A method and apparatus for identifying sedimentary microfacies based on well logs

Abstract

The application discloses a method and device for identifying sedimentary microfacies based on well logging curves, and the method comprises the following steps: acquiring all well logging curves and well logging interpretation results of a research area; pre-processing all well logging curves; obtaining a reconstructed characteristic curve by using the pre-processed well logging curves; constructing a training set and a test set by using the pre-processed well logging curve data, the reconstructed characteristic curve data and corresponding single-well sedimentary microfacies interpretation result data; training a pre-established sedimentary microfacies identification model by using the training set and the test set, so as to obtain a trained sedimentary microfacies identification model; and identifying the sedimentary microfacies of a target well by using the trained sedimentary microfacies identification model according to the pre-processed well logging curve data and the reconstructed characteristic curve data of the target well.

Description

Technical Field

[0001] This article relates to the field of geological research technology for oil and gas field exploration and development, and in particular to a method and device for identifying sedimentary microfacies based on well logging curves. Background Technology

[0002] Sedimentary microfacies are a key research area in oil and gas exploration and development, aiming to provide effective model support for reservoir parameter prediction and 3D geological modeling. Traditional methods typically rely on regional sedimentary background and paleontological and sedimentological theories, combined with core facies markers and well logging morphology characteristics, to analyze rock thickness, grain size, etc., thereby determining sedimentary types and classifying sedimentary microfacies (Su et al., 2020; Li et al., 2018). While current theories and methods for sedimentary facies analysis are relatively mature and can guarantee a certain level of accuracy, they still have significant limitations. The process is complex, requires high levels of expertise, and interpretation is subjective. This results in a large workload, long study time, and difficulty in ensuring accurate mapping between sedimentary facies and well logging data, affecting the accuracy and efficiency of oil and gas production prediction.

[0003] With the outstanding performance of artificial intelligence across disciplines, scholars both domestically and internationally have successively carried out research on reservoir logging interpretation methods based on machine learning-based intelligent algorithms. Compared with conventional methods, machine learning has higher efficiency and accuracy. It should be noted that logging curves, as geophysical responses of sediments formed under specific sedimentary environments, contain a wealth of information. This also leads to unclear key response characteristics of reservoir types. Simply building a single logging identification model based on conventional logging curves is far from sufficient for predicting sedimentary microfacies in complex reservoirs.

[0004] Therefore, there is an urgent need for an intelligent logging identification method for sedimentary microfacies based on feature reconstruction, in order to overcome the problems of low accuracy in sedimentary microfacies interpretation caused by the influence of human subjectivity and the lack of clarity in the key response characteristics of reservoir types. Summary of the Invention

[0005] To address the aforementioned technical problems, the present invention aims to provide a method and apparatus for identifying sedimentary microfacies based on well logging curves. This method, based on the analysis of sedimentary microfacies types in the target area, improves data quality, reduces uncertainty in data interpretation, and better trains the sedimentary microfacies identification model through well logging curve preprocessing and feature reconstruction, thereby improving the efficiency and accuracy of sedimentary microfacies interpretation.

[0006] In a first aspect, this application provides a method for identifying sedimentary microfacies based on well logging curves, the method comprising:

[0007] All well logging curves and interpretation results for the study area are acquired; all well logging curves are preprocessed, and reconstructed feature curves are obtained using the preprocessed well logging curves; training and testing sets are constructed using the preprocessed well logging curve data, reconstructed feature curve data, and corresponding single-well sedimentary microfacies interpretation results; a pre-established sedimentary microfacies identification model is trained using the training and testing sets to obtain a trained sedimentary microfacies identification model; based on the preprocessed well logging curve data and reconstructed feature curve data of the target well, the sedimentary microfacies of the well are identified using the trained sedimentary microfacies identification model.

[0008] Secondly, embodiments of the present invention also provide an apparatus for identifying sedimentary microfacies based on well logging curves. The apparatus includes a memory and a processor. The memory is used to store a program for identifying sedimentary microfacies based on well logging curves, and the processor is used to read and execute the program for identifying sedimentary microfacies based on well logging curves, and to execute the method described in any of the above embodiments.

[0009] Thirdly, embodiments of the present invention also provide a computer-readable storage medium storing a data processing program, wherein the data processing program is executed by a processor using the method for identifying sedimentary microfacies based on well logging curves as described in any of the above embodiments.

