LU603268B1 - Method for Intelligently Identifying Sweet Spot Types in Shale Reservoirs Based on Fractal Characteristics - Google Patents

Abstract

The present invention discloses a method for intelligently identifying sweet spot types in shale reservoirs based on fractal characteristics, which relates to the technical field of shale reservoir evaluation. The method includes the following steps: calculating fractal attribute indicators of T2 spectra, obtaining new clustering labels, acquiring optimal fractal attribute values, deriving final classification labels, and applying the final classification criteria to all optimal fractal attribute values of the target well to obtain new classification results. By performing multifractal calculations on the original T2 spectrum data obtained from nuclear magnetic logging and using effective porosity data from sealed coring of corresponding intervals as a basis, the invention categorizes varying degrees of effective porosity. It then systematically trains machine learning models for the fractal attributes corresponding to each category, establishing a classification interval for fractal attribute values primarily based on the most influential fractal attribute. This interval is applied to all fractal attribute values of the target formation, enabling accurate classification and evaluation of shale reservoirs.

Description

Specification

LU603268

Method for Intelligently Identifying Sweet Spot Types in Shale

Reservoirs Based on Fractal Characteristics

Technical Field

The present invention relates to the technical field of shale reservoir evaluation, in particular to a method for intelligently identifying sweet spot types in shale reservoirs based on fractal characteristics.

Background Technology

Shale reservoirs refer to rock layers primarily composed of fine-grained sediments (such as clay, mudstone, and siltstone) that contain abundant organic matter and can generate and store natural gas or oil. The classification and evaluation of shale reservoirs involve systematic analysis and grading based on geological characteristics, physical properties, and gas-bearing potential of shale gas reservoirs. This process aims to identify and differentiate various types of shale reservoirs, providing scientific basis for shale gas exploration and development, thereby enabling more effective resource assessment and development decisions.

Nuclear magnetic resonance (NMR) logging technology has unique advantages in evaluating complex reservoirs. It can be used to assess the saturation of the study area using NMR logging. NMR logging measures the relaxation signals of hydrogen atoms in the formation, providing porosity and fluid information independent of lithology. A set of data 1

Specification

LU603268 obtained by measuring the transverse relaxation time (Tz) distribution of fluids in rock pores is called the T: spectrum. These data reflect detailed information about the pore structure of the rock, including pore size, pore fluid type (oil, gas, water), and their saturation. The T, spectrum is commonly used to identify and evaluate reservoir pore structures, including pore size distribution, pore fluid types, and saturation. By analyzing the T: spectrum, different fluids such as oil, gas, and water can be distinguished, as they exhibit distinct relaxation characteristics on the

T, spectrum. Therefore, NMR logging technology plays a significant role in reservoir classification and evaluation.

Since the T, spectrum distribution obtained from nuclear magnetic resonance logging is closely related to the size and distribution of rock pores, as well as lithology, pore fluid properties, wettability, and formation water salinity, the position of the oil peak in the T. spectrum may vary across different wells, and the T. cutoff value may also differ.

For low-permeability or tight reservoirs, the presence of (light) oil can cause the T, spectrum distribution to widen, making it difficult in some wells to separate bound fluid from movable fluid on the T. spectrum. This results in a certain deviation between the calculated water saturation from nuclear magnetic resonance logging and the actual situation, ultimately affecting the accuracy of reservoir evaluation and classification.

Therefore, this invention proposes a method for intelligently identifying 2

Specification

LU603268 sweet spot types in shale reservoirs based on fractal characteristics to address the issues in existing technologies.

Invention Content

To address the above issues, the objective of this invention is to propose a method for intelligently identifying sweet spot types in shale reservoirs based on fractal characteristics. This method solves the problem where existing shale reservoir classification and evaluation methods, which rely on water saturation calculated from nuclear magnetic resonance logging, exhibit deviations from the actual situation, leading to inaccurate classification and evaluation of shale reservoirs.

To achieve the objective of this invention, the following technical solution is implemented: The method for intelligently identifying sweet spot types in shale reservoirs based on fractal characteristics includes the following steps:

Step 1: Acquire and analyze the original T. spectrum data of the target well depth using nuclear magnetic resonance logging technology, then calculate relevant fractal attribute values from the processed data to obtain fractal attribute indicators for each set of T. spectra.

