A segmentation and clustering method for volumetric fracturing horizontal wells based on machine learning

By using fuzzy clustering algorithm to analyze the correlation between reservoir physical properties parameters and gas production profile data in volume fracturing horizontal well engineering, the precise segmentation and clustering of long horizontal segments is achieved, and the complexity of reservoir physical properties parameters is solved. The division difficulties brought about have improved the output and economic benefits of oil and gas wells.

CN118051847BActive Publication Date: 2024-11-01CHENGDU UNIV OF INFORMATION TECH
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202410189744.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-02-20
Publication Date
2024-11-01
Estimated Expiration
2044-02-20

AI Technical Summary

Technical Problem

In volume fracturing horizontal well engineering, how to achieve accurate segmentation and clustering of long horizontal sections to improve the output and economic benefits of oil and gas wells is affected by the diversity and complexity of reservoir physical properties parameters, the existing technology has not yet been formed Mature solutions.

Method used

Using a fuzzy clustering algorithm based on machine learning, combining the reservoir physical properties parameters and gas production profile data, calculate the correlation between physical properties parameters and yield data, select significant correlation parameters for fuzzy clustering analysis, and divide reservoirs along the well depth. Physical properties are classified and segmented and clustered in combination with constraints.

Benefits of technology

The precise classification and efficient segmentation and clustering of reservoir properties have been achieved, the perforation layout strategy has been optimized, the development level and economic benefits of oil and gas fields have been improved, and the fracturing effect of volume fracturing horizontal wells and single-well oil and gas output have been improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118051847B_ABST
    Figure CN118051847B_ABST
Patent Text Reader

Abstract

The invention discloses a volume fracturing horizontal well segmentation and clustering method based on machine learning, comprising: selecting wells that have carried out production logging and have gas production profile data of different well sections from constructed wells with similar reservoir physical property parameters to the well to be constructed as reference samples; using a fuzzy clustering algorithm to combine the reservoir physical property parameters and the gas production profile data of the constructed wells to calculate the correlation between the physical property parameters and the production data; based on the correlation between the physical property parameters and the production data, selecting reservoir physical property parameters with significant correlation with the production; using the fuzzy clustering algorithm to perform fuzzy clustering analysis on corresponding reservoir physical property parameter values ​​distributed along different well depths in the well to be constructed, and obtaining reservoir physical property classification weights along the well depth; determining the classification of reservoirs at each well depth according to the obtained reservoir physical property classification weights along the well depth, and performing segmentation and clustering division of the reservoir in combination with the reservoir classification result and constraint conditions.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of drilling and completion engineering in oil and natural gas, and in particular to a volume fracturing horizontal well segmentation and clustering method based on machine learning. Background Art

[0002] With the rapid development of social economy, the demand for the exploitation of unconventional oil and gas resources (such as shale gas and tight oil) is becoming increasingly urgent. In the process of exploiting these resources, volume fracturing technology has become one of the key technologies to increase the oil and gas production of a single well. Volume fracturing technology significantly enhances the flow performance of oil and gas around the wellbore by creating a complex fracture network in the reservoir, thereby increasing the production capacity of a single well. Horizontal well drilling technology, as the preferred supporting technology for volume fracturing, enables the wellbore to be drilled closely following the direction of the reservoir, thereby greatly increasing the contact area between the wellbore and the reservoir, providing strong support for the efficient exploitation of oil and gas resources. However, to achieve the best effect of horizontal well volume fracturing technology, the premise is to accurately segment and cluster the long horizontal section. This division needs to be based on the various attributes of the drilled reservoir. First, the reservoir is divided into different sections according to the attributes, and then each section is further subdivided into clusters according to the distribution characteristics of the attributes. The accuracy of segmentation and clustering is crucial to the pertinence of subsequent construction measures and the production of oil and gas wells. However, in actual operation, due to the diversity and complexity of reservoir physical parameters, how to select appropriate parameters and how to combine them to achieve effective division of long horizontal sections has become a technical problem. This not only affects the refined implementation of construction measures, but is also directly related to the production and economic benefits of oil and gas wells. In recent years, the application of machine learning algorithms in the field of oil and gas engineering has gradually attracted attention, and such algorithms have provided new ideas for solving complex problems. Despite this, the volume fracturing horizontal well segmentation and clustering technology based on machine learning algorithms is still in the exploratory stage, and a mature and widely used technical solution has not yet been formed. Summary of the invention

[0003] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a method for segmenting and clustering volume fracturing horizontal wells based on machine learning, comprising the following steps:

[0004] Step 1: From the constructed wells with similar reservoir physical property parameters to the well to be constructed, select wells that have been subjected to production logging and have gas production profile data of different well sections as reference samples;

[0005] Step 2: Using fuzzy clustering algorithm to combine reservoir physical property parameters and gas production profile data of constructed wells, calculate the correlation between physical property parameters and production data;

[0006] Step 3: Based on the correlation between physical property parameters and production data, reservoir physical property parameters with significant correlation with production are selected; fuzzy clustering algorithm is used to perform fuzzy clustering analysis on the corresponding reservoir physical property parameter values ​​distributed along different well depths in the wells to be constructed, and reservoir physical property classification weights along the well depth are obtained;

[0007] Step 4: Determine the classification of reservoirs at each well depth based on the obtained reservoir physical property classification weights along the well depth, and divide the reservoirs into segments and clusters based on the reservoir classification results and constraint conditions.