[0010] Compared with related technologies, this application provides a method and apparatus for identifying sedimentary microfacies based on well logging curves. The method includes: acquiring all well logging curves and well logging interpretation results in the study area; preprocessing all well logging curves to obtain reconstructed feature curves; constructing a training set and a test set from the preprocessed well logging curve data, the reconstructed feature curve data, and the corresponding single-well sedimentary microfacies interpretation results data; training a pre-established sedimentary microfacies identification model using the training set and the test set to obtain a trained sedimentary microfacies identification model; and identifying the sedimentary microfacies of the target well using the trained sedimentary microfacies identification model based on the preprocessed well logging curve data and the reconstructed feature curve data. This application obtains reconstructed well logging curves by reconstructing the features of well logging curves in the study area, trains a sedimentary microfacies identification model using the preprocessed well logging curves, the reconstructed feature curves, and the corresponding single-well sedimentary microfacies interpretation results data, and then accurately identifies the sedimentary microfacies types in the study area using the trained sedimentary microfacies identification model.

[0011] Other features and advantages of this application will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the application. Other advantages of this application can be realized and obtained by means of the solutions described in the description and the accompanying drawings. Attached Figure Description

[0012] The accompanying drawings are used to provide an understanding of the technical solutions of this application and constitute a part of the specification. They are used together with the embodiments of this application to explain the technical solutions of this application and do not constitute a limitation on the technical solutions of this application.

[0013] Figure 1 This is a flowchart of the method for identifying sedimentary microphases according to an embodiment of this application;

[0014] Figure 2 This is a schematic diagram of a depositional microphase identification device according to an embodiment of this application;

[0015] Figure 3 This is a schematic diagram of the well logging curve preprocessing and feature reconstruction process in some exemplary embodiments;

[0016] Figure 4 This is a schematic diagram illustrating the algorithm model construction and optimization process in some exemplary embodiments;

[0017] Figure 5 This is a schematic diagram of the technical process of Bi-LSTM in some exemplary embodiments;

[0018] Figure 6 This is a schematic diagram of the LSTM technology process in some exemplary embodiments;

[0019] Figure 7 This is a schematic diagram illustrating sedimentary microfacies types and logging curve characteristics in some exemplary embodiments;

[0020] Figure 8 This is a schematic diagram of the confusion matrix of the algorithm model prediction results in some exemplary embodiments;

[0021] Figure 9 This is a schematic diagram illustrating the blind well identification effect of the algorithm model in some exemplary embodiments. Detailed Implementation

[0022] This application describes several embodiments, but these descriptions are exemplary and not restrictive, and it will be apparent to those skilled in the art that many more embodiments and implementations are possible within the scope of the embodiments described herein. Although many possible combinations of features are shown in the drawings and discussed in the detailed description, many other combinations of the disclosed features are also possible. Unless specifically limited, any feature or element of any embodiment may be used in combination with, or may replace, any feature or element of any other embodiment.

[0023] This application includes and contemplates combinations of features and elements known to those skilled in the art. The embodiments, features, and elements disclosed in this application may also be combined with any conventional features or elements to form a unique inventive scheme as defined by the claims. Any feature or element of any embodiment may also be combined with features or elements from other inventive schemes to form another unique inventive scheme as defined by the claims. Therefore, it should be understood that any feature shown and / or discussed in this application may be implemented individually or in any suitable combination. Therefore, the embodiments are not limited except by the limitations imposed by the appended claims and their equivalents. Furthermore, various modifications and changes may be made within the scope of the appended claims.

[0024] Furthermore, in describing representative embodiments, the specification may have presented methods and / or processes as a specific sequence of steps. However, the method or process should not be limited to the specific order of steps described herein, to the extent that it does not depend on such a specific order. As will be understood by those skilled in the art, other sequences of steps are also possible. Therefore, the specific order of steps set forth in the specification should not be construed as a limitation of the claims. Moreover, the claims concerning the method and / or process should not be limited to the steps performed in the written order, and those skilled in the art will readily understand that these orders can be varied and still remain within the spirit and scope of the embodiments of this application.

[0025] Currently, intelligent logging identification technology for sedimentary microfacies has gradually shifted from the experience-driven model of traditional geologists to automated methods based on deep learning. Traditional methods rely on manual analysis of logging curve morphology characteristics (such as bell-shaped, funnel-shaped, etc.) and core data, which suffers from high subjectivity and low efficiency. Commonly used intelligent identification methods include Support Vector Machine (SVM), Random Forest (RF), Convolutional Neural Network (CNN), and Recurrent Neural Network (RNN). These methods have made significant progress in the application of sedimentary microfacies identification, but some limitations still exist.