Step 2: Obtain movable porosity data of the target well and classify it into four categories as original classification labels. Combine the fractal attribute indicators with the original classification labels into a data table, then input the data table into a K-means clustering model. Use machine 3

Specification

LU603268 learning methods to cluster the fractal attribute values and derive corresponding new clustering labels.

Step 3: Analyze the importance of fractal attribute indicators using the random forest method, then select the fractal attribute with the highest importance from the analysis results as the final classification criterion, 1.e., the optimal fractal attribute value.

Step 4: Based on the correspondence between the original classification labels and the new clustering labels, calculate the proportion of each type within each clustering label and the proportion of each type across all clustering labels. Comprehensively analyze these proportions to establish a mapping relationship between clustering labels and original classification labels, and derive the final label for each set of data according to this mapping relationship.

Step 5: Based on the final labels and optimal fractal attribute values, determine the value ranges of optimal fractal attributes for each of the four label categories. Use these ranges as the final classification criteria and apply them to all optimal fractal attribute values of the target well, obtaining a new shale reservoir classification evaluation result primarily based on optimal fractal attributes.

A further improvement lies in: in step 1, the specific steps for analyzing and processing the original data of the target well depth T: spectrum are: performing dimensionless processing on the original T, 4

Specification ee LU603268 spectrum data using a dimensionless formula, so that all data converges within the interval [0,1].

A further improvement lies in: in step 1, using the calculation method of fractal dimension to compute and analyze the relevant fractal property values of the processed data, where the fractal property indicators of the T: spectrum include the singular intensity, offset, and maximum-minimum offset values of the T, spectrum.

A further improvement lies in: in step 2, the movable porosity data 1s obtained from the sealed coring data of the target well section, and the movable porosity data is divided into four categories in descending order.

A further improvement lies in: in step 2, the compiled data table undergoes data preprocessing before being input into the K-means clustering model. The specific steps are: first using data cleaning to remove missing values and outliers from the data table, then performing correlation analysis and dimensionality reduction on the data table, followed by standardizing the data table through normalization.

A further improvement lies in: in step 3, using the random forest method to analyze the importance of fractal property values, obtaining the ranking of the influence of three fractal property values on the clustering process, the influence degree values of the three fractal properties, and a histogram.

A further improvement lies in: in step 4, using the bidirectional

Specification ee LU603268 maximization method to calculate the proportion of each type within each cluster label and the proportion of each type across all cluster labels, with specific steps as follows: based on analyzing the differences and connections between cluster labels and original classification labels, the bidirectional maximization method is used to obtain the proportion of samples for each original label within each cluster label, as well as the proportion of each original label type across all cluster labels. A comprehensive analysis yields the mapping relationship between cluster labels and original classification labels. Based on this mapping relationship, the clustered data is reclassified to obtain the final classification labels.

A further improvement lies in: the mapping relationship is: {cluster label 0: original label 1, cluster label 1: original label 2, cluster label 2: original label 4, cluster label 3: original label 3}.

The beneficial effects of the present invention are: by performing multifractal calculations on the original T: spectrum data obtained from nuclear magnetic logging, the invention derives indicators such as the singular spectrum function and spectrum offset of the T, spectrum data.

Combining these with effective porosity data from sealed core samples of corresponding well sections, the K-means clustering method and machine learning methods like random forest are used to classify varying degrees of effective porosity. Systematic machine learning training is then 6

Specification

LU603268 conducted on the fractal attributes corresponding to each category, ultimately resulting in a classification interval for fractal attribute values primarily based on the most influential fractal attributes. This 1s applied to all fractal attribute values in the target formation, yielding a novel classification result that accurately evaluates shale reservoirs. Compared to traditional classification results, this method is more detailed and accurate.

Description of the Drawings

To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following briefly introduces the accompanying drawings required for describing the embodiments or the prior art. Obviously, the accompanying drawings in the following description are merely some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained from these accompanying drawings without creative effort.

Figure 1 is a schematic flowchart of the shale reservoir classification and evaluation method based on machine learning and optimal fractal attributes according to the present invention;

Figure 2 1s a schematic diagram of clustering results for three fractal attribute values in Embodiment 3 of the present invention;

Figure 3 is a schematic diagram of the correspondence between clustering labels and original classification labels in Embodiment 3 of the 7

Specification ee LU603268 present invention;

Figure 4 is a schematic diagram of the importance ranking results in

Embodiment 3 of the present invention;

Figure 5 is a schematic diagram of the final classification labels obtained from the mapping relationship in Embodiment 3 of the present invention;

Figure 6 1s a schematic comparison between the original classification results and the classification results of the method of the present invention in Embodiment 3.

