Geological knowledge map construction method based on multisource rock debris digitization and exploration system

Abstract

The invention belongs to the technical field of oil-gas exploration, and provides a geological knowledge map construction method based on multi-source rock debris digitization and an exploration system, and the method comprises the steps: carrying out the multi-source rock debris data collection of a target region, establishing a rock debris digital necessary parameter system containing spatial resolution, mineral volume percentage, pore complexity index and pollution mark information; according to the rock debris digital necessary parameter system, geological entity data related to the rock debris sample is extracted through a deposition detector and a diagenesis map; and according to the geological rule base and the geological entity data, reasoning the relationship between the key nodes to generate a geological knowledge graph. Closed-loop management of rock debris data from collection to intelligent application is realized by establishing a rock debris digital necessary parameter system, and geological entity data related to rock debris samples is extracted through a deposition detector and a diagenesis map, so that scanning data can be directly associated with geological significance, and data relevance is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of oil and gas exploration, and particularly relates to a multi-source rock debris digitalization-based geological knowledge graph construction method and an exploration system. BACKGROUND

[0002] With the deepening of oil and gas exploration and development, rock debris analysis as an important geological evaluation method plays a key role in oil and gas resource exploration. As a direct carrier of geological information obtained during drilling, the digital analysis of rock debris is of great significance for accurately evaluating the underground geological conditions and predicting the distribution of oil and gas reservoirs.

[0003] At present, the collection and analysis of rock debris data mainly rely on traditional manual observation and simple digital processing methods. These methods have obvious limitations when dealing with a large amount of multi-source heterogeneous rock debris data. With the application of knowledge graph technology in the geological field, geological evaluation methods based on knowledge graph have gradually attracted attention. However, there are many deficiencies in the construction of knowledge graph. First, the existing rock debris scanning parameter system is not unified, and the rock debris data parameter standards collected by different oil fields and different well areas are different, resulting in serious data island problem, which cannot be effectively compared and analyzed across wells, and restricts the formation of regional geological understanding. Second, there is a lack of effective association mechanism between the existing rock debris scanning data and the geological semantics, and the digital parameters obtained by scanning cannot be directly converted into interpretation results with geological significance, and there is a lack of effective association from data to geological knowledge.

[0004] Therefore, the traditional geological knowledge graph construction scheme has the technical problem of low accuracy caused by the non-uniform parameter system and insufficient data association. SUMMARY

[0005] The present application provides a multi-source rock debris digitalization-based geological knowledge graph construction method and an exploration system to solve the defect of low accuracy caused by the non-uniform parameter system and insufficient data association of the traditional geological knowledge graph construction scheme.

[0006] In one aspect, the present application provides a multi-source rock debris digitalization-based geological knowledge graph construction method, comprising: Collecting multi-source rock debris data in a target area, and establishing a rock debris digitalization mandatory parameter system containing spatial resolution, mineral volume percentage, pore complexity index and pollution marker information; According to the rock debris digitalization mandatory parameter system, extracting geological entity data related to rock debris samples through a pre-constructed sediment detector and diagenesis graph; According to a pre-constructed geological rule base and the geological entity data, reasoning the relationship between key nodes to generate a geological knowledge graph.

[0007] The method for constructing a geological knowledge graph based on multi-source lithic digitalization provided by the application comprises a deposition detector, which comprises: a deposition microfacies classifier, configured to output a deposition microfacies classification result according to a mineral volume percentage in the lithic digitalization optional parameter system; a deposition environment recognizer, configured to determine a mineral distribution and evolution sequence of the lithic in the target area according to the deposition microfacies classification result, and recognize a deposition environment of the lithic.

[0008] The method for constructing a geological knowledge graph based on multi-source lithic digitalization provided by the application comprises a diagenesis graph, which is obtained by the following process: obtaining porosity evolution data of the lithic sample; inputting the porosity evolution data into a pre-constructed graph network model to obtain a diagenesis graph output by the graph network model.

[0009] The method for constructing a geological knowledge graph based on multi-source lithic digitalization provided by the application further comprises: after a local lithology recognition model is trained by each oilfield data end in the target area according to local lithic data, receiving key model parameters of the local lithology recognition model uploaded by each oilfield data end in the target area; aggregating all the key model parameters by using a weighted average method to obtain global model parameters; downloading the global model parameters to each oilfield data end to update the local lithology recognition model.

