Intelligent geophysical logging full-well-section lithofacies identification method
By constructing a neural network model to clean and reduce the dimensions of historical lithophagometry data, optimize parameters, and solve the difficulty of identifying traditional lithophagometry at high data order, achieve fast and accurate lithophagometry recognition, and improve geophysics Feasibility of intelligent identification of lithophagos in the entire well section of logging.
Patent Information
- Application Number
- CN202411172057.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-26
- Publication Date
- 2024-12-13
AI Technical Summary
Traditional lithologic identification methods are difficult to adapt to the rapid growth of data dimensions and orders of magnitude in the oil and gas industry, resulting in difficulty in precise identification of complex lithologic properties in fine strata description.
Construct a neural network logging lithology recognition model, and optimize the model parameters by cleaning historical lithophytic data, labeling dimensionality reduction and parameter dimensionality reduction processing, thereby improving the accuracy of the model's drilling, and importing the lithophytic data of the well section to be studied The optimization model is used to identify it, and the lithology with the maximum probability is selected as the actual lithology.
It realizes the rapid identification of lithophytic information of the well section, avoids the difficulty of identifying traditional methods under high data dimensions and orders of magnitude, and improves the feasibility and accuracy of intelligent identification of lithophytic phases of the whole well section of the geophysical logging.
Smart Images
Figure CN119128380A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of geophysical technology, and in particular to a method for intelligently identifying lithofacies in the entire well section of geophysical logging. Background Art
[0002] Geophysical logging is a method of measuring the physical parameters of underground rock formations and the technical conditions of wells using various instruments. It is mainly used for the exploration and exploitation of oil, coal and metal ore bodies. This method analyzes the characteristics of the formation, such as resistivity, acoustic wave velocity, rock density, etc., by measuring the electrical, acoustic, and radioactive properties of the rock formation, so as to determine the important parameters of the reservoir, such as porosity, oil saturation, and permeability. Geophysical logging not only helps to divide oil, gas, and water layers, but also studies the dynamics of oil, gas, and water and the conditions of the wells during the development of oil and gas fields, providing a basis for the formulation of development plans. The lithofacies of the entire well section refers to the comprehensive description and analysis of the distribution and characteristics of different lithofacies in the entire well section. Lithofacies refers to rocks or rock combinations formed in a sedimentary environment, which reflect the paleoenvironmental conditions at the time of deposition, such as paleoclimate, paleosource, paleosalinity, and paleowater depth. Through the lithofacies analysis of the entire well section, we can understand the evolution of the sedimentary environment and then infer geological events and sedimentary effects during the geological history.
[0003] Detailed description of formations is a key task in oil and gas exploration and development, and accurate identification of complex lithology is one of the most important links in stratigraphic interpretation. With the deepening of digitalization in the oil and gas industry, traditional lithology identification methods have become difficult to adapt to the increasing dimensions and magnitude of data.
[0004] Therefore, it does not meet the existing needs. We propose a method for intelligent identification of lithofacies in the entire well section of geophysical logging. Summary of the invention
[0005] The purpose of the present invention is to provide a method for intelligent identification of lithofacies in the entire well section of geophysical logging. By constructing a neural network logging lithology identification model, historical lithofacies data are obtained for cleaning and label and parameter dimensionality reduction, thereby improving the effectiveness of the lithofacies sample set, and the lithofacies sample set is used to establish and optimize the parameters of its model, thereby improving the accuracy of model training; after the lithofacies data of the well section to be studied are cleaned, they are imported into the optimized model for training, and the probabilities of various lithologies in the well section to be studied are output, and the lithology with the maximum probability is selected as the actual lithology, thereby quickly identifying the lithofacies information of the current well section to be studied, thereby avoiding the difficulty of traditional lithology identification methods in adapting to the increasing data dimensions and magnitudes, improving the feasibility of existing logging section lithofacies identification methods, and solving the problems raised in the above-mentioned background technology.
