

[0202]

[0203] Among them, 1 is an n-dimensional vector with all components equal to 1, t is the number of iterations of the algorithm, R t is the PR value after the tth iteration, d is the damping coefficient, and n is the number of nodes in the directed graph. The stability condition is set to satisfy the difference between the PR value obtained in the previous iteration and the PR value of this iteration is less than the preset threshold. The result is considered stable and the calculation result is output at this time.

[0204] For simplicity of description, the aforementioned method embodiments are described as a series of actions. However, those skilled in the art should be aware that the present invention is not limited by the order of the actions described, as certain steps can be performed in other orders or simultaneously according to the present invention. Furthermore, those skilled in the art should also be aware that the embodiments described in this specification are preferred embodiments, and the actions and modules involved are not necessarily required for the present invention.

[0205] It should be pointed out that in other embodiments of the present invention, the method can also obtain a new fracture-cavity reservoir knowledge management method by combining one or several of the above embodiments to achieve comprehensive mining, analysis, processing and application of fracture-cavity reservoir knowledge.

[0206] It should be noted that, based on the method in any one or more of the above embodiments of the present invention, the present invention also provides a storage medium, which stores program code that can implement the method described in any one or more of the above embodiments. When the code is executed by the operating system, it can implement the efficient fracture-cavity oil reservoir knowledge management method based on knowledge mining technology as described above.

[0207] It should be understood that the embodiments disclosed herein are not limited to the specific structures, processing steps, or materials disclosed herein, but should extend to equivalent substitutions of these features understood by those skilled in the relevant art. It should also be understood that the terminology used herein is for the purpose of describing specific embodiments only and is not intended to be limiting.

[0208] The phrase "one embodiment" mentioned in the specification means that a particular feature, structure, or characteristic described in conjunction with the embodiment is included in at least one embodiment of the present invention. Therefore, the phrase "one embodiment" appearing in various places throughout the specification does not necessarily refer to the same embodiment.

[0209] Although the embodiments disclosed herein are as described above, the contents described herein are merely embodiments for facilitating understanding of the present invention and are not intended to limit the present invention. Any person skilled in the art may make any modifications and variations in the form and details of the embodiments without departing from the spirit and scope of the present invention. However, the scope of patent protection of the present invention shall remain subject to the scope defined by the appended claims.

Claims

1. An efficient knowledge management system for fracture-cavity oil reservoirs based on knowledge mining technology, characterized by: The system is built on an intelligent computer system and uses a multi-layer network architecture service, including a data acquisition layer, a data storage layer, an application support layer, and a core application layer: The knowledge acquisition module is provided in the data acquisition layer, and acquires different types of fracture-cavity reservoir knowledge files according to the multi-source knowledge approach required by fracture-cavity reservoirs, including basic data knowledge, business data knowledge and existing database objects; The knowledge storage module is provided in the data storage layer and is used to provide knowledge storage, knowledge management and auxiliary services to users based on knowledge files; The application auxiliary module is set up in the application support layer to provide permission management, file management, model management and internal data management support for knowledge management; An application management module is provided at the core application layer and is used to provide key technology management, knowledge resource management, team management, training management, and status progress management services based on knowledge documents in response to user intentions; A visualization display module, which is used to visualize the output results in response to user control based on the interactive interface; The knowledge storage module includes a professional vocabulary database, including multiple types of professional vocabulary for setting development processes and supporting processes, which is used to assist the system in implementing file processing and analysis; The application management module includes a label annotation submodule, which is configured to analyze the frequency of professional terms in the fracture-cavity reservoir knowledge file based on the professional word library, set the vocabulary label of the file and annotate it; The application management module includes a knowledge global processing submodule, which is configured to perform classification, clustering and prediction processing based on multiple fracture-cavity reservoir knowledge objects with labels to be processed, and convert the knowledge files into structured characteristic files; The application management module is provided with an operation retrieval submodule, including a knowledge retrieval engine and a sorting processing unit. The knowledge retrieval engine uses a twin BERT network similarity evaluation model with an attention pooling method to perform document similarity analysis to implement knowledge retrieval services. The sorting processing unit is used to sort the retrieved knowledge objects according to a relevance sorting algorithm for intelligent knowledge display. The application management module includes a prediction analysis submodule, which is used to perform statistical analysis algorithms and modeling techniques based on existing knowledge data to predict the trend of knowledge volume changes in future demand time.

2. The system according to claim 1, wherein: The set development process includes seismic imaging and prediction, reservoir description, geological modeling, numerical simulation, water injection and gas injection technology.

3. The system according to claim 1, wherein: The label annotation submodule is configured to read the text content, and if there are pictures, use optical character recognition technology to identify the picture information, combine it with the text content to form the file content, and then perform word segmentation on the file content according to the professional vocabulary. Based on this, the frequency of professional words is counted, and words with high frequency are selected as vocabulary tags for the text and annotated.

