Oil reservoir dynamic monitoring well pattern intelligent optimization method based on machine learning

Abstract

The invention provides an oil reservoir dynamic monitoring well pattern intelligent optimization method based on machine learning. The oil reservoir dynamic monitoring well pattern intelligent optimization method based on machine learning comprises the following steps: step 1, obtaining production monitoring data of a well; step 2, taking each piece of monitoring item data as a dimension to obtain coordinates of each well in a feature space; step 3, filtering the dimensions to obtain an optimized feature space; 4, k-means clustering is carried out, and when the sum of squares of errors is minimum, a clustering center is obtained; and 5, changing the value of K, and determining the minimum number and the corresponding well positions of the monitoring wells and the optimal number and the corresponding well positions of the monitoring wells. The oil reservoir dynamic monitoring well pattern intelligent optimization method based on machine learning can break through the limitation that an existing monitoring well pattern depends on subjective experience of an engineer and an accurate geologic model, does not depend on the subjective experience and a percolation mechanism any more, and can efficiently and dynamically adjust the monitoring well pattern.

Description

Technical Field

[0001] The present invention relates to the technical field of oilfield development, and in particular to an intelligent optimization method for a reservoir dynamic monitoring well network based on machine learning. Background Art

[0002] After years of development, dynamic monitoring technology has gradually formed technologies in basic experiments, monitoring scheme design, monitoring data collection, data interpretation and comprehensive application, becoming another important means to understand and evaluate the dynamics and effects of reservoir development in addition to reservoir engineering and reservoir numerical simulation. The traditional reservoir engineering method based on the analytical solution of the seepage mechanics equation requires a significant simplification of the reservoir model and cannot fully consider the influence of reservoir heterogeneity and complex flow mechanisms. Reservoir numerical simulation is a conventional method for reservoir dynamic analysis and prediction. Its advantage is that the correlation between parameters based on the mechanism is clear, and after years of development, the method is mature and can be implemented based on commercial software. Its main limitation is that it relies on accurate geological models and seepage mechanisms. For blocks with unclear geological models or seepage mechanisms, the error is large, and only part of the monitoring data can be considered. The data-driven method uses machine learning and deep learning methods to directly fit the nonlinear relationship between data and perform corresponding analysis and prediction. It does not require a clear seepage mechanism or geological model, but it has high requirements for the quality of the collected data.

[0003] Data analysis based on machine learning mainly includes correlation analysis, denoising and dimensionality reduction, classification and regression, prediction, inversion and control, and has formed algorithms such as k-nearest neighbor method, support vector machine, logistic regression and neural network. How to combine machine learning algorithms with specific engineering practices and establish corresponding data-driven analysis technologies is the focus of industrial big data research. For reservoir monitoring data, establishing corresponding data analysis algorithms based on machine learning is conducive to fully exploring the characteristics of monitoring data and comprehensively considering the relationship between various dimensions of data, so that monitoring data can play a greater role in guiding the efficient development of oil fields, which is of great significance to the construction of intelligent oil and gas fields. The existing monitoring well network relies on the subjective experience of engineers and accurate geological models, and lacks a data-driven reservoir dynamic monitoring well network method.

[0004] In the Chinese patent application with application number: CN202210493119.1, a method for optimizing decision-making of oilfield production system based on forced learning is involved, including: collecting dynamic production data of oilfield production site, establishing data cube for reservoir production optimization; training the preset machine learning model based on the data cube, obtaining a reservoir injection and production system proxy model based on forced learning for predicting oil production according to dynamic production data of the production site; constructing an evaluation function for gas injection reservoir production optimization; establishing a forced constraint model and boundary constraint conditions based on input parameters during the production optimization process; using the constraint model and boundary constraint conditions as constraints, based on the reservoir injection and production system proxy model, and taking the evaluation function as the optimization direction, searching for reservoir production optimization solutions and obtaining the optimal production solutions. This invention can solve the technical problems that traditional physics-based methods have limitations in calculation accuracy and time.