[0006] To achieve the above object, the present invention provides the following technical solutions:
[0007] A method for intelligently identifying lithofacies in the entire well section of geophysical logging comprises the following steps:
[0008] S1. Obtain data: collect historical lithofacies related data and collect lithofacies data of the current well section to be studied;
[0009] S2. Data processing: cleaning and preprocessing the historical lithofacies data, and performing label dimension reduction and parameter dimension reduction on the processed historical lithofacies data, and using the processed historical lithofacies data as a lithofacies sample set;
[0010] S3. Establish model parameters: construct a neural network logging lithology recognition model, import the processed historical lithofacies data into the neural network logging lithology recognition model for training and testing, and output the training results and test results in sequence, check whether the probabilities of various lithologies in the two results are consistent, and establish and optimize the lithology parameters of the neural network logging lithology recognition model based on the verification results;
[0011] S4, Analyze lithology and sedimentation: After pre-processing the lithofacies data of the current well section to be studied according to the processing step S2, import the optimized neural network logging lithology recognition model according to the step S3 to perform parameter recognition, and output the probability of various lithologies of the current well section to be studied;
[0012] S5. Establishing lithology information: Based on the parameter training results output in S4, the lithology with the maximum probability in the results is selected as the lithology information of the current well section to be studied.
[0013] Furthermore, the historical lithofacies data and current lithofacies data of the well section to be studied are collected in S1, which specifically includes the following steps:
[0014] By checking the previous logging related literature, we collected multiple sets of historical logging lithofacies data, and selected the logging lithofacies data with relatively accurate data as sample data to establish and optimize the lithology parameters of the neural network logging lithology identification model;
[0015] The mechanical notching method is used to collect geological and lithofacies data of the well section to be studied. The specific geological and lithofacies data include: rock layer, lithofacies, sedimentary structure information, logging curve and logging depth.
[0016] Furthermore, the historical lithofacies data is cleaned and preprocessed in S2, which specifically includes the following steps:
[0017] Abnormal and erroneous data in the logging lithofacies data to be processed are eliminated, missing data are supplemented, and the pre-processed data are standardized to ensure the accuracy and consistency of the lithofacies data. Then, labels are set for the lithofacies data according to their attributes, and each lithofacies data is classified with the corresponding logging curve and logging depth.
[0018] Furthermore, the label dimensionality reduction and parameter dimensionality reduction processing are performed on the historical lithofacies data in S2, which specifically includes the following steps:
[0019] Based on the label information of the pre-processed lithofacies data, similar lithologies are merged to reduce the label ratio of similar and repeated lithologies, and the lithology labels and lithology data with a ratio lower than the preset ratio are deleted;
[0020] Based on the parameter categories of the preprocessed lithofacies data, two or more parameter categories with parameters higher than the preset correlation are selected for dimensionality reduction, and parameter categories that are unrelated to each other are retained.
[0021] Furthermore, the labels obtained after the label dimensionality reduction process include seven lithologies: mudstone, sandstone, shale, dolomite, conglomerate, marl and gypsum; the parameters obtained after the parameter dimensionality reduction process include five input parameters: compensated acoustic wave, compensated electron, compensated density, natural gamma and resistivity.
[0022] Furthermore, in S3, the processed historical lithofacies data is imported into the neural network logging lithology recognition model for training and testing, which specifically includes the following steps:
[0023] The lithofacies data after dimension reduction is used as a lithofacies sample set, and the lithofacies sample set is divided into a training set and a test set. The training set is used to establish the lithology parameters of the neural network logging lithology recognition model, and the test set is used to confirm and optimize the accuracy of the neural network logging lithology recognition model.
[0024] The training set is imported into the neural network well logging lithology identification model for training, and the training results are output. The characteristic parameters of the neural network well logging lithology identification model are established according to the training results; the test set is then imported into the neural network well logging lithology identification model for testing, and the test results are output. The current test results are compared with the training results to determine whether the two results are consistent; if they are consistent, the current neural network well logging lithology identification model runs normally and the output results are normal; if they are inconsistent, the training set is imported into the neural network well logging lithology identification model again for training, and the characteristic parameters of the neural network well logging lithology identification model are continuously optimized; the test set is then imported into the neural network well logging lithology identification model for testing, and the two results are continuously compared according to the above comparison process until the two results are consistent.
