Automatic identification method for sedimentary microfacies of braided river delta

Abstract

The invention provides a braided river delta sedimentary microfacies automatic identification method, and relates to the technical field of oil and gas field development, the method comprises the following steps: S1, well logging curve preprocessing: selecting a well logging curve according to well logging data, and carrying out smoothing and layering processing on the well logging curve; s2, a sample data set is constructed, specifically, a sedimentary microfacies characteristic parameter sample set is established by calculating characteristic parameters of all layers, and the characteristic parameter sample set comprises a training sample set and a testing sample set; s3, establishing an automatic identification model, including constructing an XGoos t automatic identification sedimentary microfacies model, and using the model to train and optimize the training sample set; and S4, sedimentary microfacies automatic identification: predicting the test sample set by using XGoos t to realize sedimentary microfacies automatic identification. According to the method, the sedimentary microfacies can be quickly, effectively, quantitatively and accurately identified.

Description

Technical Field

[0001] The invention relates to the technical field of oil and gas field development, in particular to an automatic identification method for braided river delta sedimentary microfacies. Background Art

[0002] Sedimentary microfacies are the smallest sedimentary combinations within a sedimentary subfacies zone, characterized by their unique lithology, rock texture, structure, thickness, rhythmicity, and specific planar distribution patterns. From a sedimentary perspective, the sedimentary environment and conditions control the development, spatial distribution, and internal structure of sand bodies. Sand bodies formed in different sedimentary environments exhibit distinct reservoir characteristics, which in turn influence the migration, accumulation, and development of oil and gas.

[0003] Traditional sedimentary microfacies identification methods mainly determine sedimentary facies types through seismic attributes, study the sedimentological characteristics of core wells, and manually and finely characterize sedimentary microfacies based on rock type, structure, and texture combined with the combined characteristics of logging curves. During the identification process, they often face problems such as large data volumes, strong subjectivity in stratification, repetitive identification processes, and low identification efficiency. At the same time, they also require a high level of identification experts.

[0004] With the rise of machine learning, qualitative and quantitative prediction technologies have played a huge role in promoting the development of various disciplines. The identification research of sedimentary microfacies has also entered the stage of digital quantitative identification combined with computers. It is of great practical significance to find an automatic identification method for sedimentary microfacies to replace manual identification in work and achieve the purpose of improving identification efficiency.

[0005] However, existing sedimentary microfacies identification methods lack noise reduction processing for logging data and rarely achieve automatic stratification. Manual identification often faces problems such as huge core or logging data, strong subjectivity of stratification, repetitive identification process, low identification efficiency, and high expert level requirements in sedimentary microfacies identification.

[0006] Therefore, an automatic identification method for braided river delta sedimentary microfacies is urgently needed to solve the above technical problems. Summary of the Invention

[0007] The present invention aims to provide a method for automatically identifying sedimentary microfacies in braided river deltas. The method first smoothes the natural gamma ray log curve, then uses the ordered sample optimal segmentation method to extract the natural gamma ray log curve characteristics and divide the layers. Characteristic parameters for each layer are calculated, and a sample set of sedimentary microfacies characteristic parameters is established. An XGBoost machine learning algorithm model is constructed using Python. The training sample set is trained to find optimal hyperparameters, and predictions are made on the test sample set to achieve automatic identification of sedimentary microfacies. The various technical effects that can be achieved by the preferred technical solution among the many technical solutions provided by the present invention are detailed below.

[0008] To achieve the above objectives, the present invention provides the following technical solutions:

[0009] The present invention provides a method for automatically identifying sedimentary microfacies in a braided river delta, comprising the following steps:

[0010] S1: Logging curve preprocessing, including selecting logging curves according to logging data, and smoothing and layering the logging curves;

[0011] S2: constructing a sample data set, including establishing a sedimentary microfacies characteristic parameter sample set by calculating characteristic parameters of each layer, wherein the characteristic parameter sample set includes a training sample set and a test sample set;

[0012] S3: Establish an automatic identification model, including building an XGBoost automatic identification model for sedimentary microfacies, and use the model to train and optimize the training sample set;

[0013] S4: Automatic identification of sedimentary microfacies, including using XGBoost to predict the test sample set to achieve automatic identification of sedimentary microfacies.

[0014] Preferably, in step S1, the selected logging curve is a natural gamma ray curve.

[0015] Preferably, in step S1, a sliding average method is used to perform smoothing filtering on the logging curve.

[0016] Preferably, in step S1, the layered processing of the well logging curves includes:

[0017] Segmenting the well logging curve using an ordered sample optimal segmentation method;

[0018]

[0019] Among them, b(n,k) means dividing n ordered samples into k categories, using the sum of squared deviations to represent the degree of difference D(i,j) between samples of the same type, and using the loss function L to determine the optimal number of categories.