[0008] Furthermore, the reservoir physical property parameters include gamma, sonic, density, and porosity parameters obtained by logging, and parameters that can indirectly reflect reservoir characteristics calculated using the obtained gamma, sonic, density, and porosity parameters.

[0009] Furthermore, the production data includes differential production, intermediate production and high production. A fuzzy clustering algorithm is used in combination with the gas production profile data of the constructed wells to establish a membership function. The membership function accepts three parameters: intermediate production, maximum production and production to be classified. The production to be classified is the production that needs to be calculated through the membership function to correspond to the membership of the three production categories.

[0010] Furthermore, the membership function uses the following steps to calculate the membership of the output to be classified:

[0011] Step 1: Determine the production range: Calculate the intermediate production and maximum production based on the gas production profile data of the constructed wells;

[0012] Step 2: Construct membership function: For poor yield, construct a linear membership function interp1, which takes the value of 1 when the yield is 0, indicating that it completely belongs to the poor yield category; as the yield increases, the function value gradually decreases, and takes the value of 0 when the yield is two-thirds of the middle yield, indicating that it does not belong to the poor yield category at all; for middle yield, construct another linear membership function interp2, the value range of which starts from 0 and increases to 1 when the yield is two-thirds of the middle yield, indicating that it completely belongs to the middle yield category; then maintain this value until the yield increases to one-third of the difference between the middle yield and the maximum yield, and then linearly decreases to 0, indicating that it gradually does not belong to the middle yield category; for high yield, construct a third linear membership function interp3, which takes the value of 0 when the yield is one-third of the difference between the middle yield and the maximum yield, indicating that it does not belong to the high yield category at all; as the yield increases, the function value gradually increases, and increases to 1 when the yield is two-thirds of the poor yield plus the middle yield, indicating that it completely belongs to the high yield category, and maintains this value until the maximum yield;

[0013] Step 3: Calculate the degree of membership: According to the input output to be classified, determine the category it belongs to and calculate the corresponding degree of membership; if the output to be classified is less than or equal to two-thirds of the middle output, calculate the values ​​of interp1 and interp2 at this output, and the value of interp3 is 0; if the output to be classified is greater than two-thirds of the middle output and less than or equal to one-third of the difference between the middle output and the maximum output plus the middle output, the value of interp1 is 0, the value of interp2 is 1, and the value of interp3 is 0; if the output to be classified is greater than one-third of the difference between the middle output and the maximum output plus the middle output and less than or equal to two-thirds of the difference output plus the middle output, calculate the values ​​of interp2 and interp3 at this output, and the value of interp1 is 0; if the output to be classified is greater than two-thirds of the difference output plus the middle output and less than or equal to the maximum output, the values ​​of interp1 and interp2 are 0, and the value of interp3 is 1;

[0014] Step 4: Output result: The function returns a list containing three membership values, which correspond to the membership of the three categories of poor yield, medium yield and high yield respectively; by comparing these three values, the classification of the yield to be classified in different categories is determined.

[0015] Furthermore, the fuzzy clustering algorithm is used to combine the reservoir physical property parameters and gas production profile data of the constructed wells to calculate the correlation between the physical property parameters and the production data, including the following steps:

[0016] S1: Initialize the list of attributeAccumulationList to store the accumulated values ​​of reservoir physical parameters under poor production, intermediate production and high production. Initially, all elements of the list are set to 0;

[0017] S2: Calculate membership and accumulate physical property parameters: For each production data point in the gas production profile, call the membership function to calculate the membership of the three corresponding production categories;

[0018] S3: multiplying the calculated membership degree by the reservoir physical property parameter value at the corresponding well depth of the production data point;

[0019] S4: adding the product result to the element of the corresponding production category in attributeAccumulationList, where attributeAccumulationList contains the accumulated values ​​of reservoir physical property parameters under three production categories;

[0020] S5: normalize the accumulated values ​​in attributeAccumulationList. First, find the minimum and maximum values ​​in the list, map the accumulated values ​​to the range of 0 to 1 using a linear interpolation function, and generate a new list attributeNormalizationList.

[0021] S6: Calculate the absolute value of the sum of the differences between adjacent elements in attributeNormalizationList and express it as a percentage to obtain the correlation coefficient of the reservoir property change trend with production.

[0022] Furthermore, after calculating the correlation coefficient, a specific threshold is set and the reservoir properties that meet the threshold requirements are screened out.

[0023] In the well to be constructed, according to the selected reservoir physical properties, the reservoir physical property parameters of the well are selected and fuzzy cluster analysis is performed, which is divided into 3 categories. The fuzzy C-means algorithm is used for analysis, which is completed in the following steps:

[0024] Step 11: Initialization: Set the number of clusters to 3 and randomly initialize the cluster centers;

[0025] Step 12: Membership calculation: For each reservoir property data point, calculate its membership to each cluster. The membership is calculated based on the distance from the data point to the cluster center and the fuzzy factor;

[0026] Step 13: Update cluster centers: recalculate the center of each cluster based on the degree of membership of each data point;

[0027] Step 14: Iterative optimization: Repeat the above membership calculation and cluster center update steps until the change of the cluster center is less than the preset convergence threshold or the maximum number of iterations is reached.

[0028] Furthermore, the fuzzy cluster analysis obtains weight values ​​of three classifications of reservoir properties along the well depth, and the classifications are represented by numbers 0, 1, and 2 respectively. The classification number corresponding to the maximum weight value is used as the classification of the reservoir at the well depth, and the classification of reservoir properties distributed along the well depth is obtained. According to the classification of reservoir properties distributed along the well depth of the well to be constructed, the horizontal well section is divided in combination with the segmentation and clustering restriction conditions.