[0025] Furthermore, there are multiple lithofacies sample sets, and the lithological parameters of each lithofacies sample set are different. Each group is divided into a training set and a test set. According to the different lithological parameters in each lithofacies sample set, the characteristic parameters of the neural network logging lithology identification model are iteratively updated to optimize the characteristic parameters of the neural network logging lithology identification model.
[0026] Furthermore, the analysis of lithological deposits and establishment of lithological information in S4 and S5 specifically include the following steps:
[0027] Clean the currently collected lithofacies data of the well section to be studied, remove erroneous data, fill in missing data, unify data parameters, and set attribute labels for the remaining data; merge similar or consistent label parameters according to the label content, remove label parameters with a low proportion, and retain unrelated parameter categories;
[0028] The preprocessed lithofacies data of the well section to be studied is imported into the optimized neural network logging lithology identification model for parameter identification, and the probability of each lithology parameter of the current well section to be studied is output. According to the probability of the current lithology parameter category, the lithology parameter corresponding to the maximum probability is selected as the lithology information of the current well section to be studied.
[0029] Furthermore, the abnormal data and erroneous data in the logging lithofacies data to be processed are eliminated, and the eliminated abnormal data and erroneous data specifically include: eliminating the data of thin layers, lithology mutation sections and well wall collapse sections.
[0030] Furthermore, when comparing the test results with the training results, by obtaining the logging curves and logging depth range in the lithofacies sample set, the logging depth range and the corresponding logging curves are added to the training set and the test set to assist the neural network logging lithology recognition model in the exercise; and the actual lithology at the center depth point of the logging depth range is used as the actual lithology of the lithofacies sample set. By checking whether the lithology with the highest probability output by the test results and the training results is consistent with the actual lithology of the lithofacies sample set, it is obtained whether the test results are consistent with the training results and are accurate.
[0031] Compared with the prior art, the present invention has the following beneficial effects:
[0032] The present invention constructs a neural network logging lithology identification model, obtains historical lithofacies data as a lithofacies sample set, cleans it, and performs label and parameter dimensionality reduction processing, thereby improving the effectiveness of the lithofacies sample set, and uses the lithofacies sample set to establish and optimize the parameters of its model, thereby improving the accuracy of model training; after the lithofacies data of the well section to be studied are cleaned, they are introduced into the optimized neural network logging lithology identification model for training, and the probabilities of various lithologies of the well section to be studied are output, and the lithology with the maximum probability is selected as the lithology of the current well section to be studied, thereby quickly identifying the lithofacies information of the current well section to be studied, thereby avoiding the difficulty of traditional lithology identification methods in adapting to the increasing data dimensions and magnitudes, and improving the feasibility of the existing geophysical logging full-well section lithofacies intelligent identification method. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] Figure 1This is a flow chart of the geophysical logging full-well section lithofacies intelligent identification method of the present invention. DETAILED DESCRIPTION
[0034] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0035] In order to solve the technical problem that the traditional lithology identification method has been unable to adapt to the increasing data dimensions and magnitudes as the degree of digitalization of the oil and gas industry deepens, this embodiment provides the following technical solutions:
[0036] A method for intelligently identifying lithofacies in the entire well section of geophysical logging comprises the following steps:
[0037] S1. Obtaining data: Collecting historical lithofacies related data and collecting lithofacies data of the current well section to be studied; wherein, collecting historical lithofacies data and current lithofacies data of the well section to be studied specifically includes the following steps:
[0038] By checking previous logging related literature, multiple groups of historical logging lithofacies data are collected, and the logging lithofacies data with relatively accurate data are selected as sample data to establish and optimize the lithology parameters of the neural network logging lithology identification model; specifically, by collecting relevant literature on logging lithofacies, the lithofacies information of the previous logging sections is obtained as the lithofacies sample set; in order to optimize the lithology parameters of the neural network logging lithology identification model, the lithology data of the collected literature can be diverse, such as mudstone, sandstone and other lithology data, so as to enrich the model parameters of the neural network logging lithology identification model and improve the practicality and effectiveness of the neural network logging lithology identification model.