[0020] Preferably, in step S2, the sedimentary microfacies logging characteristic parameter sample set established includes the average value, relative center of gravity, average slope, mutation amplitude, and root of variance of the natural gamma ray logging curve.

[0021] Preferably, in step S3, grid search and cross-validation methods are used to find the optimal hyperparameters.

[0022] Preferably, a tree model is selected as a booster in the XGBoos t model.

[0023] Preferably, the hyperparameters include maximum depth, random sampling ratio and learning rate.

[0024] Preferably, the maximum depth has a value range of [0, 10]; the random sampling ratio has a value range of [0, 1]; and the learning rate has a value range of [0, 1].

[0025] Preferably, in step S3, the feature parameter sample set is divided into a training sample set and a test sample set in a ratio of 7:3.

[0026] This invention provides a method for automatically identifying sedimentary microfacies in braided river deltas. It uses a sliding average method to reduce noise in well logging curves and an ordered sample optimal segmentation method to automatically stratify the log curves. The method also extracts five features from the natural gamma ray curves: the mean, relative center of gravity, average slope, mutation amplitude, and root of variance variance (RSV) to fully characterize the logging characteristics of different sedimentary microfacies. The extracted features are then used to establish an identification model using the XGBoost algorithm for sedimentary microfacies prediction. In summary, this invention addresses the challenges of manual sedimentary microfacies identification, such as large data volumes, repetitive identification processes, strong subjectivity, and low efficiency, achieving rapid, effective, quantitative, and relatively accurate identification of sedimentary microfacies. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0028] Figure 1 This is a flow chart of a method for automatically identifying sedimentary microfacies in a braided river delta provided by an embodiment of the present invention;

[0029] Figure 2 This is a schematic diagram of the segmentation results and classification number selection of the optimal segmentation method for ordered samples provided by an embodiment of the present invention;

[0030] In the figure: (a) Segmentation result of the optimal segmentation method; (b) Selection of the number of categories; (c) Comparison between manual stratification and automatic stratification;

[0031] Figure 3 It is a box plot of characteristic parameters of different sedimentary microfacies provided by the embodiments of the present invention.

[0032] Figure 4 These are examples of sample characteristics of different sedimentary microfacies provided by the embodiments of the present invention.

[0033] Figure 5 is a confusion matrix of sedimentary microfacies identification results provided by an embodiment of the present invention.

[0034] Figure 6 This is the sedimentary microfacies identification result of a certain well provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0035] To make the objectives, technical solutions, and advantages of the present invention more apparent, the technical solutions of the present invention will be described in detail below. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other implementations obtained by those of ordinary skill in the art without inventive effort are within the scope of protection of the present invention.

[0036] The technical solution of the present invention will be further described below with reference to specific embodiments.

[0037] The specific implementation scheme of the present invention is illustrated by taking the identification of sedimentary microfacies in the sedimentary environment of a shallow braided river delta in the study area as an example.

[0038] The present invention provides an automatic identification method for braided river delta sedimentary microfacies, such as Figure 1 As shown, the following steps are included:

[0039] S1: Logging curve preprocessing, including selecting logging curves according to logging data, and smoothing and stratifying the logging curves.

[0040] S2: constructing a sample data set, including establishing a sedimentary microfacies characteristic parameter sample set by calculating characteristic parameters of each layer, wherein the characteristic parameter sample set includes a training sample set and a test sample set;

[0041] S3: Establish an automatic identification model, including building an XGBoost automatic identification model for sedimentary microfacies, and use the model to train and optimize the training sample set;

[0042] S4: Automatic identification of sedimentary microfacies, including using XGBoost to predict the test sample set to achieve automatic identification of sedimentary microfacies.

[0043] As an emerging algorithm, XGBoost has excellent performance in both classification and regression prediction. It is simple and efficient, has low computational complexity, and is highly fault-tolerant, effectively preventing overfitting during prediction. This example uses the XGBoost algorithm to establish a sedimentary microfacies identification model.

[0044] Specifically, in step S1 : in the well logging curve preprocessing, firstly the well logging data is comprehensively considered and a suitable well logging curve is selected.

[0045] This embodiment fully considers the logging curve data, morphological characteristics, and other factors, selecting the natural gamma ray (GR) curve. The sliding average method is used to smooth the logging curve. This method reduces noise by calculating the average value of data points within a certain window while preserving the data trend. The window size determines the degree of smoothing. A larger window will result in a smoother curve but will slow the reflection of trends, while a smaller window will be more sensitive to data fluctuations. Selecting an appropriate number of window points is crucial for use. This embodiment uses the sliding average method to smooth the logging curve, reducing the impact of extreme values and severely jagged data. It also makes the curve morphology more distinct, facilitating automatic stratification.