[0010] The method for constructing a geological knowledge graph based on multi-source lithic digitalization provided by the application aggregates all the key model parameters by using a weighted average method to obtain global model parameters, which comprises: determining respective weight values corresponding to the key model parameters uploaded by different oilfield data ends; multiplying each key model parameter by a corresponding weight value to obtain a weighted model parameter; performing a mean value operation on each weighted model parameter to obtain global model parameters.

[0011] The method for constructing a geological knowledge graph based on multi-source lithic digitalization provided by the application determines respective weight values corresponding to the key model parameters uploaded by different oilfield data ends, which comprises: obtaining actual geological data of a target basin in which each oilfield data end is located; comparing the actual geological data with respective typical geological data of a plurality of typical basins to obtain a geological similarity between the target basin and each typical basin; According to the geological similarity between the target basin and each typical basin, weight values corresponding to key model parameters uploaded by different oilfield data terminals are obtained.

[0012] According to the present application, the method for constructing a geological knowledge graph based on multi-source rock digitalization further comprises: According to the updated local lithology identification model, the lithology category and attribute information are obtained according to the updated local lithology identification model; According to the lithology category and attribute information, the geological knowledge graph is updated.

[0013] According to the present application, the method for constructing a geological knowledge graph based on multi-source rock digitalization further comprises: The multi-source rock data in the target area is collected; It is judged whether all the required parameters are included in the multi-source rock data, and a judgment result is obtained; If the judgment result is no, the multi-source rock data in the target area is collected again until the judgment result is yes, and then a rock digitalization required parameter system containing spatial resolution, mineral volume percentage, pore complexity index and pollution marker information is established according to the currently collected multi-source rock data.

[0014] In another aspect, the present application further provides an exploration system based on multi-source rock digitalization, comprising: A distributed rock scanning terminal is used to collect multi-source rock data in a target area and establish a rock digitalization required parameter system containing spatial resolution, mineral volume percentage, pore complexity index and pollution marker information. A geological knowledge graph visualization platform is used to extract geological entity data related to rock samples according to the rock digitalization required parameter system and a pre-constructed sedimentary information identifier and diagenesis graph, and infer the relationship between key nodes according to a pre-constructed geological rule base and the geological entity data to generate a geological knowledge graph.

[0015] According to the present application, the exploration system based on multi-source rock digitalization further comprises an aggregation server. The polymeric server is used for receiving key model parameters of the local lithology identification model uploaded by each oilfield data end in a target area after the local lithology identification model is trained by the local rock data of each oilfield data end in the target area; all the key model parameters are aggregated by using a weighted average method to obtain global model parameters; and the global model parameters are distributed to each oilfield data end to update the local lithology identification model.

[0016] The application provides a multi-source rock digitalization-based geological knowledge graph construction method and an exploration system. The multi-source rock digitalization-based geological knowledge graph construction method and the exploration system are provided, multi-source rock data of a target area is collected, a rock digitalization essential parameter system containing spatial resolution, mineral volume percentage, pore complexity index and pollution marker information is established, geological entity data related to rock samples is extracted through a pre-constructed sediment detector and a diagenesis graph according to the rock digitalization essential parameter system, and the relationship between key nodes is inferred according to a pre-constructed geological rule base and the geological entity data to generate a geological knowledge graph. The multi-source rock digitalization essential parameter system containing spatial resolution, mineral volume percentage, pore complexity index and pollution marker information is established, closed-loop management of rock data from collection to intelligent application is realized, the data island problem is solved, rock data of different wells can be standardized and compared, the pre-constructed sediment detector and the diagenesis graph are used to extract geological entity data related to rock samples, the scanning data can be directly associated with geological significance, the data correlation is improved, and the accuracy of the geological knowledge graph construction link is improved. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present application or prior art, the following will briefly introduce the drawings needed in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0018] Figure 1 Fig. 1 is a flowchart of the multi-source rock digitalization-based geological knowledge graph construction method provided by the embodiments of the present application; Figure 2 Fig. 2 is a schematic diagram of the distribution state of key nodes in the embodiments of the present application; Figure 3 Fig. 3 is a structural schematic diagram of the multi-source rock digitalization-based exploration system provided by the embodiments of the present application. DETAILED DESCRIPTION

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0020] The following is combined Figures 1 to 3 This invention describes in detail the geological knowledge map construction method and exploration system based on multi-source rock cuttings digitization provided in the embodiments of the present invention.