[0029] The beneficial effects of the present invention are: by introducing machine learning algorithms, especially fuzzy clustering algorithms, accurate classification and efficient segmentation and clustering of reservoir physical properties are achieved; limiting conditions such as casing collar positions are fully considered to avoid potential interference and risks; and the perforation arrangement strategy is optimized to improve the development level and economic benefits of oil and gas fields. The present invention is of great significance for improving the fracturing effect of volume fracturing horizontal wells and the oil and gas production of single wells. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 A flow chart of a method for segmenting and clustering volumetric fractured horizontal wells based on machine learning;

[0031] Figure 2 Schematic diagram of the analysis process involved in the method of the present invention;

[0032] Figure 3 It is a schematic diagram of the process of sequentially segmenting and clustering the horizontal well sections involved in the method of the present invention;

[0033] Figure 4 A schematic diagram of the distribution of fuzzy clustering of wells to be constructed along the depth of the well involved in the method embodiment of the present invention;

[0034] Figure 5 It is a schematic diagram of the division of perforation clusters of the well to be constructed involved in the method embodiment of the present invention. DETAILED DESCRIPTION

[0035] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings, but the protection scope of the present invention is not limited to the following.

[0036] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is further described in detail in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention, that is, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0037] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative work are within the scope of protection of the present invention. It should be noted that relational terms such as the terms "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0038] Moreover, the terms "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus that includes a list of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or apparatus. In the absence of more restrictions, an element defined by the phrase "comprises a ..." does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0039] The features and performance of the present invention are further described in detail below in conjunction with the embodiments.

[0040] like Figure 1 As shown, a method for segmenting and clustering volume fracturing horizontal wells based on machine learning is characterized by comprising the following steps:

[0041] Step 1: From the constructed wells with similar reservoir physical property parameters to the well to be constructed, select wells that have been subjected to production logging and have gas production profile data of different well sections as reference samples;

[0042] Step 2: Using fuzzy clustering algorithm to combine reservoir physical property parameters and gas production profile data of constructed wells, calculate the correlation between physical property parameters and production data;

[0043] Step 3: Based on the correlation between physical property parameters and production data, reservoir physical property parameters with significant correlation with production are selected; fuzzy clustering algorithm is used to perform fuzzy clustering analysis on the corresponding reservoir physical property parameter values ​​distributed along different well depths in the wells to be constructed, and reservoir physical property classification weights along the well depth are obtained;

[0044] Step 4: Determine the classification of reservoirs at each well depth based on the obtained reservoir physical property classification weights along the well depth, and divide the reservoirs into segments and clusters based on the reservoir classification results and constraint conditions.

[0045] The reservoir physical property parameters include gamma, sonic, density, porosity parameters obtained by logging, and parameters that can indirectly reflect reservoir characteristics calculated by the obtained gamma, sonic, density, porosity parameters.

[0046] The production data includes differential production, intermediate production and high production. A fuzzy clustering algorithm is used in combination with the gas production profile data of constructed wells to establish a membership function. The membership function accepts three parameters: intermediate production, maximum production and production to be classified. The production to be classified is the production that needs to be calculated through the membership function to correspond to the membership of the three production categories.

[0047] The membership function uses the following steps to calculate the membership of the output to be classified:

[0048] Step 1: Determine the production range: Calculate the intermediate production and maximum production based on the gas production profile data of the constructed wells;

[0049] Step 2: Construct membership function: For poor yield, construct a linear membership function interp1, which takes the value of 1 when the yield is 0, indicating that it completely belongs to the poor yield category; as the yield increases, the function value gradually decreases, and takes the value of 0 when the yield is two-thirds of the middle yield, indicating that it does not belong to the poor yield category at all; for middle yield, construct another linear membership function interp2, the value range of which starts from 0 and increases to 1 when the yield is two-thirds of the middle yield, indicating that it completely belongs to the middle yield category; then maintain this value until the yield increases to one-third of the difference between the middle yield and the maximum yield, and then linearly decreases to 0, indicating that it gradually does not belong to the middle yield category; for high yield, construct a third linear membership function interp3, which takes the value of 0 when the yield is one-third of the difference between the middle yield and the maximum yield, indicating that it does not belong to the high yield category at all; as the yield increases, the function value gradually increases, and increases to 1 when the yield is two-thirds of the poor yield plus the middle yield, indicating that it completely belongs to the high yield category, and maintains this value until the maximum yield;

[0050] Step 3: Calculate the degree of membership: According to the input output to be classified, determine the category it belongs to and calculate the corresponding degree of membership; if the output to be classified is less than or equal to two-thirds of the middle output, calculate the values ​​of interp1 and interp2 at this output, and the value of interp3 is 0; if the output to be classified is greater than two-thirds of the middle output and less than or equal to one-third of the difference between the middle output and the maximum output plus the middle output, the value of interp1 is 0, the value of interp2 is 1, and the value of interp3 is 0; if the output to be classified is greater than one-third of the difference between the middle output and the maximum output plus the middle output and less than or equal to two-thirds of the difference output plus the middle output, calculate the values ​​of interp2 and interp3 at this output, and the value of interp1 is 0; if the output to be classified is greater than two-thirds of the difference output plus the middle output and less than or equal to the maximum output, the values ​​of interp1 and interp2 are 0, and the value of interp3 is 1;

[0051] Step 4: Output result: The function returns a list containing three membership values, which correspond to the membership of the three categories of poor yield, medium yield and high yield respectively; by comparing these three values, the classification of the yield to be classified in different categories is determined.