[0005] In the Chinese patent application with application number: CN202010572648.1, a machine learning method for oil reservoir production based on a parallel proxy model is involved. The parallel oil field production optimization method based on the proxy model can simultaneously obtain multiple better candidate solutions in each iteration, and then use MATLAB to call the reservoir numerical simulation software Eclipse in parallel to conduct real evaluation on them at the same time, which can greatly reduce the optimization time of complex problems. Through the method of the present invention, the speed of oil field production optimization can be greatly accelerated, and the optimization efficiency and the final optimization effect can be improved. At the same time, in addition to adjusting the production system of oil wells and water wells in the oil field, the method of the invention can also be used to optimize the well network, historical fitting, etc.

[0006] In the Chinese patent application with application number: CN202010292889.0, a large-scale reservoir injection and production optimization method based on a machine learning proxy model is involved. The method is a solution to the shortcomings of traditional evolutionary algorithms in calling numerical simulations many times during optimization design. It combines the evolutionary algorithm framework of the proxy model and the dimensionality reduction algorithm, and uses the Salmon mapping as a dimensionality reduction algorithm to reduce the dimension of the original data. The calculation is simple and easy to implement. The method first constructs a mathematical model with a small amount of calculation and fast calculation speed through a small number of test samples under the premise of ensuring accuracy, but the calculation results are similar to those obtained by the numerical simulator. Based on the high-dimensional interpolation theory in the machine learning method, a fast-calculating mathematical model is used to approximate the target complex system and replace it to complete the optimization design. In order to improve the accuracy of the proxy model, an evolutionary algorithm is used to dynamically sample and update the proxy model, so as to obtain the optimal oil and water well production plan with a higher economic net present value, which has a good promotion and application value.

[0007] In the Chinese patent application with application number: CN202110028282.6, a data-driven water injection reservoir optimization method and system are involved, and the method includes: establishing a preliminary reservoir model based on static data and dynamic data collected in real time; randomly selecting historical production parameters about reservoir parameters to perform numerical simulation of the preliminary reservoir model; when the difference between the numerical simulation results of the reservoir model and the production history is less than a second threshold, using the preliminary reservoir model as the reservoir model for subsequent water injection and drainage optimization; when the difference between the numerical simulation results of the reservoir model and the production history is greater than the second threshold, selecting at least one first parameter with a correlation greater than the first threshold based on the correlation analysis between the dynamic data and the numerical simulation results, and adjusting the preliminary reservoir model by optimizing at least one of the first parameters so that the difference between the numerical simulation results and the production history is less than the second threshold.

[0008] The above existing technologies are significantly different from the present invention and fail to solve the technical problem we want to solve. For this reason, we have invented a new intelligent optimization method for reservoir dynamic monitoring well network based on machine learning. Summary of the invention

[0009] The purpose of the present invention is to provide an intelligent optimization method for reservoir dynamic monitoring well network based on machine learning, which breaks through the limitation that the existing monitoring well network relies on the subjective experience of engineers and accurate geological models.

[0010] The purpose of the present invention can be achieved by the following technical measures: an intelligent optimization method for oil reservoir dynamic monitoring well network based on machine learning, the intelligent optimization method for oil reservoir dynamic monitoring well network based on machine learning includes:

[0011] Step 1, obtaining production monitoring data of the well;

[0012] Step 2, taking each monitoring project data as a dimension, obtain the coordinates of each well in the feature space;

[0013] Step 3, filter the dimensions to obtain an optimized feature space;

[0014] Step 4: Perform k-means clustering and obtain the cluster center when the sum of squared errors is minimized;

[0015] Step 5, change the value of K to determine the minimum number of monitoring wells and corresponding well locations as well as the optimal number of monitoring wells and corresponding well locations.

[0016] The purpose of the present invention can also be achieved by the following technical measures:

[0017] In step 1, production monitoring data of a well is obtained, and the production monitoring data includes data of multiple monitoring items.