[0026] Well logging curve feature reconstruction refers to the process of preprocessing and extracting features from raw well logging data to generate new curves or feature sets that better reflect subsurface geological characteristics. Its purpose is to eliminate noise, enhance useful signals, and extract key features, thereby improving the performance of subsequent intelligent recognition models. Addressing the noise sensitivity and long-range dependency issues of high-sampling-rate well logging signals (length > 1000), this method employs feature construction techniques to preprocess well logging curves, constructing feature variables such as geological trend decomposition, median filtering, and K-means clustering, and then combining this with the Bi-LSTM algorithm for sequence modeling. This method significantly improves model robustness by eliminating noise interference and extracting multi-scale features, making it suitable for complex reservoir environments.

[0027] Bi-LSTM (Bidirectional Long Short-Term Memory) is an improved recurrent neural network (RNN) specifically designed for processing sequential data. Compared to traditional RNNs, Bi-LSTM, by introducing a bidirectional structure (forward and backward LSTM networks), can simultaneously capture past and future information in sequential data, thus overcoming the gradient vanishing and short-term memory problems inherent in RNNs with long sequences. RNNs and their variants (such as LSTM) have already seen some applications in sedimentary microfacies identification, for example, by processing well logging sequence data to identify sedimentary microfacies. However, traditional RNNs are prone to the gradient vanishing problem when processing long sequences, resulting in insufficient ability to capture long-term dependencies. Bi-LSTM, through its bidirectional structure, significantly improves the model's ability to model sequential data, better capturing complex patterns and geological features in well logging curves.

[0028] Different neural network methods have significant differences in input data formats and parameter settings. RNN / LSTM algorithms require one-dimensional time-series data (single logging curve) as input and are only suitable for short sequences and low noise. CNN models can accept multi-dimensional data (multiple logging curves) and two-dimensional images as input, and can extract spatial features, making them suitable for situations with imaging logging. XGBoost models require logging curves and structured features (statistics without physical meaning) as input, and can perform multi-scale feature modeling of complex nonlinear relationships between logging curves, making them suitable for small samples and no sequence. Bi-LSTM models support multi-dimensional spatiotemporal joint feature matrices (original logging curves, reconstructed features, and depth sequences), are suitable for long sequences and high noise data, and have good generalization ability while ensuring accuracy.

[0029] This invention provides a method for identifying sedimentary microfacies based on well logging curves, such as... Figure 1 As shown, the method includes steps S1-S5:

[0030] S1: Obtain all logging curves and logging interpretation results for the study area;

[0031] S2: Preprocess all logging curves and use the preprocessed logging curves to obtain the reconstructed characteristic curves;

[0032] S3: The training set and test set are constructed from the preprocessed well logging curve data, the reconstructed characteristic curve data and the corresponding single-well sedimentary microfacies interpretation results data;

[0033] S4: The pre-established sedimentary microfacies identification model is trained using the training set and the test set to obtain a trained sedimentary microfacies identification model;

[0034] S5: Based on the preprocessed logging curve data and reconstructed characteristic curve data of the target well, the sedimentary microfacies of the well are identified using the trained sedimentary microfacies identification model.

[0035] In one exemplary embodiment, based on previous research, a detailed macroscopic and microscopic description of the core samples obtained from the target area is performed, including rock type, structure, sedimentary structures, and fossils. Based on the core analysis, the possible sedimentary microfacies types in the study area are identified according to the core analysis and testing data. For example, if the study area is located at the delta front, the sedimentary microfacies types include: mouth bars, distributary channels, bar margins, and interdistributary bays. The primary focus is on identifying mouth bars, distributary channels, and bar margins. For interdistributary bays, since this sedimentary microfacies mainly consists of mudstone, it is generally not identified.

[0036] In this embodiment, core analysis and testing generally include thin section photographs (ordinary photographs, cast thin sections), nuclear magnetic resonance data, mercury intrusion porosimetry data (high-pressure mercury intrusion porosimetry, constant-rate mercury intrusion porosimetry), electron micrographs, fluid inclusion data, cathodoluminescence, etc. No specific limitations are imposed on this; any techniques feasible in this field can be used to identify the possible sedimentary microfacies types in the study area.

[0037] In one exemplary embodiment, the preprocessing of the logging curve ( Figure 3 This includes: depth correction, outlier and missing value handling, baseline correction and smoothing, and normal transformation.

[0038] Due to geological factors and the influence of logging instruments, logging curves exhibit depth deviations, and the data range of the same logging data varies between different wells, which can affect the accuracy of subsequent identification. Therefore, based on depth correction of logging curve data in the study area using core data and analytical test data, further processing of outliers and missing values, baseline correction and smoothing, and normal transformation were performed to eliminate the influence of the logging instruments themselves on the logging data and ensure data reliability.