[0052] The method of using a fuzzy clustering algorithm to combine reservoir physical property parameters and gas production profile data of constructed wells to calculate the correlation between physical property parameters and production data includes the following steps:

[0053] S1: Initialize the list of attributeAccumulationList to store the accumulated values ​​of reservoir physical parameters under poor production, intermediate production and high production. Initially, all elements of the list are set to 0;

[0054] S2: Calculate membership and accumulate physical property parameters: For each production data point in the gas production profile, call the membership function to calculate the membership of the three corresponding production categories;

[0055] S3: multiplying the calculated membership degree by the reservoir physical property parameter value at the corresponding well depth of the production data point;

[0056] S4: adding the product result to the element of the corresponding production category in attributeAccumulationList, where attributeAccumulationList contains the accumulated values ​​of reservoir physical property parameters under three production categories;

[0057] S5: normalize the accumulated values ​​in attributeAccumulationList. First, find the minimum and maximum values ​​in the list, map the accumulated values ​​to the range of 0 to 1 using a linear interpolation function, and generate a new list attributeNormalizationList.

[0058] S6: Calculate the absolute value of the sum of the differences between adjacent elements in attributeNormalizationList and express it as a percentage to obtain the correlation coefficient of the reservoir property change trend with production.

[0059] After calculating the correlation coefficient, a specific threshold is set and the reservoir properties that meet the threshold requirements are screened out.

[0060] In the well to be constructed, according to the selected reservoir physical properties, the reservoir physical property parameters of the well are selected and fuzzy cluster analysis is performed, which is divided into 3 categories. The fuzzy C-means algorithm is used for analysis, which is completed in the following steps:

[0061] Step 11: Initialization: Set the number of clusters to 3 and randomly initialize the cluster centers;

[0062] Step 12: Membership calculation: For each reservoir property data point, calculate its membership to each cluster. The membership is calculated based on the distance from the data point to the cluster center and the fuzzy factor;

[0063] Step 13: Update cluster centers: recalculate the center of each cluster based on the degree of membership of each data point;

[0064] Step 14: Iterative optimization: Repeat the above membership calculation and cluster center update steps until the change of the cluster center is less than the preset convergence threshold or the maximum number of iterations is reached.

[0065] The fuzzy cluster analysis obtains the weight values ​​of the reservoir properties along the well depth in three categories, and the categories are represented by numbers 0, 1, and 2 respectively. The classification number corresponding to the maximum weight value is used as the classification of the reservoir at the well depth, and the reservoir physical property classification distributed along the well depth is obtained. According to the reservoir physical property classification distributed along the well depth of the well to be constructed, the horizontal well section is divided in combination with the segmentation and clustering restriction conditions.

[0066] Specifically, Figure 2 As shown in the figure, a volume fracturing horizontal well segmentation and clustering method based on machine learning algorithm is provided, so that operation engineers can accurately adjust the fracturing construction parameters according to the algorithm analysis results, ensuring that the various design parameters in the fracturing construction process can adapt to the changes in reservoir characteristics in a more targeted manner, thereby effectively improving the fracturing effect and increasing the oil and gas production of a single well.

[0067] To achieve the above-mentioned purpose, the technical solution adopted by the present invention is: based on a volume fracturing horizontal well segmentation clustering method based on a machine learning algorithm, first, from the constructed wells with similar reservoir properties to the wells to be constructed, wells that have carried out production logging and have gas production profile data of different well sections are screened out as reference samples. Then, the reservoir physical property parameters of the constructed wells are deeply combined with the gas production profile data using the fuzzy clustering algorithm in machine learning, and the correlation between the physical property parameters and the production data is calculated. Based on the correlation calculation results, reservoir physical property parameters with obvious correlation with production are selected as an important basis for segmentation and clustering of the wells to be constructed. Subsequently, the fuzzy clustering algorithm is used again, and based on the screened reservoir physical property parameters, the corresponding values ​​distributed along different well depths in the wells to be constructed are extracted, and fuzzy clustering analysis is performed. After completing the fuzzy clustering analysis, the reservoir physical property classification weights along the well depth are obtained, and the classification number corresponding to the maximum weight value is used as the classification of the reservoir at the well depth. Finally, the reservoir segmentation and clustering are performed in combination with the reservoir classification results and constraints. The analysis process is as follows: Figure 1 shown.

[0068] Furthermore, the present invention needs to record the reservoir physical properties and gas production profile data of the constructed wells, and the reservoir physical properties of the wells to be constructed. The reservoir physical properties referred to here include but are not limited to the parameters directly reflecting the physical properties of the reservoir such as gamma, sonic, density, porosity, etc. obtained by well logging, and also refer to the parameters calculated by these parameters that can indirectly reflect the reservoir characteristics.

[0069] Furthermore, the production is divided into three categories: poor production, intermediate production and high production. The fuzzy clustering algorithm is combined with the gas production profile data of the constructed wells to establish a membership function, which accepts three parameters: intermediate production, maximum production and production to be classified. The production to be classified refers to the production that needs to be calculated by the membership function to correspond to the membership of the three production categories. The membership function uses the following steps to calculate the membership of the production to be classified:

[0070] Step 1: Determine the production range: Calculate the intermediate production and maximum production based on the gas production profile data of the constructed wells.