[0018] In step 1, the monitoring project data include wellhead temperature, bottom hole temperature, oil nozzle, wellhead pressure, bottom hole flow pressure, oil production, gas production, water production, liquid production, water content, pump frequency, pump current, pump voltage, pump leakage current, pump vibration, pump inlet temperature, pump motor temperature, pump inlet pressure and pump outlet pressure.

[0019] In step 3, the production monitoring data of all wells are spliced ​​into a whole, and the dimensions are filtered according to the overall statistical feature contribution rate based on principal component analysis to obtain an optimized feature space.

[0020] In step 3, the dimensions are filtered based on the principal component analysis according to the overall statistical feature contribution rate, including:

[0021] Step 31, determining the principal component according to the cumulative variance contribution rate;

[0022] Step 32: If the weight index of a dimension in each principal component is 0, then the dimension is eliminated.

[0023] In step 31, the components with cumulative variance contribution rates of more than 80% are taken as principal components; there are five principal components, which are represented as Y1, Y2, Y3, Y4, and Y5 respectively.

[0024] In step 31, the principal component Y1 comprehensively reflects the influence of wellhead temperature, bottom hole temperature, oil nozzle, wellhead pressure, bottom hole flowing pressure, oil production, water production, liquid production, water content and pump parameters on the development dynamics of the block;

[0025] The principal component Y2 reflects the negative effects of the wellhead-bottomhole temperature difference, the wellhead-bottomhole pressure difference, the pump leakage current, the pump vibration, the pump inlet temperature, the pump motor temperature, the pump inlet pressure, and the pump outlet pressure;

[0026] The principal component Y3 reflects the negative influence of oil production and the positive influence of water production and water content;

[0027] The main components Y4 and Y5 highlight the influence of gas production.

[0028] In step 4, based on the optimized feature space, K objects are randomly selected as initial cluster centers, and k-means clustering is performed. When the sum of squared errors is minimized, the cluster center is obtained and used as the well location for regular monitoring.

[0029] In step 4, the sum of squared errors is calculated using the following formula:

[0030]

[0031] Among them, C i is the i-th class, p is C i Well in, m i It is C i center.

[0032] In step 5, the value of K is changed to determine the minimum number of monitoring wells and the corresponding well locations, including:

[0033] Change the value of K and determine the sum of squared errors corresponding to the K value;

[0034] Cluster analysis is performed using the K value corresponding to the minimum sum of squared errors as the number of clusters to determine the minimum number of monitoring wells and the corresponding well locations.

[0035] In step 5, the value of K is changed to determine the optimal number of monitoring wells and the corresponding well locations, which specifically includes:

[0036] Change the value of K and determine the sum of squared errors corresponding to the K value;

[0037] Cluster analysis was performed using the K value corresponding to the error square sum less than 0.5 as the number of clusters to determine the optimal number of monitoring wells and the corresponding well locations.

[0038] The intelligent optimization method for dynamic reservoir monitoring well network based on machine learning also includes, after step 5, repeating steps 1-5 within a preset time interval to update the minimum number of monitoring wells and the optimal number of monitoring wells to achieve dynamic adjustment of the monitoring well network.

[0039] The intelligent optimization method for reservoir dynamic monitoring well network based on machine learning in the present invention can break through the limitation that the existing monitoring well network relies on the subjective experience of engineers and accurate geological models. Compared with the prior art, the present invention has at least the following beneficial effects:

[0040] 1. The present invention can break through the limitation that the existing monitoring well network relies on the subjective experience of engineers and accurate geological models. Based on the machine learning method, it can relatively accurately determine the regular monitoring well locations, the minimum / optimal number of monitoring wells and their corresponding well locations, and no longer rely on subjective experience and seepage mechanism.