[0039] In one exemplary embodiment, the process of obtaining the reconstructed characteristic curve using the preprocessed logging curve is as follows:

[0040] The preprocessed well logging curves were decomposed using a time series decomposition algorithm with different odd-frequency windows to remove periodic components and residual terms, thereby obtaining multiple geological trend component features.

[0041] Using the same frequency window as the trend decomposition, median filtering was applied to the preprocessed logging curves to obtain multiple median filtering features.

[0042] Multi-scale curve slope features are extracted from the preprocessed logging curves using the spline interpolation algorithm.

[0043] The slope of the curve is obtained by using the spline interpolation algorithm to calculate the continuous second derivative, thereby obtaining the curve's inclination characteristics.

[0044] In one exemplary embodiment, the reconstructed characteristic curve is obtained using the preprocessed logging curve, including:

[0045] Step 1: Use the time series decomposition algorithm to decompose the preprocessed logging curves using different odd-frequency windows, remove periodic components and residual terms, and obtain multiple geological trend component features.

[0046] Step 2: Using the same frequency window as the trend decomposition, perform median filtering on the preprocessed logging curves to obtain multiple median filtering features;

[0047] Step 3: Extract multi-scale curve slope features from the preprocessed logging curves using the spline difference algorithm;

[0048] Step 4: Use the spline interpolation algorithm to obtain the continuous second derivative of the curve slope to obtain the curve inclination characteristics.

[0049] In this embodiment, specifically: the morphological changes of the logging curve can reflect the sedimentary cycle and sedimentary background, and are closely related to the amplitude, top-to-bottom contact relationship, smoothness, shape, and tooth centerline of the logging curve. Corresponding features can be reconstructed based on the logging curve. The specific technical roadmap for feature reconstruction and optimization is as follows: Figure 3 As shown below:

[0050] Step 31: Decompose the logging curve using different odd-frequency windows according to the STL algorithm, remove the periodic component and the remainder term, and obtain multiple geological trend component features;

[0051] The Time Series Decomposition (STL) algorithm uses different odd-numbered frequency windows for well logging curves, with values ​​ranging from 3 to 20. The filter window value needs to be tested multiple times and selected based on actual needs; larger values ​​result in smoother logging. Compared to even-numbered filtering, odd-numbered filtering has a center point and is more sensitive to the overall shape and edge features of the well logging curve, allowing for more effective extraction of edge information.

[0052] The characteristic curve after removing periodic components and residual terms can not only remove high-frequency noise from the original curve, retain geological trend characteristics and effective edge morphology characteristics, but also fully reflect the spatiotemporal correlation of the curve sequence.

[0053] Step 32: Using the same frequency window as the trend decomposition, perform median filtering on the original curve to obtain multiple median filtering features, which can effectively suppress and eliminate noise interference, making the intelligent interpretation of sedimentary microfacies by well logging data more accurate.

[0054] Step 33: Extract multi-scale curve slope features from the logging curves using the spline difference algorithm.

[0055] In this embodiment, the scale is divided into three types: large-scale curve slope characteristics, medium-scale curve slope characteristics, and small-scale curve slope characteristics. The scale mainly refers to the thickness of the sand body.

[0056] Large-scale sand bodies are mainly multiple sand bodies stacked together vertically. The thickness of the composite sand body in the large-scale area is similar to that in the target area (the thickness of the composite sand body is generally >20m). The slope characteristics of the curve in the large-scale area reflect the characteristics of the micro-phase composite.

[0057] The mesoscale is approximately the same as the scale of a single sand body in the target area (the thickness of the composite sand body is generally <20m), and the slope characteristics of the curve at the mesoscale reflect the slope characteristics of the curve of a single microfacies.

[0058] The thickness at small scales ranges from 0.125 m to 1 m. The changes in the slope of the small-scale curves reflect the internal structural characteristics of a single microfacies. In summary, the multi-scale slopes are a good indicator of the contact relationships and internal characteristics of sedimentary microfacies. The geological factors affecting the changes in the curve slopes are mainly lithology and physical properties. A certain degree of heterogeneity also exists within a single microfacies, which in turn leads to changes in the small-scale curve slopes.

[0059] Step 34: Use the spline interpolation algorithm to obtain the continuous second derivatives based on the curve slope to obtain the curve's inclination characteristics.

[0060] In one exemplary embodiment, selecting sensitive reconstructed logging characteristic curves from the reconstructed logging characteristic curves includes:

[0061] The first step is to extract and reconstruct well logging characteristic curves and single-well sedimentary microfacies interpretation results using the same sampling interval, and calculate the correlation coefficient between the two.