[0071] Step 2: Construct membership function: For the first category (poor output), construct a linear membership function interp1. The function takes the value of 1 when the output is 0, indicating that it completely belongs to the poor output category; as the output increases, the function value gradually decreases, and takes the value of 0 when the output is two-thirds of the middle output, indicating that it does not belong to the poor output category at all. For the second category (middle output), construct another linear membership function interp2. The value range of this function starts from 0, increases to 1 when the output is two-thirds of the middle output, indicating that it completely belongs to the middle output category; then maintains this value until the output increases to one-third of the difference between the middle output and the maximum output, and then linearly decreases to 0, indicating that it gradually does not belong to the middle output category. For the third category (high output), construct a third linear membership function interp3. The function takes the value of 0 when the output is one-third of the difference between the middle output and the maximum output, indicating that it does not belong to the high-yield category at all; as the output increases, the function value gradually increases, and increases to 1 when the output is two-thirds of the difference output plus the middle output, indicating that it completely belongs to the high-yield category, and maintains this value until the maximum output.

[0072] Step 3: Calculate the degree of membership: According to the input output to be classified, determine the category to which it belongs, and calculate the corresponding degree of membership. The specific judgment logic is as follows: If the output to be classified is less than or equal to two-thirds of the middle output, calculate the values ​​of interp1 and interp2 at this output, and the value of interp3 is 0. If the output to be classified is greater than two-thirds of the middle output and less than or equal to one-third of the difference between the middle output and the maximum output plus the middle output, the value of interp1 is 0, the value of interp2 is 1, and the value of interp3 is 0. If the output to be classified is greater than one-third of the difference between the middle output and the maximum output plus the middle output and less than or equal to two-thirds of the difference output plus the middle output, calculate the values ​​of interp2 and interp3 at this output, and the value of interp1 is 0. If the output to be classified is greater than two-thirds of the difference output plus the middle output and less than or equal to the maximum output, the values ​​of interp1 and interp2 are 0, and the value of interp3 is 1.

[0073] Step 4: Output: The function returns a list of three membership values, which correspond to the membership of the three categories of poor yield, medium yield, and high yield. By comparing these three values, we can determine the classification of the yield to be classified in different categories.

[0074] Furthermore, according to the reservoir property values ​​corresponding to the well depth of the gas production profile in the constructed well, a correlation coefficient reflecting the trend of reservoir property changes with production is calculated:

[0075] Step 1: Initialize a list named attributeAccumulationList to store the accumulated values ​​of reservoir physical parameters under three production categories (poor production, medium production, and high production). Initially, all elements of the list are set to 0.

[0076] Step 2: Calculate the membership and accumulate physical property parameters: For each production data point in the gas production profile, call the previously defined membership function to calculate its membership corresponding to the three production categories.

[0077] Step 3: Multiply the calculated membership degree by the reservoir physical property parameter value at the corresponding well depth of the production data point.

[0078] Step 4: Add the product result to the element of the corresponding production category in attributeAccumulationList. After this step is completed, attributeAccumulationList will contain the accumulated values ​​of reservoir physical property parameters under the three production categories, which reflect the physical property trends with production changes.

[0079] Step 5: In order to eliminate the influence of different physical property parameter units and dimensions, the accumulated values ​​in attributeAccumulationList need to be normalized. First, find the minimum and maximum values ​​in the list. Then, use the linear interpolation function to map the accumulated values ​​to the range of 0 to 1 and generate a new list attributeNormalizationList. This step ensures the comparability between different physical property parameters.

[0080] Step 6: Finally, calculate the absolute value of the sum of the differences between adjacent elements in attributeNormalizationList and express it as a percentage to obtain a correlation coefficient that reflects the trend of reservoir properties changing with production.

[0081] Furthermore, after calculating the correlation coefficient, a specific threshold is set, and reservoir properties that meet the threshold requirements are screened out. These screened reservoir properties will serve as the basis for segmenting and clustering the wells to be constructed.

[0082] Furthermore, in the well to be constructed, according to the previously screened reservoir properties, the reservoir properties parameters of the well are selected and fuzzy cluster analysis is performed to divide them into three categories. The analysis is performed using the Fuzzy C-Means (FCM) algorithm, which is completed in the following steps:

[0083] Step 1: Initialization: Set the number of clusters to 3 and randomly initialize the cluster centers.

[0084] Step 2: Membership calculation: For each reservoir property data point, calculate its membership to each cluster. The membership is calculated based on the distance from the data point to the cluster center and the fuzzy factor.

[0085] Step 3: Update cluster centers: Recalculate the center of each cluster based on the degree of membership of each data point.

[0086] Step 4: Iterative optimization: Repeat the above membership calculation and cluster center update steps until the change of the cluster center is less than the preset convergence threshold or the maximum number of iterations is reached.

[0087] Furthermore, fuzzy cluster analysis was used to obtain the weight values ​​of the three categories of reservoir properties along the well depth, and the categories were represented by numbers 0, 1, and 2. The classification number corresponding to the maximum weight value was used as the classification of the reservoir at the well depth, and the reservoir property classification distributed along the well depth could be obtained.

[0088] Furthermore, according to the reservoir physical property classification along the well depth distribution of the well to be constructed, the horizontal well section can be divided in combination with the segmentation and clustering restriction conditions. The following steps can be used to complete the analysis process. Figure 3 As shown:

[0089] The first part is the horizontal well section division step, including step 1 to step 4.

[0090] Step 1: Set segmentation parameters and clarify segmentation constraints.