[0021] like Figure 1 As shown in the figure, the geological knowledge map construction method based on multi-source rock cuttings digitization provided by the embodiments of the present invention mainly includes the following steps: Step 110: Collect multi-source rock cuttings data from the target area and establish a mandatory parameter system for digitizing rock cuttings, including spatial resolution, mineral volume percentage, pore complexity index, and contamination labeling information.

[0022] In the essential parameter system for digitizing rock cuttings, spatial resolution is usually measured in micrometers to characterize the fineness of rock cutting samples. In practical applications, the spatial resolution of CT scans needs to reach below 5 micrometers to clearly capture the microstructure inside the rock cuttings.

[0023] Mineral volume percentage is used to quantify the content distribution of various minerals in rock cuttings. In practical applications, it is necessary to accurately identify and obtain the volume percentage of minerals such as quartz, clay, and carbonate minerals. These minerals are common in oil and gas reservoirs and have an important impact on reservoir properties. Accurately determining their content helps in judging reservoir type and quality.

[0024] The pore complexity index reflects the complexity of the pore structure in rock cuttings. In practical applications, the pore complexity index can be calculated using the CT three-dimensional fractal dimension.

[0025] Contamination labeling information is used to identify whether rock cuttings samples are contaminated by mud, ensuring data accuracy. In practical applications, contamination can be indicated by detecting mud additive elements, such as calculating the Ba / Sr ratio. Mud contamination can interfere with the acquisition of accurate geological information from rock cuttings samples; accurate labeling of contamination helps in subsequent data correction and analysis.

[0026] Step 120: Based on the essential parameter system for digitizing rock fragments, extract geological entity data related to the rock fragment samples using a pre-constructed sedimentary detector and diagenetic map.

[0027] Through the analysis of the deposition detector and the diagenesis atlas, geological entity data related to the cutting sample can be extracted, which includes lithology information, mineral composition, pore characteristics, sedimentary environment characteristics, and diagenetic types, and the construction of the cutting-geological entity mapping relationship can be realized.

[0028] Step 130: According to the pre-constructed geological rule base and the geological entity data, the relationship between the key nodes is inferred, and the geological knowledge atlas is generated.

[0029] In this embodiment, the geological knowledge atlas can not only clearly show the correlation between various geological elements, but also provide reservoir quality evaluation and engineering suggestions, thereby providing decision support for oil and gas exploration and development.

[0030] In an embodiment, multi-source cutting data of a target area is collected, and a cutting digitization mandatory parameter system including spatial resolution, mineral volume percentage, pore complexity index, and pollution marker information is established, which specifically includes: First, multi-source cutting data in the target area is collected.

[0031] In practical applications, multi-source cutting data can come from cutting samples, core analysis data, logging data, and other sources during drilling.

[0032] Then, it is determined whether all mandatory parameters are included in the collected multi-source cutting data, and a determination result is obtained.

[0033] It can be understood that the mandatory parameters involve spatial resolution, mineral volume percentage, pore complexity index, and pollution marker information, that is, the necessary data in the cutting digitization mandatory parameter system.

[0034] Finally, if the determination result is no, the multi-source cutting data in the target area needs to be collected again until the determination result is yes, and at this time, the cutting digitization mandatory parameter system including the above four mandatory parameters is established according to the currently collected multi-source cutting data.

[0035] In this embodiment, the data integrity is checked in real time during the parameter scanning process, and if the mandatory parameters are missing, the resampling mechanism is triggered immediately to ensure that the collected data meet the requirements of subsequent analysis, thereby ensuring the data quality.

[0036] In an embodiment, the deposition detector includes a sedimentary microfacies classifier and a sedimentary environment identifier.

[0037] The sedimentary microfacies classifier is used to output a sedimentary microfacies classification result according to the mineral volume percentage in the cutting digitization mandatory parameter system.

[0038] In this embodiment, the sedimentary microfacies classifier can adopt a data classifier based on a convolutional neural network, which can classify the sedimentary microfacies according to the mineral volume percentage of the cutting sample. Since different sedimentary microfacies have specific mineral combination characteristics, the data classifier based on the convolutional neural network can accurately identify the sedimentary microfacies to which the cutting sample belongs by training a large amount of labeled data.

[0039] It can be understood that the sedimentary microfacies classification result can include different types of sedimentary microfacies such as sandstone, mudstone, and carbonate rock.