[0062] The second step is to identify reconstructed logging characteristic curves with correlation coefficients greater than a threshold as reconstructed logging characteristic curves that are sensitive to sedimentary microfacies.

[0063] The correlation coefficient mentioned above is the Pearson correlation coefficient.

[0064] The preferred process can be as follows: using the same sampling interval (e.g., 0.125m), extract the single-well sedimentary microfacies interpretation and reconstruction characteristic curves, and calculate the correlation coefficient (e.g., Pearson correlation coefficient); the closer the correlation coefficient is to 1, the better the reconstruction characteristic curve matches the sedimentary microfacies interpretation. A threshold can be set according to the specific situation of the study area. For example, if the threshold is set to 0.5, then the reconstruction characteristic curve with a correlation coefficient greater than 0.5 is determined as the sensitive curve.

[0065] In one exemplary embodiment, in step S3, a training set and a test set are constructed from the preprocessed well logging curve data, reconstructed characteristic curve data, and corresponding single-well sedimentary microfacies interpretation results data, including:

[0066] Step S31: Preprocess the well logging curve data that has been interpreted, and use the preprocessed well logging curves to obtain the reconstructed characteristic curves;

[0067] Step S32: Based on the correlation coefficient between the well logging interpretation results data and the reconstructed characteristic curve, select the most sensitive reconstructed characteristic curve;

[0068] Step S33: Using the selected sensitive reconstructed feature curves and preprocessed logging curve data, combined with the corresponding logging interpretation results, construct a sample dataset;

[0069] The sample dataset consists of preprocessed logging curves, reconstructed characteristic curves, and single-well sedimentary microfacies interpretations that correspond one-to-one.

[0070] In one exemplary embodiment, in step S4, the sample dataset established using the preprocessed well logging curve data, reconstructed characteristic curve data, and single-well sedimentary microfacies interpretation results data is divided into a training sample dataset and a test sample dataset in an 8:2 ratio; alternatively, a different ratio, such as 7:3, can be used depending on the specific circumstances. The pre-established sedimentary microfacies identification model is trained using the training set and the test set to obtain a trained sedimentary microfacies identification model; the specific training process is as follows:

[0071] Step S41: Train the sedimentary microfacies recognition model using the training sample set and the Bi-LSTM algorithm, and perform cross-validation to obtain the initial sedimentary microfacies recognition model;

[0072] Step S42: If the accuracy of cross-validation is less than 90%, optimize the model parameters using Bayesian methods and re-execute S41; if the accuracy of cross-validation is greater than or equal to 90%, perform blind well test using the test sample dataset and continue to execute step S43.

[0073] Step S43: If the blind well test result is less than 80%, then reselect the logging curve and reconstruct the characteristic curve, and repeat the above operations (S41, S42 and S43); if the blind well test result is greater than or equal to 80%, then determine that the initial sedimentary microfacies identification model is a trained sedimentary microfacies identification model.

[0074] In step S41, a sedimentary microfacies identification model is trained using the training sample set and cross-validation is performed to obtain an initial sedimentary microfacies identification model, including:

[0075] Step S411: Set the initial parameters of the model. The main parameters to be set include the number of layers of transmission neurons and the number of iterations.

[0076] Step S412: Input the training sample set into the sedimentary microfacies recognition model for training, and perform a 5-fold cross-validation to obtain the initial sedimentary microfacies recognition model;

[0077] The sedimentary microfacies identification model takes as input the preprocessed logging curves and reconstructed feature curves as input and outputs the sedimentary microfacies type as output.

[0078] In this embodiment, the sedimentary microfacies identification model uses Bi-LSTM (Bidirectional Long Short-Term Memory), which is an improved recurrent neural network (RNN). Figure 5 This is a schematic diagram illustrating the overall principle of the Bi-LSTM model (the Bi-LSTM algorithm is a collection of multiple LSTM models). Figure 6 This is a schematic diagram illustrating the working principle of a single LSTM model. The input to the sedimentary microfacies identification model is the reconstructed feature curve, and the output is the sedimentary microfacies type.

[0079] The present invention has the following effects:

[0080] 1) This invention combines well logging curves with geological model understanding, enabling batch intelligent and automated interpretation of sedimentary microfacies of sand bodies in a single well, which greatly improves the efficiency of sedimentary microfacies interpretation.