[0091] (1) Determine the upper and lower limits of the segment length.

[0092] (2) Determine the maximum and minimum distances between segments.

[0093] (3) Determine the step value for increasing the segment length.

[0094] (4) Identify and mark the locations of casing couplings that must be avoided.

[0095] Step 2: Divide the first section starting from the bottom of the well.

[0096] (1) Select the starting position: start dividing from the bottom of the horizontal wellbore (the bottom of the well).

[0097] (2) Set the initial segment length: The lower limit of the segment length is used as the initial value of the first segment length.

[0098] (3) Determine the reservoir category: Based on the results of fuzzy cluster analysis, determine the reservoir physical property category with the highest proportion in the first section.

[0099] (4) Gradually increase the segment length: Gradually increase the segment length according to the step value of the segment length increase, and calculate the change in the proportion of reservoir physical property categories at the same time.

[0100] (5) Determine the stop condition: When the percentage of the reservoir physical property category with the highest proportion decreases or reaches the upper limit of the segment length, stop increasing the segment length and determine the current length as the final segment length of the first segment.

[0101] (6) Check the casing coupling: confirm whether the boundary position of the first length interval avoids the casing coupling. If necessary, adjust the section length appropriately according to the step value to ensure safety.

[0102] Step 3: Continuously divide subsequent well sections.

[0103] (1) Based on the determined minimum and maximum segment spacings, the starting position of the second segment is calculated with the minimum segment spacing as the starting value.

[0104] (2) If the starting position is a casing coupling, the starting position is adjusted according to the step value based on the minimum spacing, and the non-casing coupling position is found as the starting interval of the second segment.

[0105] (3) Follow the method in step 2 to divide the second segment from the new starting interval.

[0106] Step 4: Complete the division of the entire well section: Repeat step 3 until the entire horizontal well section is completely divided.

[0107] The second part is the intra-segment cluster division, including steps 5 and 6.

[0108] Step 5: Set cluster partitioning parameters.

[0109] (1) Determine the perforation length of each cluster.

[0110] (2) Set the maximum cluster spacing and the minimum cluster spacing.

[0111] (3) Set the maximum boundary distance and minimum boundary distance between the perforation point and the well section.

[0112] Step 6: Perform intra-cluster perforation placement.

[0113] (1) First, determine the main categories within the well section, then find the location of the first perforation cluster and continue to add more perforation clusters within the well section until the end of the well section is reached or no more perforation clusters can be added.

[0114] (2) When adding the first perforation cluster, find the corresponding position of the reservoir property category with the highest proportion of well depth in this section between the minimum and maximum boundary distances. This position is the first perforation cluster.

[0115] (3) When adding subsequent perforation clusters, find the corresponding position of the reservoir property category with the highest proportion of the well depth in this section between the minimum and maximum cluster spacings. This position is the perforation cluster.

[0116] (4) Repeat the steps in (3) and add perforation clusters in sequence until the position of the perforation cluster plus the minimum boundary distance exceeds the well section boundary.

[0117] The embodiment provides a method for segmenting and clustering volume fracturing horizontal wells based on a machine learning algorithm, and the specific steps are as follows:

[0118] Step 1: Screen reference samples. From the constructed wells with similar reservoir properties to the well to be constructed, select wells that have been subjected to production logging and have gas production profile data of different well sections. These wells will serve as reference samples for analysis.

[0119] Step 2: Calculate the correlation coefficient between the reservoir properties and production change trends in the well to be constructed.

[0120] (1) Initialize a list to store the accumulated values ​​of reservoir physical parameters under three production categories (poor production, intermediate production, and high production).

[0121] (2) For each production data point, calculate its membership degree to the three production categories and multiply it by the reservoir physical property parameter value at the corresponding well depth.

[0122] (3) Accumulate the product results to the elements of the corresponding output category in the list.

[0123] (4) Normalize the accumulated values ​​to eliminate the influence of different units and dimensions of physical property parameters.

[0124] (5) Calculate the absolute value of the sum of the differences between adjacent elements in the normalized list and convert it into a percentage to obtain the correlation coefficient of the reservoir property change trend with production.

[0125] Step 3: Screen reservoir properties.

[0126] (1) Set a threshold and select reservoir properties that meet the threshold requirements.

[0127] (2) These selected reservoir properties will serve as the basis for segmenting and clustering the wells to be constructed.

[0128] Step 4: Conduct fuzzy cluster analysis on the wells to be constructed.

[0129] (1) Extract the reservoir physical property parameter values ​​distributed along the depth of the well to be constructed.

[0130] (2) Based on the selected reservoir physical property parameters, fuzzy cluster analysis is performed. The number of categories is set to 3 (the numbers 0, 1, and 2 are used to represent the three categories, respectively). After the calculation is completed, the weight values ​​of the three categories corresponding to different well depths are obtained. Then, the category with the highest weight is used as the classification of reservoir physical properties at the well depth, and the reservoir physical property categories distributed along the well depth in the well to be constructed can be obtained.

[0131] Step 5: Divide the well into horizontal sections.

[0132] (1) Set segment parameters, including segment length increase, segment length upper and lower limits, maximum and minimum segment spacing, and casing coupling positions that must be avoided.

[0133] (2) Divide the first section from the bottom of the well, gradually increase the section length according to the section length increase value, and calculate the change in the proportion of reservoir physical property categories.

[0134] (3) When the maximum proportion category value of the reservoir physical property decreases or reaches the upper limit of the segment length, stop increasing the length and determine the current length as the final length of the first segment.