[0040] The sedimentary environment identifier is used to determine the mineral distribution and evolution sequence of the cutting in the target area according to the sedimentary microfacies classification result, and to identify the sedimentary environment of the cutting, such as different sedimentary environment types such as shallow sea, deep sea, lacustrine facies, and fluvial facies.

[0041] In this embodiment, the sedimentary environment identifier can be implemented based on a recurrent neural network with an attention mechanism. According to the sedimentary microfacies classification result and the mineral distribution and evolution sequence of the cutting sample, the dynamic identification of the sedimentary environment can be realized. The recurrent neural network can more accurately capture the evolution characteristics of the sedimentary environment by learning the change rule of the mineral combination in the historical sedimentary sequence, thereby providing reliable data support for the construction of the subsequent geological knowledge graph.

[0042] In some embodiments, natural language processing technology can also be used to extract key geological information such as mineral distribution and sedimentary structure from the relevant descriptive text of the cutting sample as supplementary data to improve the node content of the geological entity. At the same time, reinforcement learning algorithm can be used to optimize the sedimentary microfacies and sedimentary environment identification process, and the related model parameters can be adjusted through continuous trial and error and reward mechanism to further improve the accuracy and stability of the geological entity extraction.

[0043] In other embodiments, a conditional random field model can also be used to accurately model and feature extract the pore network based on the mineralogical and structural parameter information of the cutting sample, providing an accurate data basis for subsequent diagenesis analysis. At the same time, the complex relationships between geological entities such as sedimentary microfacies and sedimentary environment can be modeled by combining the Transformer architecture in deep learning, and the potential geological evolution law can be mined, so that the entity content of the geological knowledge graph can be further enriched.

[0044] In an embodiment, the diagenesis graph can be obtained by the following process: First, the porosity evolution data of the cutting sample is obtained.

[0045] It can be understood that the porosity evolution data reflects the change of porosity of the rock in the geological history.

[0046] Then, the porosity evolution data is input into a pre-constructed graph network model which can analyze the relationship between porosity evolution and diagenesis, and finally output a diagenesis graph.

[0047] Since the information involved in the porosity evolution process such as pore structure change, mineral dissolution and precipitation can be effectively modeled by graph neural networks, the process and mechanism of diagenesis are revealed.

[0048] In this embodiment, the diagenesis graph shows the diagenetic process that the rocks in the target area have undergone, including compaction, cementation, dissolution, etc.

[0049] Through the cooperative use of the above-mentioned multiple algorithms, geological entity information in the digitalized core parameter system can be extracted from multiple dimensions and in-depth, providing rich and accurate node content for the knowledge graph, and further laying a solid foundation for subsequent relationship reasoning.

[0050] In this embodiment, a pre-defined geological rule library can be established, which contains a large amount of geological knowledge and experience rules, such as the correspondence between different mineral combinations and sedimentary environments, the influence law of diagenesis on pore network, and the reservoir quality evaluation standard. For example, it can include geological rules such as "high quartz + dissolution pore → high-quality sandstone reservoir" and "clay > 30% + microfracture → fracturing potential layer". Through these geological rules, combined with the extracted geological entity data, the relationship between the key nodes such as reservoir quality and engineering recommendations can be further reasoned, so as to build a complete geological knowledge graph.

[0051] In one specific implementation, as shown in Figure 2 the key nodes specifically include core samples, mineral combinations, pore networks, sedimentary environments, diagenesis, reservoir quality, and engineering recommendations.

[0052] Among them, the core sample is the starting node of the geological knowledge graph, which contains various core sample information in the digitalized core parameter system. The mineral combination is determined by the mineral identification parameters in the core sample, and different mineral combinations reflect different sedimentary environments and geological evolution history. The pore network is constructed based on the pore structure information obtained by CT scanning, and the characteristics of the pore network are closely related to diagenesis. The sedimentary environment is inferred from the mineral combination and other information, and the sedimentary environment has a decisive influence on the formation and distribution of the reservoir. Diagenesis is analyzed from information such as pore network, which changes the physical and chemical properties of rocks and further affects the reservoir quality. Reservoir quality can be evaluated comprehensively from sedimentary environment, diagenesis and other factors, and is one of the core indicators of oil and gas exploration. Engineering recommendations can be generated according to reservoir quality and other information, which can provide guidance for actual drilling and completion engineering.