[0039] The geological and lithofacies data of the current well section to be studied are collected by adopting the mechanical groove method. The specific geological and lithofacies data include: rock layer, lithofacies, sedimentary structure information, logging curve and logging depth; specifically, by selecting a well section in a certain area as the target well section to be studied, the geological and lithofacies data of the well section to be studied are collected by adopting the mechanical groove method, and the data are used as the data basis for identifying the lithofacies information in the well section to be studied.
[0040] S2. Processing data: cleaning and preprocessing the historical lithofacies data, and performing label dimension reduction and parameter dimension reduction on the processed historical lithofacies data, and using the processed historical lithofacies data as a lithofacies sample set; wherein, cleaning and preprocessing the historical lithofacies data specifically includes the following steps:
[0041] Abnormal data and erroneous data in the logging lithofacies data to be processed are eliminated, missing data are supplemented, and the pre-processed data are standardized to ensure the accuracy and consistency of the lithofacies data. Then, labels are set up for the lithofacies data according to their attributes, and each lithofacies data is classified with the corresponding logging curve and logging depth; wherein, abnormal data and erroneous data in the logging lithofacies data to be processed are eliminated, and the eliminated abnormal data and erroneous data specifically include: eliminating thin layers, lithology mutation sections and well wall collapse sections; specifically, by treating thin layers, lithology mutation sections and well wall collapse sections as abnormal and erroneous data and eliminating them, the invalid data of historical lithofacies data is reduced, and the historical data is improved. The effectiveness of lithofacies data; the data is then standardized, the main purpose of which is to eliminate the differences in dimensions and numerical magnitudes in the data, so that different characteristic variables have the same scale, thereby making the subsequent analysis of the data through the neural network logging lithology identification model more accurate and effective; this process adjusts the data values to a unified range to ensure that each feature has an equal influence on the final result, avoiding the situation where some features dominate the analysis results because the original values are too large or too small; further, the data are labeled according to the data category, so as to more intuitively understand the content covered by the current lithofacies sample set, further facilitate the streamlining of effective data, and improve the accuracy of model exercises.
[0042] Secondly, the historical lithofacies data are processed for label dimension reduction and parameter dimension reduction, which specifically includes the following steps:
[0043] Based on the label information of the preprocessed lithofacies data, similar lithologies are merged to reduce the label ratio of similar and repeated lithologies, and the lithology labels and lithology data with a ratio lower than the preset ratio are deleted; the labels obtained after label dimension reduction processing include seven lithologies: mudstone, sandstone, shale, dolomite, conglomerate, marl and gypsum; specifically, in order to further streamline the effective data in the lithofacies sample set, similar lithologies are merged to effectively reduce the label ratio of repeated or similar lithologies, and at the same time, the lithology labels and lithology data with a smaller ratio are deleted.
[0044] In this embodiment, for example, the preset proportion should be greater than or equal to 5%. After checking, it is found that the proportion of granite labels is 4%, which means that granite is lower than the preset proportion, and the possibility that granite is the lithofacies of the current well section is low. Therefore, the granite label and granite data can be deleted, thereby effectively reducing useless data.
[0045] Based on the parameter categories of the pre-processed lithofacies data, two or more parameter categories with parameters higher than the preset correlation are selected for dimensionality reduction, and parameter categories that are not related to each other are retained; the parameters obtained after the parameter dimensionality reduction processing include five input parameters: compensation sound waves, compensation electrons, compensation density, natural gamma and resistivity; specifically, as described in the above embodiment, the parameters are reduced in the same steps by dimensionality reduction processing, so as to optimize the effectiveness of the lithofacies sample set; at the same time, the lithofacies data of the current well section to be studied can be optimized and simplified in the same way as the above operation, thereby improving the feasibility of the lithofacies intelligent identification method for the entire well section.