4. The system according to claim 1, wherein: The global knowledge processing submodule is configured to use the vocabulary tags of the file as file features, convert the file feature text into text vector form through the input layer of the BERT model, and then merge the word vector, text vector, and position vector in the input layer through the word embedding layer to obtain the word embedding representation of each word. Based on the word embedding representation, the encoding layer of the attention mechanism is used to further extract the features, and finally the attention pooling layer is used to further extract the features to obtain the final document and citation features of the file, forming a structured feature file.

5. The system according to claim 1, wherein: The knowledge retrieval engine introduces the idea of ​​attention into the twin BERT network similarity evaluation model, uses both whole sentence feature representation and single character feature representation to perform attention pooling, completes the extraction of important features, and realizes document similarity calculation.

6. The system according to claim 1 or 5, characterized in that The knowledge retrieval engine uses the twin BERT network similarity evaluation model with attention pooling method to calculate document similarity according to the following operations: The similarity between the document and each cited document is calculated using two similarity evaluation functions: inner product similarity and cosine similarity. Calculate the proportion of similarity of all citations pointed to by the document to the total; The calculated proportion is used to replace the transfer matrix of the average distribution method, and the PageRank algorithm is iteratively executed until the set stability requirements are met. The calculation result is output as the document similarity result to realize knowledge document retrieval.

7. The system according to claim 6, characterized in that The knowledge search engine calculates the similarity between the document and each cited document by the following operations: Compare the overall features and individual character features of the BERT word embedding layer and the BertEncoder layer, calculate the attention weight, and then determine the attention pooling feature extraction formula based on it; The output results of the attention pooling layer of the two twin sub-networks are obtained respectively, and the cosine similarity of the two output result vectors is calculated as the similarity between the document and the cited document.

8. The system according to claim 6, wherein: The knowledge search engine calculates the proportion of the similarity of all citations pointed to by the document to the total by the following operation: A random walk model is defined on a directed graph constructed based on documents and citations, and the proportion of the similarity of all citations pointed to by the documents is calculated using the basic transfer matrix of the random walk model and the completely random transfer matrix.

9. The system according to claim 6, wherein: The knowledge retrieval engine determines the following attention pooling feature extraction formula: Among them, the attention weight α i =Softmax(sim(x i ·e i )), x i 、e i ∈R d , R d represents the set of all d-dimensional real vectors, d is the dimension of the word vector, S represents the output of the attention pooling layer, x i represents the overall features of the ith citation document after the BERT word embedding layer and the BertEncoder layer, e i represents the single character feature of the ith citation document after the BERT word embedding layer and the BertEncoder layer; N represents the number of citation documents; R k Represents a set of k-dimensional real vectors.

10. The system according to claim 6, wherein: The knowledge search engine calculates the overall proportion of similarity of all citations pointed to by the document according to the following formula: Among them, W i S is the proportion of the similarity of the i-th reference of the document to the total similarity of the references i is the similarity between the document and the i-th reference, and N is the number of references corresponding to the document.

11. The system according to claim 6, wherein: The knowledge retrieval engine iteratively executes the PageRank algorithm as follows: Among them, 1 is an n-dimensional vector with all components equal to 1, t is the number of iterations of the algorithm, R t is the PR value after the tth iteration, d is the damping coefficient, and n is the number of nodes in the directed graph. The stability condition is set to satisfy the difference between the PR value obtained in the previous iteration and the PR value of this iteration is less than the preset threshold. The result is considered stable and the calculation result is output at this time.

12. An efficient knowledge management method for fracture-cavity reservoirs based on knowledge mining technology, characterized in that: The method is applied to the system according to any one of claims 1 to 11, and the method comprises: The knowledge acquisition module is used to obtain different types of fracture-cavity reservoir knowledge files, including basic data knowledge, business data knowledge and existing database objects, based on the multi-source knowledge approach required by fracture-cavity reservoirs. The knowledge storage module provides users with knowledge storage, knowledge management and auxiliary services based on knowledge files. It uses a preset professional vocabulary to assist the system in file processing and analysis. The professional vocabulary includes various professional terms for setting development processes and supporting processes. Use application auxiliary modules to provide authority management, file management, model management and internal data management support for knowledge management; Provide key technology management, knowledge resource management, team management, training management, and status progress management services based on knowledge documents in response to user intent through the application management module; Using a visualization display module to visualize the output results in response to user control based on an interactive interface; The process of providing key technology management and knowledge resource management services through the application management module includes: Analyzing the frequency of professional terms in the fracture-cavity reservoir knowledge file based on the professional term database, setting vocabulary tags for the file and annotating it; Based on the multiple labeled fracture-cavity reservoir knowledge objects to be processed, classification, clustering and prediction processing are performed to convert the knowledge files into structured characteristic files; Use the twin BERT network similarity evaluation model with attention pooling method to perform document similarity analysis to realize knowledge retrieval service; sort the retrieved knowledge objects according to the relevance ranking algorithm; Based on the statistical analysis algorithms and modeling techniques of existing knowledge data, the trend of changes in knowledge volume in future demand time is predicted.

Images