[0053] In an embodiment, the method for constructing a multi-source rock debris digitalized geological knowledge graph described above can further include the following steps: First, after each oilfield data end in the target region trains a local lithology identification model according to local rock debris data, the key model parameters of the local lithology identification model uploaded by each oilfield data end in the target region are received.

[0054] In actual application, the target region can include multiple oilfields, and each oilfield has its own rock debris data characteristics. Therefore, each oilfield data end first trains a local lithology identification model suitable for local geological conditions using local rock debris data. After the training is completed, each oilfield data end uploads the key model parameters of the local lithology identification model to the aggregation server. These key model parameters usually include the weight matrix, bias vector, activation function, and other parameters of the model.

[0055] By uploading only the key model parameters, the original data privacy of each oilfield data end is protected, and the model performance is improved by using multi-source data.

[0056] Then, the weighted average method is used to aggregate all the key model parameters to obtain global model parameters.

[0057] In a specific implementation, the weighted average method is used to aggregate all the key model parameters to obtain global model parameters, including: First, determine the respective weight values of the key model parameters uploaded by different oilfield data ends.

[0058] Further, the process of determining the respective weight values of the key model parameters uploaded by different oilfield data ends includes: First, obtain the actual geological data of the target basin where each oilfield data end is located.

[0059] Then, compare the actual geological data with the respective typical geological data of multiple typical basins to obtain the geological similarity between the target basin and each typical basin.

[0060] Finally, according to the geological similarity between the target basin and each typical basin, the respective weight values of the key model parameters uploaded by different oilfield data ends are obtained.

[0061] In actual application, the calculation of geological similarity can be based on multiple geological features, such as structural features, sedimentary environment, lithology distribution, etc. Specifically, the similarity of geological features of different basins can be quantitatively analyzed by using cosine similarity, Euclidean distance, and other measurement methods to obtain the geological similarity.

[0062] For example, the carbonate reservoirs are mainly in the Middle East region, and the weight value of the carbonate model parameters can be set to 0.8; and the weight value of the shale gas basin model parameters can be set to 0.2. In this way, the aggregated model can better adapt to the geological characteristics of different regions, thereby improving the prediction accuracy of the model in each region.

[0063] Secondly, multiply each key model parameter by the corresponding weight value to obtain a weighted model parameter.

[0064] Thirdly, perform a mean value operation on each weighted model parameter to obtain a global model parameter.

[0065] Before and after the mean value operation, the weighted model parameters can be standardized and de-standardized to eliminate the dimension influence before the mean value operation and restore the actual physical meaning of the global model parameter after the mean value operation.

[0066] Through the weighted average method, the differences in different oilfield geological conditions can be fully considered, so that the global model is more representative and applicable.

[0067] Finally, the global model parameters are distributed to each oilfield data terminal to update the local lithology identification model.

[0068] After the global model parameters are distributed to each oilfield data terminal, the local lithology identification model can be updated using the global model parameters. This federated learning method not only protects the data privacy of each oilfield, but also improves the overall performance of the model. The updated local lithology identification model can more accurately identify the local lithology category and improve the accuracy of geological interpretation.

[0069] In an embodiment, the method for constructing a geological knowledge graph based on multi-source digitalized cuttings can further include: First, the updated local lithology identification model is used to obtain lithology categories and attribute information according to the digitalized cutting mandatory parameter system.

[0070] The updated local lithology identification model can analyze the cutting samples based on each parameter in the digitalized cutting mandatory parameter system and output more accurate lithology categories and attribute information. The lithology categories include different types such as sandstone, mudstone, and carbonate rock, and the attribute information includes characteristic parameters such as mineral composition, porosity, and permeability.

[0071] Then, the geological knowledge graph is updated according to the lithology categories and attribute information.

[0072] In this embodiment, the obtained lithology category and attribute information are integrated into the constructed geological knowledge graph to update and improve the constructed geological knowledge graph. The updated geological knowledge graph can more accurately reflect the geological characteristics and reservoir properties of the target region, providing a more reliable decision basis for oil and gas exploration and development.

[0073] Through the above steps, the embodiment realizes the construction of a geological knowledge graph based on multi-source digitalized cuttings, which can intuitively display the correlation between geological elements and provide scientific decision support for oil and gas exploration and development. At the same time, through federated learning, the collaborative use of multi-oilfield data is realized, which protects data privacy while improving model performance and the accuracy of geological interpretation.