[0046] After the curve is smoothed, the curve is layered by extracting its morphological features to reasonably determine the minimum vertical sedimentation unit. In this embodiment, the ordered sample optimal segmentation method is used to segment the natural gamma ray curve to achieve layered processing of the curve.

[0047] The optimal segmentation method for ordered samples is a linear discriminant method based on the idea of variance analysis that can better distinguish between populations. If there are n data points, there will be 2n-1 possible division methods. Among all the divisions, a division method is found that minimizes the difference between samples within each segment and maximizes the difference between segments.

[0048] b(n,k) represents a method for classifying n ordered samples into k categories. The sum of squared deviations is used to represent the degree of difference between samples of the same type (which can be called the diameter of the group D(i,j)). This minimizes the difference between samples of the same type and maximizes the difference between samples of different categories. The loss function L is used to determine the optimal number of categories. The more categories, the smaller the loss value and the smaller the error. However, the more categories, the better. It is recommended to select an appropriate number of categories around the inflection point of the loss curve.

[0049] Diameter D:

[0050]

[0051] The loss function is:

[0052]

[0053] In this example, the number of classifications when the slope of the loss curve is 0.05 is selected according to the requirements of the subdivision of the sedimentary microfacies in the study area. The segmentation results, the selection of the number of classifications, and the schematic diagram of the comparison with the manual stratification are shown in Figure 2 .

[0054] Step S2: Construct a sample data set. After smoothing and stratifying the GR curve, combine core observation, logging interpretation, and expert identification results to divide the sedimentary microfacies. Calculate the quantitative identification characteristic parameters of each layer and establish a sample set of logging characteristic parameters for each sedimentary microfacies.

[0055] Among them, the established sedimentary microfacies logging characteristic parameter sample set includes the average value, relative center of gravity, average slope, mutation amplitude, and root of variance of the natural gamma ray logging curve.

[0056] The above feature parameters are used as the input features X( Figure 3 ) to establish an automatic identification model for sedimentary microfacies, and the established model was generalized and verified based on the test set.

[0057] In this embodiment, based on the sedimentary microfacies type of the braided river delta, Y = {0, 1, 2, 3} is used as the output category, where category "0" represents distributary channel microfacies, category "1" represents interchannel microfacies, category "2" represents distributary sand bar microfacies, and category "3" represents sheet sand microfacies.

[0058] After determining the input features and output results, we further extracted sample data to construct the sample dataset required for the XGBoost classification model. The dataset was divided into a training set and a test set. The training set was used to train and adjust the parameters of the XGBoost model, while the test set was used to verify the model's recognition accuracy and generalization capabilities.

[0059] When allocating sample data for the training set and the test set, the ratio of the training set to the test set is controlled at about 7:3. Later, when predicting the surrounding wells, the entire data set will be used as the training set. In this example, 2957 sedimentary microfacies samples were selected as training samples and test samples for the XGBoos t model, divided into 2070 training sets and 887 test sets. See (part of) the training sample data set. Figure 4 .

[0060] Step S3: In establishing an automatic identification model, Python is used to build an XGBoost machine learning algorithm model, that is, to automatically identify the sedimentary microfacies model.

[0061] The model was used to train the training sample set and predict the test sample set. The grid search and cross-validation methods were used to find the optimal hyperparameters and obtain the mapping relationship between sedimentary microfacies and logging curve characteristics.

[0062] Among them, the commonly used boosters in XGBoost are tree models and linear models. Since linear models can only find linear segmentation, while tree models can find nonlinear segmentation, and tree models are closer to human thinking, they can generate visual classification rules, and the generated model is interpretable, so tree models are selected as boosters.

[0063] Before training the model, you need to set hyperparameters. The hyperparameters that have the greatest impact on model training in the tree model are the maximum depth (max_depth), learning rate (learning_rate), and random sampling ratio (subsample). That is, the maximum depth of the tree. The larger the value, the more branches the tree has and the easier it is to cause overfitting. The value range of max_depth is [0,+∞]. Increasing this value will increase the complexity of the model. In this embodiment, it is [0,10]. Random sampling ratio (subsample): This parameter controls the random sampling ratio of each tree. It is a parameter that prevents overfitting of the model, but setting it too small will lead to underfitting. The value range is [0,1]. Learning rate (learning_rate): Prevents overfitting by controlling the step size. The value range is [0,1].