[0041] 2. After the production data changes, the monitoring well network can be efficiently and dynamically adjusted based on the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a flowchart of a specific embodiment of the intelligent optimization method for reservoir dynamic monitoring well network based on machine learning of the present invention;

[0043] Figure 2 is a graph showing the relationship between the sum of squared errors (SSE) and the number of clusters k in Example 1 of the present invention;

[0044] Figure 3 This is a monitoring well network distribution diagram of Example 1 of the present invention (the number of clusters k=3);

[0045] Figure 4 The monitoring well network distribution diagram of Example 1 of the present invention (the number of clusters k=5);

[0046] Figure 5 is a graph showing the relationship between the sum of squared errors (SSE) and the number of clusters k in Example 2 of the present invention;

[0047] Figure 6 This is a monitoring well network distribution diagram of Example 2 of the present invention (the number of clusters k=3);

[0048] Figure 7 is a graph showing the relationship between the sum of squared errors (SSE) and the number of clusters k in Example 3 of the present invention;

[0049] Figure 8 This is the monitoring well network distribution diagram of Example 3 of the present invention (the number of clusters k=6). DETAILED DESCRIPTION

[0050] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0051] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations and / or combinations thereof.

[0052] The intelligent optimization method for reservoir dynamic monitoring well pattern based on machine learning of the present invention comprises:

[0053] Step 1, obtain the production monitoring data of the well, and the production monitoring data includes multiple monitoring item data; the monitoring item data includes but is not limited to wellhead temperature, bottom hole temperature, oil nozzle, wellhead pressure, bottom hole flow pressure, oil production, gas production, water production, liquid production, water content, pump frequency, pump current, pump voltage, pump leakage current, pump vibration, pump inlet temperature, pump motor temperature, pump inlet pressure and pump outlet pressure.

[0054] Step 2, taking each monitoring project data as a dimension, obtain the coordinates of each well in the feature space;

[0055] Step 3: The production monitoring data of all wells are spliced ​​into a whole, and the dimensions are filtered according to the overall statistical feature contribution rate based on principal component analysis to obtain an optimized feature space;

[0056] The filtering of dimensions based on principal component analysis according to the overall statistical feature contribution rate specifically includes:

[0057] Determine the principal components based on the cumulative variance contribution rate;

[0058] If the weight index of a dimension in each principal component is 0, the dimension will be eliminated.

[0059] Determining the principal component according to the cumulative variance contribution rate specifically includes:

[0060] The components with cumulative variance contribution rate above 80% were selected as principal components;

[0061] There are five principal components, which are represented by Y1, Y2, Y3, Y4, and Y5 respectively; among which:

[0062] The principal component Y1 comprehensively reflects the influence of wellhead temperature, bottom hole temperature, nozzle, wellhead pressure, bottom hole flowing pressure, oil production, water production, liquid production, water content and pump parameters on the development dynamics of the block;

[0063] The principal component Y2 reflects the negative effects of the wellhead-bottomhole temperature difference, the wellhead-bottomhole pressure difference, the pump leakage current, the pump vibration, the pump inlet temperature, the pump motor temperature, the pump inlet pressure, and the pump outlet pressure;

[0064] The principal component Y3 reflects the negative influence of oil production and the positive influence of water production and water content;

[0065] The main components Y4 and Y5 highlight the influence of gas production.

[0066] Step 4: Based on the optimized feature space, randomly select K objects as initial cluster centers, perform k-means clustering, and obtain the cluster center when the error square sum is minimized. The cluster center is used as the regular monitoring well location;

[0067] The sum of squared errors is calculated using the following formula:

[0068]

[0069] Among them, C i is the i-th class, p is C i Well in, m i It is C i center.

[0070] Step 5, change the value of K to determine the minimum number of monitoring wells and corresponding well locations as well as the optimal number of monitoring wells and corresponding well locations.

[0071] Change the value of K to determine the minimum number of monitoring wells and the corresponding well locations, including:

[0072] Change the value of K and determine the sum of squared errors corresponding to the K value;

[0073] Cluster analysis is performed using the K value corresponding to the minimum sum of squared errors as the number of clusters to determine the minimum number of monitoring wells and the corresponding well locations.