Specific Embodiments

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. It 1s evident that the described embodiments are only a part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

Embodiment 1

Referring to Figure 1, this embodiment provides a method for intelligently identifying sweet spot types of shale reservoirs based on fractal characteristics, including the following steps:

Step 1: Parallelly analyze and process the raw data of the T, 8

Specification

Se LU603268 spectrum for the target well depth, and calculate the fractal attribute indicators of the T, spectrum.

First, use nuclear magnetic resonance logging technology to obtain and analyze the original data of the T: spectrum of the target well depth in parallel. Apply a dimensionless formula to process the original T: spectrum data, ensuring all data converges within the [0,1] interval to eliminate the influence of dimensions and improve the performance of subsequent model processing. Then, use the calculation method of fractal dimensions to compute relevant fractal attribute values for the processed data, obtaining fractal attribute indicators for each group of T: spectra, specifically including the singular intensity, offset, and maximum-minimum offset values of the T2 spectrum.

Step 2: Cluster the fractal attribute values using machine learning methods and obtain newclustering labels.

Obtain movable porosity data from the sealed core samples of the target well section. Categorize the movable porosity data of the target well into four classes in descending order as the original classification labels. Then, compile the fractal property indicators calculated in Step 1 with the original classification labels into a data table. Use this compiled data table as the input for the K-means clustering model. Apply machine learning methods to cluster the fractal property values, resulting in corresponding new cluster labels. Analogous to the classification of 9

Specification

LU603268 movable porosity into four categories in descending order in sealed core samples, this example also divides the cluster labels into four categories to facilitate the next step of establishing correspondences. After clustering, four distinct clustering results are obtained.

In this embodiment, the compiled data table undergoes data preprocessing before being input into the K-means clustering model. The specific steps are as follows: First, use data cleaning to remove missing values (check for missing values in the data table and choose to fill or delete them based on the situation) and outliers (identify and handle outliers to prevent adverse effects on clustering results). Then, perform correlation analysis (analyze the correlation between each feature and the fractal property values, selecting the most relevant features) and dimensionality reduction (if the feature dimensionality is high, use

Principal Component Analysis for dimensionality reduction to reduce computational load and improve model performance). Finally, standardize the data table using normalization (normalize the data to ensure all features are on the same scale, avoiding excessive influence on clustering results due to scale differences) to ensure data quality and model effectiveness.

Step 3: Analyze the importance degree of the fractal attribute values, and take the fractal attribute with the highest importance degree as the optimal fractal attribute value.

Specification ee LU603268

Use the random forest method to analyze the importance of the fractal property values obtained in Step 1. Obtain the ranking of the influence of the three fractal property values on the clustering process, their respective influence values, and a histogram. Then, select the fractal property with the highest importance from the analysis results as the final classification criterion, i.e., the optimal fractal property value, which serves as the primary basis for subsequent final classification.

Step 4: Obtain the final classification labels based on the mapping relationship between cluster labels and original classification labels.

Based on the correspondence between the original classification labels in Step 2 and the newly obtained cluster labels, analyze the differences and connections between cluster labels and original classification labels. Use a bidirectional maximization method to determine the proportion of each original label sample within each cluster label, as well as the proportion of each original label type across all cluster labels. Through comprehensive analysis, derive the mapping relationship between cluster labels and original classification labels.

Based on this mapping relationship, reclassify the clustered data to obtain the final classification labels, 1.e., the ultimate labels.

Step 5: Apply the final classification basis to all the optimal fractal attribute values of the target well to obtain the new classification results.

Based on the ultimate labels mapped in Step 4 and the optimal 11

Specification

LU603268 fractal attribute values obtained in Step 3, determine the intervals of optimal fractal attribute values for each of the four label categories. Use these intervals as the final classification criteria and apply them to all optimal fractal attribute values of the target well. This ultimately yields a new shale reservoir classification and evaluation result primarily based on optimal fractal attributes, achieving a shale reservoir classification and evaluation based on machine learning and optimal fractal attributes.

Embodiment 2

The data used in this embodiment is sourced from the Jimsar shale oil reservoir. The specific steps for shale reservoir classification and evaluation are as follows:

S1: Organize and select nuclear magnetic resonance (NMR) logging data and sealed core data from the target well. Calculate the fractal dimension of the NMR T. spectrum data to obtain the singularity intensity, offset, and maximum/minimum offset values of the T. spectrum.