[0074] Taking the cuttings in a well in the Sichuan Basin as an example, CT3 micron scanning is performed on the cuttings in the well, and it is detected that the clay mineral content is 32%. Through CT three-dimensional fractal dimension calculation, the pore complexity index is 2.17. According to the above geological knowledge graph construction scheme provided by the embodiment, the geological knowledge graph is constructed, and the following information can be output: the mineral combination analysis shows that the mineral combination characteristics of the cuttings point to a deep water shelf facies sedimentary environment. Based on the analysis of the pore network characteristics, it is judged that it has experienced strong compaction diagenesis. Integrating the sedimentary environment and diagenesis, the reservoir quality of the region is evaluated as grade III. For grade III reservoir quality, the engineering suggestion is given: close-cut volume fracturing to improve the efficiency of shale gas exploitation.

[0075] After adding the above related data into the federated learning system, the lithology recognition accuracy of the Guizhou shale gas well is effectively improved. This shows that through federated learning, using data from different regions for model training and optimization can effectively improve the generalization ability of the model, so that it can play a better role in shale gas exploration in different regions.

[0076] As can be seen from this embodiment, the geological knowledge graph construction method based on multi-source digitalized cuttings has strong data processing and analysis capabilities in shale gas block applications. Not only can it accurately analyze the geological information contained in the cuttings to provide scientific decision basis for exploration and development, but also effectively enhances the generalization performance of the model through federated learning.

[0077] Based on the same overall inventive concept, the present application also protects an exploration system based on multi-source digitalized cuttings. The exploration system based on multi-source digitalized cuttings provided by the present application is described below, and the exploration system based on multi-source digitalized cuttings described below can be mutually corresponding and referred to the geological knowledge graph construction method based on multi-source digitalized cuttings described above.

[0078] As Figure 3As shown, the multi-source rock debris digitalization-based exploration system provided by the embodiment of the present application specifically comprises: a distributed rock debris scanning terminal 210 and a geological knowledge graph visualization platform 220.

[0079] The distributed rock debris scanning terminal 210 is configured to collect multi-source rock debris data of a target area, and establish a rock debris digitalization optional parameter system containing spatial resolution, mineral volume percentage, pore complexity index and pollution marking information.

[0080] In the embodiment, the distributed rock debris scanning terminal is equipped with a high-precision optical scanner and an X-ray diffraction analysis device, and can collect physical property and mineral composition data of rock debris samples in real time. The spatial resolution is usually set to 5-50 microns, which can be adjusted according to the characteristics of different rock debris samples; the pore complexity index adopts a quantitative standard of 1-10, wherein 1 represents a simple pore structure and 10 represents an extremely complex pore network; and the pollution marking information is realized through a specific fluorescent marking technology, which can distinguish drilling fluid pollution and original fluid characteristics.

[0081] The geological knowledge graph visualization platform 220 is configured to extract geological entity data related to rock debris samples according to the rock debris digitalization optional parameter system, through a pre-constructed sediment detector and diagenesis graph, and infer the relationship between key nodes according to a pre-constructed geological rule base and the geological entity data, to generate a geological knowledge graph.

[0082] In the embodiment, the geological knowledge graph visualization platform uses a deep learning algorithm to analyze the rock debris digitalization optional parameter system, which can automatically identify key mineral combinations and microstructure characteristics.

[0083] In an embodiment, as shown in Figure 2 The multi-source rock debris digitalization-based exploration system can further comprise: An aggregation server 230 is configured to, after each oilfield data end in a target area trains a local lithology recognition model according to local rock debris data, receive key model parameters of the local lithology recognition model uploaded by each oilfield data end in the target area, aggregate all the key model parameters in a weighted average manner to obtain global model parameters, and distribute the global model parameters to each oilfield data end to update the local lithology recognition model.

[0084] In the embodiment, the aggregation server can use a high-performance computing cluster architecture, which has a CPU with 16 cores or more and at least 64 GB of memory, and can ensure efficient processing of large-scale model parameters from multiple oilfields.

[0085] The process of updating the local lithology identification model can adopt an incremental learning manner, retains the adaptability of the local model to the characteristics of a specific region, and fuses global knowledge to improve the overall identification accuracy.

[0086] In the embodiment, the system can protect the data privacy of each oil field, realize knowledge sharing and model optimization, and significantly improve the accuracy and application range of lithology identification by combining distributed learning with central aggregation.