[0081] 2) This invention significantly reduces the bias that may be caused by human subjective judgment, realizes quantitative and high-precision comprehensive sedimentary microfacies identification, and improves the accuracy of sand body sedimentary microfacies identification;

[0082] 3) This invention has good generalization ability and can be learned and trained according to the actual geological conditions of the research area. It can perform sedimentary microfacies identification work for blocks with different sedimentary backgrounds.

[0083] Secondly, embodiments of the present invention also provide a device for identifying sedimentary microfacies based on well logging curves, such as... Figure 2 As shown, the device includes a memory 200 and a processor 210; the memory is used to store a program for identifying sedimentary microfacies based on well logging curves, and the processor is used to read and execute the program for identifying sedimentary microfacies based on well logging curves, and to execute the method described in any of the above embodiments.

[0084] Thirdly, embodiments of the present invention also provide a computer-readable storage medium storing a data processing program, wherein the data processing program is executed by a processor using the method for identifying sedimentary microfacies based on well logging curves as described in any of the above embodiments.

[0085] Example 1

[0086] This example demonstrates the process of training, selecting, and optimizing a model for sedimentary microfacies types and well logging curves using machine learning algorithms. The technical roadmap for algorithm model construction and optimization is shown below. Figure 4 As shown, the details are as follows:

[0087] Step 401: Data preparation.

[0088] The preprocessed well logging data, the optimized reconstructed feature curves, and the corresponding single-well sedimentary microfacies interpretations were used as the sample dataset.

[0089] The sample dataset includes preprocessed logging curves, reconstructed characteristic curves, depth, and interpretation results of single-well sedimentary microfacies.

[0090] Step 402, Model Construction and Training:

[0091] The Bi-LSTM algorithm model was used to construct and train a sedimentary microfacies logging interpretation model by dividing the training set and the test set in a ratio of 8:2.

[0092] Among them, the Bi-LSTM algorithm model is a specific technical roadmap for the sedimentary microfacies interpretation model constructed based on the Bi-LSTM algorithm, such as... Figure 5 As shown, the details are as follows:

[0093] (a) Initialize network parameters, including weights and biases;

[0094] (b) Forward layer: The forward LSTM layer computes the sequence sequentially from the first element to the last element. The reverse LSTM layer computes the sequence in reverse order from the last element to the first element.

[0095] (c) Gating mechanism: The forget gate decides to discard information that is not relevant to the interpretation of sedimentary microfacies, the input gate decides to update part of the data of the cell state, and the output gate decides the final output.

[0096] (d) Update the hidden state and cell state according to the gating mechanism.

[0097] (e) Reverse layer: Calculate the gradient of the loss function with respect to the network parameters.

[0098] (f) Parameter update: Update network parameters using an optimization algorithm.

[0099] (g) Output merging: The outputs of the forward and backward LSTMs are merged to form the final bidirectional hidden state.

[0100] (h) Loss calculation: Calculate the loss based on the task type.

[0101] (i) Iterative optimization: Repeat the forward propagation, back propagation and parameter update steps until the model converges.

[0102] The specific technical roadmap of the LSTM algorithm model is as follows: Figure 6 As shown, the specific steps are the same as those described in (c) to (h) above;

[0103] Step 403: Cross-validation. To improve the reliability and generalization ability of the model, cross-validation is used to test the model's performance.

[0104] Step 404: Bayesian hyperparameter optimization.

[0105] Since the algorithm model contains multiple algorithm modules and has a large number of parameters, manual parameter tuning is inefficient and yields poor results. Therefore, the Bayesian hyperparameter intelligent optimization algorithm is used to optimize the model parameters.

[0106] Step 405, Blind well test.

[0107] To evaluate the practical application effect of the model, blind wells (i.e., test sample set) that were not used in the training were applied to the study area.

[0108] Step 406: Obtain the trained sedimentary microfacies interpretation model constructed based on the Bi-LSTM algorithm.

[0109] The sedimentary microfacies interpretation model that meets the requirements of blind well testing is used as the well-trained sedimentary microfacies interpretation model constructed based on the Bi-LSTM algorithm.

[0110] Example 2

[0111] This example uses Block A of the XX oilfield to demonstrate in detail the specific implementation process of identifying sedimentary microfacies based on well logging curves, as follows:

[0112] Step 1: Analysis of sedimentary microfacies types in the target area.

[0113] Based on previous research, a detailed macroscopic and microscopic description of the core samples obtained from the target area was conducted, clarifying the existence of four sedimentary microfacies in the study area: the main body of the estuary bar, the distributary channel, the bar margin, and the interdistributary bay. Figure 7 The diagram shows the sedimentary microfacies types and logging curve characteristics.