S2: Conduct a joint analysis with the movable porosity data obtained from sealed coring. For the movable porosity from sealed coring, divide it into four categories from largest to smallest according to preset standards, serving as the original classification labels.

For the fractal attribute values derived from the T, spectrum, first perform dimensionless processing to ensure all data converges within the [0,1] interval, eliminating the influence of units and improving 12

Specification

LU603268 subsequent model performance. Then, use multifractal calculation methods to compute the three main fractal attribute values of the T: spectrum. Apply the K-means method to cluster these three fractal attribute values. Analogous to the sealed core data where movable porosity is divided into four categories from largest to smallest, the cluster labels are also divided into four categories here to facilitate the next step of finding correspondences. After clustering, four distinct cluster results are obtained.

S3: Use the random forest method to rank the influence of three fractal attribute values on the clustering process, obtain the influence degree values and histograms of the three fractal attributes, and record the fractal attribute with the highest influence as the optimal fractal attribute, which serves as the main basis for subsequent final classification. After the entire processing is completed, obtain the number of original classification samples corresponding to each cluster label. For example, under the cluster label 0, count how many samples have the original label 1, how many have the original label 2, and so on.

S4: Based on analyzing the differences and connections between cluster labels and original classification labels, use the bidirectional maximization method to obtain the proportion of each original label sample in each cluster label, as well as the proportion of each original label type across all cluster labels. Through comprehensive analysis, 13

Specification ee LU603268 establish the mapping relationship between cluster labels and original classification labels as: {Cluster label 0: Original label 1, Cluster label 1:

Original label 2, Cluster label 2: Original label 4, Cluster label 3: Original label 3}.

According to the above mapping relationship, reclassify the clustered data to obtain the final classification labels.

S5: Based on the optimal fractal attribute values corresponding to the final classification labels, determine the interval of the optimal fractal attribute values for each final label, which serves as the final classification standard for the entire classification model.

Since the data used for model training only corresponds to the movable porosity of sealed core samples and not all T, data, this embodiment applies the obtained final classification standard to the optimal fractal attribute values derived from all T, data. This divides all measured T: data intervals into four entirely new categories. Compared to traditional reservoir classification evaluation methods, the interpretation based on machine learning and optimal multifractal attributes is more detailed, yielding more reasonable results.

Embodiment 3

Refer to Figures 2, 3, 4, 5, and 6. The specific steps for shale reservoir classification evaluation in this embodiment are as follows:

A1. First, organize the obtained T, raw data into a standard data file, 14

Specification

LU603268 process it using a Python script with a dimensionless formula to achieve dimensionless treatment, and then organize the results into a standard data format for fractal calculations.

A2. Import the organized data into a Matlab script to calculate fractal attributes, obtaining three fractal property values: Aa, B, and Af.

Simultaneously, prepare movable porosity data obtained from sealed core samples, and categorize them into four classes from large to small according to specific standards as original class labels.

A3. Combine the T. fractal property values corresponding to the depths of movable porosity with the original classification labels as raw data for subsequent machine learning methods. Import the compiled data into a Python script for K-means clustering. Along with obtaining clustering results, use the random forest method to rank the importance of fractal property values. The ranking results are shown in Figure 4. After running the script, the distribution of original label samples for each clustering label and the importance ranking of the three fractal properties are obtained. The clustering results of the three fractal property values are shown in Figure 2.

A4. Use the fractal property value with the highest importance as the optimal fractal attribute, serving as the primary indicator for subsequent classification. Perform bidirectional maximization analysis on the distribution of original label samples corresponding to clustering

Specification ee LU603268 labels —calculating the proportion of different original label samples under each clustering label and the proportion of different clustering labels for each original sample label. By comprehensively considering all proportional distributions, identify the most reasonable one-to-one correspondence as the mapping relationship between clustering labels and final classification labels. Based on this mapping relationship, reclassify the clustering results, with the reclassified results serving as the final classification results. The correspondence between clustering labels and original classification labels is shown in Figure 3, and the final classification labels derived from the mapping relationship are shown in

Figure 5.

AS. Apply the distribution range of the optimal fractal attribute corresponding to the final classification labels to all T, main fractal attributes of the target well to obtain the final classification results.