[0135] (4) Continuously divide the subsequent well sections until the entire horizontal well section is completely divided.

[0136] Step 6: Intra-segment clustering.

[0137] (1) Setting cluster division parameters: including the length of each cluster perforation, the maximum and minimum spacing between clusters, and the maximum and minimum distance between the perforation point and the well section boundary.

[0138] (2) Initialize the perforation arrangement within the cluster: Before starting the perforation cluster arrangement, first identify the main reservoir property categories in the well section. Then, under the premise of meeting the boundary distance requirements, find the location corresponding to the reservoir property category with the highest well depth ratio as the arrangement point of the first perforation cluster.

[0139] (3) Gradually adding perforation clusters: After determining the first perforation cluster, continue to search for the location corresponding to the reservoir property category with the highest well depth ratio in the well section according to the set cluster spacing range, and arrange subsequent perforation clusters.

[0140] (4) Complete the perforation cluster layout: Repeat the previous step and continue to add perforation clusters until the end of the well section is reached or no more perforation clusters can be added. During the layout process, check the distance between the perforation cluster and the well section boundary to ensure that it does not exceed the set minimum boundary distance.

[0141] The embodiments are further described below in conjunction with specific examples.

[0142] Step 1: Record the well depth, well inclination, reservoir physical parameters and corresponding gas production of the constructed wells. As shown in Table 1, it includes well depth, well inclination, resistivity, gamma, density, porosity, permeability and gas production.

[0143] Table 1 Parameters of constructed wells

[0144]

[0145]

[0146] Step 2: According to the data in Table 1, the correlations between well inclination angle, resistivity, gamma, density, porosity, permeability and gas production are calculated to be 61.2%, 46.1%, 74.7%, 70.4%, 69.6% and 94.3% respectively.

[0147] Step 3: Based on the calculation results of the second step, 50% is selected as the limit. It can be concluded that well inclination, gamma, density, porosity, permeability and gas production have a high correlation. These parameters are selected as the basic data for the analysis of the wells to be constructed.

[0148] Step 4: Select the existing parameters of the wells to be constructed and meet the gas production correlation requirements to analyze the wells to be constructed. The data of the wells to be constructed are shown in Table 2. Fuzzy clustering analysis is performed on the well inclination, gamma, density, and porosity data. Three categories are selected for fuzzy clustering. The first category is represented by 0, the second category is represented by 1, and the third category is represented by 2. The results are shown in Figure 3 .

[0149] Table 2 Parameters of wells to be constructed

[0150]

[0151]

[0152] Figure 4 Schematic diagram of the distribution of fuzzy clustering of wells to be constructed along the depth of the well involved in the embodiment of the method of the present invention

[0153] Step 5: Divide the horizontal sections, where the positions of the casing couplings of the wells to be constructed are shown in Table 3. The upper and lower limits of the section length are set to 60m and 40m respectively, the maximum and minimum section spacing are set to 5m and 0m respectively, and the section length increase is 0.5m. The calculated sections are shown in Table 4.

[0154] Table 3 Casing couplings for wells to be constructed

[0155]

[0156]

[0157] Table 4 Sectional table of wells to be constructed

[0158]

[0159] Step 6: Clustering within the segment. Set the length of each cluster perforation to 0.5 m, the maximum and minimum spacing between clusters to 11 m and 8.5 m, and the maximum and minimum distances from the perforation point to the well segment boundary to 4.5 m and 1.5 m, respectively. The cluster data within the segment can be calculated. Figure 5 .

[0160] The above is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art shall not deviate from the spirit and scope of the present invention, and shall be within the scope of protection of the claims attached to the present invention.

Claims

1. A method for segmenting and clustering volume fracturing horizontal wells based on machine learning, characterized in that: The steps include: Step 1: From the constructed wells with similar reservoir physical property parameters to the well to be constructed, select wells that have been subjected to production logging and have gas production profile data of different well sections as reference samples; Step 2: Using fuzzy clustering algorithm to combine reservoir physical property parameters and gas production profile data of constructed wells, calculate the correlation between physical property parameters and production data; Step 3: Based on the correlation between physical property parameters and production data, reservoir physical property parameters with significant correlation with production are selected; fuzzy clustering algorithm is used to perform fuzzy clustering analysis on the corresponding reservoir physical property parameter values ​​distributed along different well depths in the wells to be constructed, and reservoir physical property classification weights along the well depth are obtained; Step 4: Determine the classification of reservoirs at each well depth according to the obtained reservoir physical property classification weights along the well depth, and divide the reservoirs into segments and clusters based on the reservoir classification results and constraint conditions; The method of using a fuzzy clustering algorithm to combine reservoir physical property parameters and gas production profile data of constructed wells to calculate the correlation between physical property parameters and production data includes the following steps: S1: Initialize the list of attributeAccumulationList to store the accumulated values ​​of reservoir physical parameters under poor production, intermediate production and high production. Initially, all elements of the list are set to 0; S2: Calculate membership and accumulate physical property parameters: For each production data point in the gas production profile, call the membership function to calculate its membership of the three corresponding production categories; S3: multiplying the calculated membership degree by the reservoir physical property parameter value at the corresponding well depth of the production data point; S4: adding the product result to the element of the corresponding production category in attributeAccumulationList, where attributeAccumulationList contains the accumulated values ​​of reservoir physical property parameters under three production categories; S5: normalize the accumulated values ​​in attributeAccumulationList. First, find the minimum and maximum values ​​in the list, map the accumulated values ​​to the range of 0 to 1 using a linear interpolation function, and generate a new list attributeNormalizationList. S6: Calculate the absolute value of the sum of the differences between adjacent elements in attributeNormalizationList and express it as a percentage to obtain the correlation coefficient of the reservoir property with the production change trend; After calculating the correlation coefficient, a threshold is set and the reservoir properties that meet the threshold are screened out. In the well to be constructed, according to the selected reservoir physical properties, the reservoir physical property parameters of the well are selected and fuzzy cluster analysis is performed, which is divided into 3 categories. The fuzzy C-means algorithm is used for analysis, which is completed in the following steps: Step 11: Initialization: Set the number of clusters to 3 and randomly initialize the cluster centers; Step 12: Membership calculation: For each reservoir property data point, calculate its membership to each cluster; The membership degree is calculated based on the distance from the data point to the cluster center and the fuzzy factor; Step 13: Update cluster centers: recalculate the center of each cluster based on the degree of membership of each data point; Step 14: Iterative optimization: Repeat the above membership calculation and cluster center update steps until the change of the cluster center is less than the preset convergence threshold or the maximum number of iterations is reached.