[0046] S3. Establish model parameters: construct a neural network logging lithology recognition model, import the processed historical lithofacies data into the neural network logging lithology recognition model for training and testing, and output the training results and test results in turn, check whether the probabilities of various lithologies in the two results are consistent, and establish and optimize the lithology parameters of the neural network logging lithology recognition model based on the verification results; wherein, the lithofacies sample sets are multiple groups, and the lithology parameters of each lithofacies sample set are different, and each group is correspondingly divided into a training set and a test set. According to the different lithology parameters in each lithofacies sample set, the characteristic parameters of the neural network logging lithology recognition model are iteratively updated to optimize the characteristic parameters of the neural network logging lithology recognition model; import the processed historical lithofacies data into the neural network logging lithology recognition model for training and testing, specifically including the following steps:
[0047] The lithofacies data after dimension reduction is used as a lithofacies sample set, and the lithofacies sample set is divided into a training set and a test set. The training set is used to establish the lithology parameters of the neural network logging lithology recognition model, and the test set is used to confirm and optimize the accuracy of the neural network logging lithology recognition model.
[0048] The training set is imported into the neural network well logging lithology recognition model for training, and the training results are output. The characteristic parameters of the neural network well logging lithology recognition model are established according to the training results; the test set is then imported into the neural network well logging lithology recognition model for testing, and the test results are output. The current test results are compared with the training results to determine whether the two results are consistent; if they are consistent, the current neural network well logging lithology recognition model is running normally and the output results are normal; if they are inconsistent, the training set is imported into the neural network well logging lithology recognition model again for training, and the characteristic parameters of the neural network well logging lithology recognition model are continuously optimized; the test set is then imported into the neural network well logging lithology recognition model The test is carried out in the model, and the two results are compared according to the above comparison process until the two results are consistent. When comparing the test results with the training results, the logging curves and logging depth ranges in the lithofacies sample set are obtained, and the logging depth ranges and corresponding logging curves are added to the training set and the test set to assist the neural network logging lithology recognition model in the exercise. The actual lithology at the center depth point of the logging depth range is taken as the actual lithology of the lithofacies sample set. By checking whether the lithology with the highest probability output by the test results and the training results is consistent with the actual lithology of the lithofacies sample set, it is obtained whether the test results are consistent with the training results and are accurate.
[0049] In this implementation, for example, after the lithofacies sample set is preprocessed, it is divided into a training set and a test set. First, the training set is introduced into the neural network logging lithology recognition model for training. This step is not only used to establish the characteristic parameters of the neural network logging lithology recognition model, but also to obtain the output training results, that is, the various lithofacies probabilities of the entire well section corresponding to the lithology sample set, such as: the output training results are 35% mudstone, 40% sandstone, 25% shale, and 55% dolomite, which is the training result of this time; then the test set is introduced into the trained neural network logging lithology recognition model for testing. This step is intended to verify the accuracy of the training results, and the actual lithofacies information of the entire well section corresponding to the lithology sample set is obtained according to the accuracy; such as: the output test results are 35% mudstone, 40% sandstone, 25% shale, and 55% dolomite, which is the training result of this time; after comparison, it is found that the two output results are consistent, and the case with the largest probability value is selected as the actual lithofacies information of the current entire well section, that is, dolomite. Based on this method, the lithofacies information of the entire well section can be quickly identified.
[0050] S4. Analyze lithologic deposits: pre-process the lithofacies data of the current well section to be studied according to the processing step S2, and then import the optimized neural network logging lithologic identification model according to the step S3 to perform parameter identification, and output the probability of various lithologies of the current well section to be studied; specifically, after the neural network logging lithologic identification model is optimized according to the above embodiment, the accuracy of the output of the neural network logging lithologic identification model is relatively accurate at this time; the lithofacies data of the current well section to be studied can be cleaned and dimensionally reduced according to the above steps, and imported into the optimized neural network logging lithologic identification model for parameter identification until the identification result is output, and then the parameter with the largest probability value in the screening area is used as the actual lithofacies information of the current well section to be studied, thereby improving the lithofacies information identification function of the existing geophysical full-well section.