[0064] In order to find the most suitable hyperparameters for classification, parameter optimization is required. This embodiment uses the more mainstream grid search and cross-validation method to optimize the process as follows:

[0065] Based on experience, the parameter ranges are selected. In this embodiment, the grid search range of the maximum depth is set to [1, 11] with a step size of 1; the learning rate and random sampling ratio range are set to [0, 1] with a step size of 0.01. Through 5-fold cross validation, the maximum depth is finally determined to be 10, the learning rate is 0.2, and the random sampling ratio is 1. It is known that 2070 training samples are used for training and 887 test samples are used for verification. Figure 5 As shown in the figure, among the 887 test samples, the prediction accuracy can reach 92.67%, which is in line with the expected effect as a whole, indicating that the XGBoost classification model has good classification and recognition effects and meets the needs of practical applications.

[0066] Step S4: In the automatic identification of sedimentary microfacies, the XGBoost automatic identification model is used to predict the test samples to verify the model's recognition accuracy and generalization ability, thereby achieving intelligent prediction of sedimentary microfacies.

[0067] Specifically, all 2957 samples were used as training sets to predict the surrounding wells. The prediction results were compared with the manually divided sedimentary microfacies. The accuracy of identifying facies zones above 3m (channels, sand bars, and inter-channels) exceeded 96%, and the accuracy of identifying facies zones between 1 and 2m exceeded 87% (e.g. Figure 6 The reasons for the errors are: first, the limited number of training and prediction data sets; second, the large variation in sand body thickness, which makes it difficult for the extracted characteristic parameters to fully reflect the characteristics of the sedimentary microfacies; and third, the complex sedimentary conditions, which means that the characteristics of the same sedimentary microfacies on the well logging curve may not be the same. However, a 5%-15% error rate is sufficient to meet the needs of practical geological applications.

[0068] The single-well sedimentary microfacies prediction results are processed, that is, adjacent interpretation results of the same sedimentary microfacies are merged to complete the final single-well sedimentary microfacies interpretation.

[0069] This automatic identification method for braided river delta sedimentary microfacies can achieve rapid, large-scale and effective identification of sedimentary microfacies types through machine learning, greatly improving work efficiency and effectively solving problems such as strong subjectivity, large workload and repetitive processes in manual identification of sedimentary microfacies.

[0070] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for automatically identifying sedimentary microfacies in braided river deltas, characterized in that: The steps include: S1: Logging curve preprocessing, including selecting logging curves according to logging data, and smoothing and layering the logging curves; S2: constructing a sample data set, including establishing a sedimentary microfacies characteristic parameter sample set by calculating characteristic parameters of each layer, wherein the characteristic parameter sample set includes a training sample set and a test sample set; S3: Establish an automatic identification model, including building an XGBoost automatic identification model for sedimentary microfacies, and use the model to train and optimize the training sample set; S4: Automatic identification of sedimentary microfacies, including using XGBoost to predict the test sample set to achieve automatic identification of sedimentary microfacies.

2. The method for automatically identifying sedimentary microfacies of a braided river delta according to claim 1, characterized in that: In step S1, the selected logging curve is a natural gamma ray curve.

3. The method for automatically identifying sedimentary microfacies of a braided river delta according to claim 2, characterized in that: In step S1, a sliding average method is used to perform smoothing and filtering on the logging curve.

4. The method for automatically identifying sedimentary microfacies of a braided river delta according to claim 3, characterized in that: In step S1, the layered processing of the well logging curves includes: Segmenting the well logging curve using an ordered sample optimal segmentation method; Among them, b(n,k) means dividing n ordered samples into k categories, using the sum of squared deviations to represent the degree of difference D(i,j) between samples of the same type, and using the loss function L to determine the optimal number of categories.

5. The method for automatically identifying sedimentary microfacies of a braided river delta according to any one of claims 2 to 4, characterized in that: In step S2, a sample set of sedimentary microfacies logging characteristic parameters is established, wherein the characteristic parameters include the average value, relative center of gravity, average slope, mutation amplitude, and root of variance of the natural gamma ray logging curve.

6. The method for automatically identifying sedimentary microfacies of a braided river delta according to any one of claims 1 to 4, characterized in that: In step S3, grid search and cross-validation methods are used to find the optimal hyperparameters.

7. The method for automatically identifying sedimentary microfacies of a braided river delta according to claim 6, characterized in that: The tree model is used as the booster in the XGBoost model.

8. The automatic identification method of braided river delta sedimentary microfacies according to claim 7, characterized in that: The hyperparameters include maximum depth, random sampling ratio, and learning rate.

9. The method for automatically identifying sedimentary microfacies of a braided river delta according to claim 8, characterized in that: The maximum depth ranges from [0, 10]; the random sampling ratio ranges from [0, 1]; and the learning rate ranges from [0, 1].

10. The method for automatically identifying sedimentary microfacies of a braided river delta according to any one of claims 1 to 4, characterized in that: In step S3, the feature parameter sample set is divided into a training sample set and a test sample set in a ratio of 7:3.

Images