2. The method for segmenting and clustering volume fracturing horizontal wells based on machine learning according to claim 1, characterized in that: The reservoir physical property parameters include gamma, sonic, density, porosity parameters obtained by logging, and parameters that can indirectly reflect reservoir characteristics calculated by the obtained gamma, sonic, density, porosity parameters.

3. The method for segmenting and clustering volume fracturing horizontal wells based on machine learning according to claim 2, characterized in that: The production data includes differential production, intermediate production and high production. A fuzzy clustering algorithm is used in combination with the gas production profile data of constructed wells to establish a membership function. The membership function accepts three parameters: intermediate production, maximum production and production to be classified. The production to be classified is the production that needs to be calculated through the membership function to correspond to the membership of the three production categories.

4. The method for segmenting and clustering volume fracturing horizontal wells based on machine learning according to claim 3, characterized in that: The membership function uses the following steps to calculate the membership of the output to be classified: Step 1: Determine the production range: Calculate the intermediate production and maximum production based on the gas production profile data of the constructed wells; Step 2: Construct membership function: For poor yield, construct a linear membership function interp1, which takes the value of 1 when the yield is 0, indicating that it completely belongs to the poor yield category; as the yield increases, the function value gradually decreases, and takes the value of 0 when the yield is two-thirds of the middle yield, indicating that it does not belong to the poor yield category at all; for middle yield, construct another linear membership function interp2, the value range of which starts from 0 and increases to 1 when the yield is two-thirds of the middle yield, indicating that it completely belongs to the middle yield category; then maintain this value until the yield increases to one-third of the difference between the middle yield and the maximum yield, and then linearly decreases to 0, indicating that it gradually does not belong to the middle yield category; for high yield, construct a third linear membership function interp3, which takes the value of 0 when the yield is one-third of the difference between the middle yield and the maximum yield, indicating that it does not belong to the high yield category at all; as the yield increases, the function value gradually increases, and increases to 1 when the yield is two-thirds of the poor yield plus the middle yield, indicating that it completely belongs to the high yield category, and maintains this value until the maximum yield; Step 3: Calculate the degree of membership: According to the input output to be classified, determine the category it belongs to and calculate the corresponding degree of membership; if the output to be classified is less than or equal to two-thirds of the middle output, calculate the values ​​of interp1 and interp2 at this output, and the value of interp3 is 0; if the output to be classified is greater than two-thirds of the middle output and less than or equal to one-third of the difference between the middle output and the maximum output plus the middle output, the value of interp1 is 0, the value of interp2 is 1, and the value of interp3 is 0; if the output to be classified is greater than one-third of the difference between the middle output and the maximum output plus the middle output and less than or equal to two-thirds of the difference output plus the middle output, calculate the values ​​of interp2 and interp3 at this output, and the value of interp1 is 0; if the output to be classified is greater than two-thirds of the difference output plus the middle output and less than or equal to the maximum output, the values ​​of interp1 and interp2 are 0, and the value of interp3 is 1; Step 4: Output result: The function returns a list containing three membership values, which correspond to the membership of the three categories of poor yield, medium yield and high yield respectively; By comparing these three values, the classification of the output to be classified in different categories is determined.

5. The method for segmenting and clustering volume fracturing horizontal wells based on machine learning according to claim 1, characterized in that: The fuzzy cluster analysis obtains the weight values ​​of the reservoir properties along the well depth in three categories, and the categories are represented by numbers 0, 1, and 2 respectively. The classification number corresponding to the maximum weight value is used as the classification of the reservoir at the well depth, and the reservoir physical property classification distributed along the well depth is obtained. According to the reservoir physical property classification distributed along the well depth of the well to be constructed, the horizontal well section is divided in combination with the segmentation and clustering restriction conditions.

Citation Information

Patent Citations

  • Volume fracturing horizontal well segmented clustering method and device

    CN109977586A

  • Horizontal well multistage-fracturing segmenting and clustering method based on logging curve

    CN111456709A

  • Combination optimization method for unconventional reservoir fracturing perforation segment cluster parameters

    CN114880957A

  • Intelligent segmentation and clustering method for fracturing grades of horizontal well

    CN116025324A

  • Differentiation evaluation method for heterogeneity of unconventional reservoir fractured horizontal well

    CN116484236A