[0051] S5, establishing lithology information: according to the parameter training results output in S4, the lithology with the maximum probability in the results is selected as the lithology information of the current well section to be studied; wherein, analyzing lithology deposition and establishing lithology information specifically includes the following steps:
[0052] The lithofacies data of the well section to be studied currently collected are cleaned, erroneous data are eliminated, missing data are filled, data parameters are unified, and attribute labels are set for the remaining data; similar or consistent label parameters are merged according to the label content, label parameters with a low proportion are eliminated, and irrelevant parameter categories are retained; the preprocessed lithofacies data of the well section to be studied are imported into the optimized neural network logging lithology identification model for parameter identification, and the probability of each lithology parameter of the current well section to be studied is output. According to the probability of the current lithology parameter category, the lithology parameter corresponding to the maximum probability is selected as the lithology information of the current well section to be studied.
[0053] The beneficial effects achieved by the above scheme are as follows: through the above operation, the effectiveness of the lithofacies sample set is improved, thereby improving the accuracy of the neural network logging lithology identification model training, thereby avoiding the difficulty of traditional lithology identification methods in adapting to the increasing data dimensions and magnitudes, and improving the feasibility of existing logging section lithofacies identification methods.
[0054] Working principle: By constructing a neural network logging lithology identification model, historical lithofacies data are obtained for cleaning and label and parameter dimension reduction, and the lithofacies sample set is used to establish and optimize the parameters of the model; after the lithofacies data of the well section to be studied are cleaned, they are imported into the optimized model for drills, and the probabilities of various lithologies of the well section to be studied are output, and the lithology with the maximum probability is selected as the actual lithology. Based on this method, the lithofacies information of the current well section to be studied can be quickly identified.
[0055] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.
[0056] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and alterations may be made to the embodiments without departing from the principles and spirit of the invention.
Claims
1. A method for intelligent identification of lithofacies in the entire well section of geophysical logging, comprising the following steps: S1. Obtain data: collect historical lithofacies related data and collect lithofacies data of the current well section to be studied; S2. Data processing: cleaning and preprocessing the historical lithofacies data, and performing label dimension reduction and parameter dimension reduction on the processed historical lithofacies data, and using the processed historical lithofacies data as a lithofacies sample set; S3. Establish model parameters: construct a neural network logging lithology recognition model, import the processed historical lithofacies data into the neural network logging lithology recognition model for training and testing, and output the training results and test results in sequence, check whether the probabilities of various lithologies in the two results are consistent, and establish and optimize the lithology parameters of the neural network logging lithology recognition model based on the verification results; S4, Analyze lithology and sedimentation: After pre-processing the lithofacies data of the current well section to be studied according to the processing step S2, import the optimized neural network logging lithology recognition model according to the step S3 to perform parameter recognition, and output the probability of various lithologies of the current well section to be studied; S5. Establishing lithology information: Based on the parameter training results output in S4, the lithology with the maximum probability in the results is selected as the lithology information of the current well section to be studied.
2. The method for intelligent identification of lithofacies in the entire well section of geophysical logging according to claim 1, characterized in that: The historical lithofacies data and current lithofacies data of the well section to be studied are collected in S1, which specifically includes the following steps: By checking the previous logging related literature, we collected multiple sets of historical logging lithofacies data, and selected the logging lithofacies data with relatively accurate data as sample data to establish and optimize the lithology parameters of the neural network logging lithology identification model; The mechanical notching method is used to collect geological and lithofacies data of the well section to be studied. The specific geological and lithofacies data include: rock layer, lithofacies, sedimentary structure information, logging curve and logging depth.
3. The method for intelligent identification of lithofacies in the entire well section of geophysical logging according to claim 1, characterized in that: In S2, the historical lithofacies data is cleaned and preprocessed, specifically comprising the following steps: Abnormal and erroneous data in the logging lithofacies data to be processed are eliminated, missing data are supplemented, and the pre-processed data are standardized to ensure the accuracy and consistency of the lithofacies data. Then, labels are set for the lithofacies data according to their attributes, and each lithofacies data is classified with the corresponding logging curve and logging depth.