[0114] Step 2: Preprocess the logging curves.

[0115] Building upon deep correction, further steps are taken to handle outliers and missing values, perform baseline correction and smoothing, and apply a normal transformation. The specific process is as follows: Figure 3 As shown:

[0116] The original logging data was depth-corrected using core data and analytical test data. The depth-corrected original logging data was then preprocessed with logging curves, including outlier and missing value handling, baseline correction and smoothing, and normal transformation.

[0117] Step 3: Feature reconstruction and optimization.

[0118] Step 31: Reconstruct well logging characteristic curves

[0119] The original curve is reconstructed in multiple dimensions using trend decomposition and median filtering to calculate the slope, tilt angle, and degree of tilt of the curve.

[0120] The specific implementation process of reconstructing the well logging characteristic curve is as follows:

[0121] Step 311: Use the time series decomposition algorithm to decompose the preprocessed logging curves using different odd-frequency windows, remove periodic components and residual terms, and obtain multiple geological trend component features.

[0122] Step 312: Using the same frequency window as the trend decomposition, perform median filtering on the preprocessed logging curves to obtain multiple median filtering features;

[0123] Step 313: Extract multi-scale curve slope features from the preprocessed logging curves using the spline difference algorithm;

[0124] Step 314: Use the spline interpolation algorithm to obtain the continuous second derivative of the curve slope to obtain the curve inclination characteristics.

[0125] Step 32: Optimize and reconstruct the characteristic curve

[0126] After the reconstruction of the characteristic curves is completed, a correlation analysis is performed on the interpretation of sedimentary microfacies in a single well and the reconstruction characteristic curves, and the reconstruction characteristic curves that are sensitive to the interpretation of sedimentary microfacies are selected.

[0127] Step 4: Training and optimization of the sedimentary microfacies identification model.

[0128] Sample data was constructed using sensitive reconstruction feature curves to train and optimize the sedimentary microfacies identification model.

[0129] It is important to note that the interflow bays in the study area are considered as background facies and are not the focus of this research. Instead, sedimentary microfacies identification models were established for three key favorable sedimentary microfacies types in the study area. Evaluation showed that the sedimentary microfacies identification models achieved an interpretation accuracy of 86.3%, and the prediction accuracy for all three sedimentary microfacies types was relatively high. Figure 8 The diagram shows the confusion matrix of the prediction results of the algorithm model.

[0130] A confusion matrix is ​​a commonly used tool for evaluating the performance of classification models, particularly suitable for classification problems in supervised learning. It displays the model's performance in matrix form by comparing the model's predictions with the actual labels. Figure 8 The confusion matrix of the sedimentary microfacies identification model prediction results shown indicates that the trained sedimentary microfacies identification model meets the evaluation criteria.

[0131] Step 5: Intelligent interpretation of sedimentary microfacies in a single well.

[0132] The trained sedimentary microfacies recognition model is used to interpret the sedimentary microfacies of a single well, such as... Figure 9 The diagram shows the recognition effect based on the sedimentary microfacies recognition model.

[0133] The model predicted three sedimentary microfacies—distributary channels, mouth bars, and bar margins—with an overall accuracy of 86.3%. Specifically, the prediction accuracy for distributary channels was 85.4%, for mouth bars 89.3%, and for bar margins 82.3%.

[0134] In this embodiment, the predicted parameters demonstrate that the intelligent identification method for sedimentary microfacies based on the Bi-LSTM algorithm effectively reduces the subjective influence of human factors, improves the accuracy and reliability of predictions, and provides an innovative and efficient intelligent identification solution for the oil and gas exploration field.

[0135] It will be understood by those skilled in the art that all or some of the steps, systems, or apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all components may be implemented as software executed by a processor, such as a digital signal processor or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software may be distributed on a computer-readable medium, which may include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

Claims

1. A method for identifying sedimentary microfacies based on well logging curves, characterized in that, The method includes: Obtain all logging curves and logging interpretation results for the study area; All logging curves are preprocessed, and the reconstructed characteristic curves are obtained using the preprocessed logging curves. Training and testing sets were constructed from preprocessed well logging curve data, reconstructed characteristic curve data, and corresponding single-well sedimentary microfacies interpretation results. The pre-established sedimentary microfacies identification model is trained using the training set and the test set to obtain a trained sedimentary microfacies identification model; Based on the preprocessed logging curve data and reconstructed characteristic curve data of the target well, the sedimentary microfacies of the well are identified using a trained sedimentary microfacies recognition model.