Import the final classification results into well-logging curves. Compared with the original classification results, it is found that the classification results from this model provide a more thorough and reasonable explanation for well-testing results. A comparison between the original classification results and the classification results of this invention is illustrated in Figure 6. In Figure 6, Track 1 shows the original classification results, Track 2 shows the classification results of this invention, and Track 3 shows the well-testing interpretation results. 16

Specification

LU603268

The above descriptions are merely preferred embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention shall be included within the protection scope of the present invention. 17

Claims (8)

Claims LU603268 Claims

1. A method for intelligently identifying sweet spot types in shale reservoirs based on fractal characteristics, characterized by comprising the following steps: Step 1: Acquire and analyze the original T. spectrum data of the target well depth using nuclear magnetic resonance logging technology then calculate relevant fractal attribute values from the processed data to obtain fractal attribute indicators for each set of T2 spectra; Step 2: Obtain movable porosity data of the target well and classify it into four categories as original classification labels. Combine the fractal attribute indicators with the original classification labels into a data table, then input the data table into a K-means clustering model. Use machine learning methods to cluster the fractal attribute values and derive corresponding new clustering labels; Step 3: Analyze the importance of fractal attribute indicators using the random forest method, then select the fractal attribute with the highest importance from the analysis results as the final classification criterion,

1.e., the optimal fractal attribute value; Step 4: Based on the correspondence between the original classification labels and the new clustering labels, calculate the proportion of each type within each clustering label and the proportion of 1

Claims LU603268 each type across all clustering labels. Comprehensively analyze these proportions to establish a mapping relationship between clustering labels and original classification labels, and derive the final label for each set of data according to this mapping relationship; Step 5: Based on the final labels and optimal fractal attribute values, determine the value ranges of optimal fractal attributes for each of the four label categories. Use these ranges as the final classification criteria and apply them to all optimal fractal attribute values of the target well, obtaining a new shale reservoir classification evaluation result primarily based on optimal fractal attributes.

2. The method for intelligently identifying sweet spot types in shale reservoirs based on fractal characteristics according to claim 1, characterized in that: in step 1, the specific steps for analyzing and processing the original data of the target well's T, spectrum are: performing dimensionless processing on the original T, spectrum data using a dimensionless formula, ensuring all data converges within the [0,1] interval.

3. The method for intelligently identifying sweet spot types in shale reservoirs based on fractal characteristics according to claim 1, characterized in that: in step 1, the fractal dimension calculation method 1s used to compute relevant fractal attribute values of the processed data. The fractal attribute indicators of the T, spectrum include the singular 2

Claims LU603268 intensity, offset, and maximum-minimum offset values of the T: spectrum.

4. The method for intelligently identifying sweet spot types in shale reservoirs based on fractal characteristics according to claim 1, characterized in that: in step 2, the movable porosity data is obtained from the sealed coring data of the target well's interval, and the movable porosity data is divided into four categories in descending order.

5. The method for intelligently identifying sweet spot types in shale reservoirs based on fractal characteristics according to claim 1, characterized in that: in step 2, the compiled data table undergoes data preprocessing before being input into the K-means clustering model. The specific steps are: first, using data cleaning to remove missing values and outliers from the data table, then performing correlation analysis and dimensionality reduction on the data table, followed by standardizing the data table through normalization.

6. The method for intelligently identifying sweet spot types in shale reservoirs based on fractal characteristics according to claim 1, characterized in that: in step 3, the random forest method is used to analyze the importance of fractal attribute values, obtaining the ranking of influence levels of the three fractal attribute values on the clustering process, the influence level values of the three fractal attributes, and a histogram.

7. The method for intelligently identifying sweet spot types in shale 3

Claims LU603268 reservoirs based on fractal characteristics according to claim 1, characterized in that: in step 4, the bidirectional maximization method is used to calculate the proportion of each type within each cluster label and the proportion of each type across all cluster labels, with the specific steps being: based on analyzing the differences and connections between cluster labels and original classification labels, the bidirectional maximization method is employed to obtain the proportion of each original label sample within each cluster label, as well as the proportion of each original label type across all cluster labels. A comprehensive analysis yields the mapping relationship between cluster labels and original classification labels. According to this mapping relationship, the clustered data is reclassified to obtain the final classification labels.

8. The method for intelligently identifying sweet spot types in shale reservoirs based on fractal characteristics according to claim 7, characterized in that: the mapping relationship is: {cluster label 0: original label 1, cluster label 1: original label 2, cluster label 2: original label 4, cluster label 3: original label 3}. 4