4. The method for intelligent identification of lithofacies in the entire well section of geophysical logging according to claim 3, characterized in that: In S2, the historical lithofacies data are subjected to label dimensionality reduction and parameter dimensionality reduction processing, which specifically includes the following steps: Based on the label information of the pre-processed lithofacies data, similar lithologies are merged to reduce the label ratio of similar and repeated lithologies, and the lithology labels and lithology data with a ratio lower than the preset ratio are deleted; Based on the parameter categories of the preprocessed lithofacies data, two or more parameter categories with parameters higher than the preset correlation are selected for dimensionality reduction, and parameter categories that are unrelated to each other are retained.
5. The method for intelligent identification of lithofacies in the entire well section of geophysical logging according to claim 4, characterized in that: The labels obtained after the label dimensionality reduction process include seven lithologies: mudstone, sandstone, shale, dolomite, conglomerate, marl and gypsum; the parameters obtained after the parameter dimensionality reduction process include five input parameters: compensated acoustic wave, compensated electron, compensated density, natural gamma and resistivity.
6. The method for intelligent identification of lithofacies in the entire well section of geophysical logging according to claim 1, characterized in that: In S3, the processed historical lithofacies data is imported into the neural network logging lithology recognition model for training and testing, which specifically includes the following steps: The lithofacies data after dimension reduction is used as a lithofacies sample set, and the lithofacies sample set is divided into a training set and a test set. The training set is used to establish the lithology parameters of the neural network logging lithology recognition model, and the test set is used to confirm and optimize the accuracy of the neural network logging lithology recognition model. The training set is imported into the neural network well logging lithology identification model for training, and the training results are output. The characteristic parameters of the neural network well logging lithology identification model are established according to the training results; the test set is then imported into the neural network well logging lithology identification model for testing, and the test results are output. The current test results are compared with the training results to determine whether the two results are consistent; if they are consistent, the current neural network well logging lithology identification model runs normally and the output results are normal; if they are inconsistent, the training set is imported into the neural network well logging lithology identification model again for training, and the characteristic parameters of the neural network well logging lithology identification model are continuously optimized; the test set is then imported into the neural network well logging lithology identification model for testing, and the two results are continuously compared according to the above comparison process until the two results are consistent.
7. The method for intelligent identification of lithofacies in the entire well section of geophysical logging according to claim 6, characterized in that: There are multiple lithofacies sample sets, and the lithological parameters of each lithofacies sample set are different. Each group is divided into a training set and a test set. According to the different lithological parameters in each lithofacies sample set, the characteristic parameters of the neural network logging lithology identification model are iteratively updated to optimize the characteristic parameters of the neural network logging lithology identification model.
8. The method for intelligent identification of lithofacies in the entire well section of geophysical logging according to claim 1, characterized in that: The analysis of lithological deposits and establishment of lithological information in S4 and S5 specifically include the following steps: Clean the currently collected lithofacies data of the well section to be studied, remove erroneous data, fill in missing data, unify data parameters, and set attribute labels for the remaining data; merge similar or consistent label parameters according to the label content, remove label parameters with a low proportion, and retain unrelated parameter categories; The preprocessed lithofacies data of the well section to be studied is imported into the optimized neural network logging lithology identification model for parameter identification, and the probability of each lithology parameter of the current well section to be studied is output. According to the probability of the current lithology parameter category, the lithology parameter corresponding to the maximum probability is selected as the lithology information of the current well section to be studied.
9. The method for intelligent identification of lithofacies in the entire well section of geophysical logging according to claim 3, characterized in that: The abnormal data and erroneous data in the logging lithofacies data to be processed are eliminated, and the eliminated abnormal data and erroneous data specifically include: eliminating the data of thin layers, lithology mutation sections and well wall collapse sections.
10. The method for intelligent identification of lithofacies in the entire well section of geophysical logging according to claim 6, characterized in that: When comparing the test results with the training results, the logging curves and logging depth ranges in the lithofacies sample set are obtained, and the logging depth ranges and corresponding logging curves are added to the training set and the test set to assist the neural network logging lithology recognition model in the exercise; and the actual lithology at the center depth point of the logging depth range is taken as the actual lithology of the lithofacies sample set. By checking whether the lithology with the highest probability output by the test results and the training results is consistent with the actual lithology of the lithofacies sample set, it is determined whether the test results are consistent with the training results and are accurate.