2. The method for identifying sedimentary microfacies based on well logging curves according to claim 1, characterized in that, The step of training a pre-established sedimentary microfacies identification model using the training set and the test set to obtain a trained sedimentary microfacies identification model includes: The pre-established sedimentary microfacies identification model was trained using sample data from the training set to obtain the initial sedimentary microfacies identification model; The initial sedimentary microfacies identification model was tested using sample data from the test set. If the blind well test result was less than 80%, the logging curves and reconstructed features were reselected, and the above training process was repeated. If the blind well inspection test result is greater than or equal to 80%, then the initial sedimentary microfacies identification model is determined to be a trained sedimentary microfacies identification model.

3. The method for identifying sedimentary microfacies based on well logging curves according to claim 2, characterized in that, The process of training a pre-established sedimentary microfacies identification model using sample data from the training set to obtain an initial sedimentary microfacies identification model includes: Set the relevant parameters for the sediment microfacies recognition model, including the number of layers of transit neurons and the number of iterations; The sedimentary microfacies identification model was trained using sample data from the training set, and cross-validation was performed. If the accuracy of cross-validation is less than 90%, Bayesian optimization is used to optimize the parameters and continue training the sedimentary microfacies identification model. If the cross-validation accuracy is greater than or equal to 90%, then the sedimentary microfacies identification model will be used as the initial sedimentary microfacies identification model. The sedimentary microfacies identification model takes as input the preprocessed logging curves and reconstructed feature curves as input and outputs the sedimentary microfacies type as output.

4. The method for identifying sedimentary microfacies based on well logging curves according to claim 2, characterized in that, The construction of training and testing sets from preprocessed well logging data, reconstructed characteristic curve data, and corresponding single-well sedimentary microfacies interpretation results includes: The well logging curve data that has been interpreted is preprocessed, and the reconstructed characteristic curve is obtained using the preprocessed well logging curves; Based on the correlation coefficient between the well logging interpretation data and the reconstructed characteristic curve, the most sensitive reconstructed characteristic curve is selected. By utilizing the selected sensitive reconstructed feature curves and preprocessed logging curve data, combined with the corresponding logging interpretation results, a sample dataset is constructed. The well logging interpretation data includes the types of sedimentary microfacies.

5. The method for identifying sedimentary microfacies based on well logging curves according to claim 4, characterized in that, The preprocessed logging curve data and reconstructed characteristic curve data in the sample dataset correspond to the sedimentary microfacies types in the logging interpretation results data.

6. The method for identifying sedimentary microfacies based on well logging curves according to claim 4, characterized in that, The step of selecting sensitive reconstruction feature curves based on the correlation coefficient between the well logging interpretation data and the reconstruction feature curves includes: Sample points for reconstructing characteristic curves and sample points for sedimentary microfacies interpretation results were extracted using the same sampling interval; The correlation coefficient between the reconstructed characteristic curve and the sedimentary microfacies interpretation results was calculated using the extracted sample points; reconstructed characteristic curves with a correlation coefficient greater than 0.5 were identified as sensitive reconstructed characteristic curves. The correlation coefficient mentioned above is the Pearson correlation coefficient.

7. The method for identifying sedimentary microfacies based on well logging curves according to claim 1, characterized in that, The preprocessing of the logging curves includes: depth correction, outlier and missing value processing, baseline correction and smoothing, and normal transformation; The reconstructed feature curves include: geological trend components, noise reduction features, multi-scale curve slope, and inclination degree.

8. The method for identifying sedimentary microfacies based on well logging curves according to claim 7, characterized in that, The process of obtaining the reconstructed characteristic curve from the preprocessed logging curve is as follows: The preprocessed well logging curves were decomposed using a time series decomposition algorithm with different odd-frequency windows to remove periodic components and residual terms, thereby obtaining multiple geological trend component features. Using the same frequency window as the trend decomposition, median filtering was applied to the preprocessed logging curves to obtain multiple median filtering features. Multi-scale curve slope features are extracted from the preprocessed logging curves using the spline interpolation algorithm. The slope of the curve is obtained by using the spline interpolation algorithm to calculate the continuous second derivative, thereby obtaining the curve's inclination characteristics.

9. A device for identifying sedimentary microfacies based on well logging curves, characterized in that, The device includes a memory and a processor; the memory is used to store a program for identifying sedimentary microfacies based on well logging curves, and the processor is used to read and execute the program for identifying sedimentary microfacies based on well logging curves, and to execute the method according to any one of claims 1-8.

10. A computer-readable storage medium storing a data processing program, the data processing program being executed by a processor according to any one of claims 1-8, the method for identifying sedimentary microfacies based on well logging curves.

Images