[0087] As to the system in the above embodiments, the specific manner in which each part performs the operation has been described in detail in the embodiments of the method, and will not be described in detail here.

[0088] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for constructing a geological knowledge graph based on multi-source digitalized cuttings, characterized in that, The method comprises the following steps: Collecting multi-source rock debris data of a target area, and establishing a rock debris digitization essential parameter system comprising spatial resolution, mineral volume percentage, pore complexity index and pollution marker information; Extracting geological entity data related to the rock debris sample according to the rock debris digitization essential parameter system and through a pre-constructed sediment detector and diagenesis atlas; Inferencing the relationship between key nodes according to a pre-constructed geological rule base and the geological entity data, and generating a geological knowledge graph. 2.The method of claim 1, wherein, The sediment detector comprises: A sedimentary microfacies classifier configured to output a sedimentary microfacies classification result according to the mineral volume percentage in the rock debris digitization essential parameter system; A sedimentary environment identifier configured to determine the mineral distribution and evolution sequence of the rock debris in the target area according to the sedimentary microfacies classification result, and identify the sedimentary environment of the rock debris. 3.The method of claim 1, wherein, The diagenesis atlas is obtained through the following process: Obtaining porosity evolution data of the rock debris sample; Inputting the porosity evolution data into a pre-constructed graph network model to obtain the diagenesis atlas output by the graph network model. 4.The method of claim 1, wherein, The method further comprises: After each oilfield data terminal in the target area trains a local lithology recognition model based on local rock debris data, receiving key model parameters of the local lithology recognition model uploaded by each oilfield data terminal in the target area; Aggregating all the key model parameters by using a weighted average method to obtain global model parameters; Downloading the global model parameters to each oilfield data terminal to update the local lithology recognition model. 5.The method of claim 4, wherein, The aggregation of all the key model parameters by using the weighted average method to obtain the global model parameters comprises: Determining the respective weight values of the key model parameters uploaded by different oilfield data terminals; Multiplying each key model parameter by the corresponding weight value to obtain a weighted model parameter; Performing a mean value operation on each weighted model parameter to obtain the global model parameter. 6.The method of claim 5, wherein, The determination of the respective weight values of the key model parameters uploaded by different oilfield data terminals comprises: Obtaining actual geological data of the target basin where each oilfield data terminal is located; Comparing the actual geological data with respective typical geological data of multiple typical basins to obtain the geological similarity between the target basin and each typical basin; According to the geological similarity between the target basin and each typical basin, the respective weight values of the key model parameters uploaded by different oilfield data terminals are obtained. 7.The method of claim 4, wherein, The method further comprises: Obtaining lithology categories and attribute information according to the rock debris digitization essential parameter system and through the updated local lithology recognition model; Updating the geological knowledge graph according to the lithology categories and attribute information. 8.The method of claim 1, wherein, Collecting multi-source rock debris data of a target area, and establishing a rock debris digitization essential parameter system comprising spatial resolution, mineral volume percentage, pore complexity index and pollution marker information, comprises: Collecting multi-source rock debris data in the target area; Determining whether all essential parameters are included in the multi-source rock debris data to obtain a determination result; If the judgment result is no, multi-source cutting data in the target area is collected again until the judgment result is yes, and then a cutting digitization essential parameter system including spatial resolution, mineral volume percentage, pore complexity index and pollution marker information is established according to the current collected multi-source cutting data.

9. A prospecting system based on multi-source digitalization of cuttings, characterized in that, The system comprises: a distributed cutting scanning terminal, which is used for collecting multi-source cutting data of a target area and establishing a cutting digitization essential parameter system including spatial resolution, mineral volume percentage, pore complexity index and pollution marker information; a geological knowledge graph visualization platform, which is used for extracting geological entity data related to cutting samples according to the cutting digitization essential parameter system through a pre-constructed sedimentary information identifier and diagenesis graph, and reasoning the relationship between key nodes according to a pre-constructed geological rule base and the geological entity data to generate a geological knowledge graph.

10. The multi-source cuttings digitization based exploration system of claim 9, wherein, The system further comprises an aggregation server. The aggregation server is used for receiving key model parameters of the local lithology recognition model uploaded by each oilfield data terminal in the target area after the local lithology recognition model is trained by local cutting data, aggregating all the key model parameters in a weighted average manner to obtain global model parameters, and downloading the global model parameters to each oilfield data terminal to update the local lithology recognition model.

Images