[0074] By changing the value of K, the optimal number of monitoring wells and the corresponding well locations are determined, including:

[0075] Change the value of K and determine the sum of squared errors corresponding to the K value;

[0076] Cluster analysis was performed using the K value corresponding to the error square sum less than 0.5 as the number of clusters to determine the optimal number of monitoring wells and the corresponding well locations.

[0077] The method further comprises:

[0078] Within a preset time interval, the above steps are repeated to update the minimum number of monitoring wells and the optimal number of monitoring wells to achieve dynamic adjustment of the monitoring well network.

[0079] The following are several specific embodiments of the present invention.

[0080] Example 1

[0081] In a specific embodiment 1 of the present invention, Figure 1 A flow chart of a method for intelligent optimization of a reservoir dynamic monitoring well network based on machine learning according to an embodiment of the present invention is shown. The method begins with step S100, obtaining production monitoring data of a well, wherein the production monitoring data includes data of multiple monitoring items.

[0082] The monitoring project data include but are not limited to wellhead temperature, bottomhole temperature, oil nozzle, wellhead pressure, bottomhole flow pressure, oil production, gas production, water production, liquid production, water content, pump frequency, pump current, pump voltage, pump leakage current, pump vibration, pump inlet temperature, pump motor temperature, pump inlet pressure and pump outlet pressure.

[0083] Step S200, taking each monitoring project data as a dimension, obtaining the coordinates of each well in the feature space.

[0084] Step S300, the production monitoring data of all wells are spliced ​​into a whole, and the dimensions are filtered according to the overall statistical feature contribution rate based on principal component analysis to obtain an optimized feature space.

[0085] In this embodiment, there are 19 monitoring project data, so the dimension of the feature space is 19, and the monitoring data comes from the production wells in the block. The data of the same indicator of all wells are spliced ​​into the same dimension as the data of the entire block.

[0086] In some embodiments, the main influencing factors of block development dynamics are analyzed based on the principal component expression and the cumulative variance contribution rate. The cumulative variance contribution rate of the five principal components is more than 80%, which reflects the main statistical characteristics of the block. The expressions of the five principal components are shown in Table 1. The principal component Y1 comprehensively reflects the impact of indicators such as wellhead temperature, bottom hole temperature, oil nozzle, wellhead pressure, bottom hole flow pressure, oil production, water production, liquid production, water content and pump parameters on the development dynamics of the block. There is no weighted significant indicator, but the weight of gas production is very small. The principal component Y2 reflects the negative effects of the wellhead and bottom hole temperature difference, the wellhead and bottom hole pressure difference, and some pump parameters. The principal component Y3 reflects the negative impact of oil production, and the positive impact of water production and water content. The principal components Y4 and Y5 highlight the impact of gas production. Therefore, there is no dimension that needs to be eliminated.

[0087] Table 1 Coefficient table of principal component expression of monitoring data

[0088]

[0089] Step S400, based on the optimized feature space, randomly select K objects as initial cluster centers, perform k-means clustering, and obtain the cluster center when the error square sum is minimized. The cluster center is used as the regular monitoring well location.

[0090] Based on the data of 19 monitoring indicators, k-means clustering is performed. As the number of clusters k increases, the corresponding sum of squared errors (SSE) gradually decreases. As the k value continues to increase, the change of SSE tends to be flat, that is, the relationship between SSE and k is an elbow shape ( Figure 2 ), and the k value corresponding to this elbow is the actual number of clusters of the data. Core indicator error sum of squares

[0091]

[0092] Among them, C i is the i-th class, p is C i Well in, m i It is C i SSE is the clustering error of all well locations, which represents the quality of clustering effect.

[0093] Finally, in step S500, the value of K is changed to determine the minimum number of monitoring wells and the optimal number of monitoring wells.

[0094] In this embodiment, according to Figure 2 The minimum number of clusters is selected as 3, and cluster analysis is performed on the 19-dimensional feature space. The results are as follows Figure 3As shown in the figure, according to the distance of each well from the cluster center, the regular monitoring wells are selected as 7, 15, and 17. SSE < 0.5 is set as the optimal number of monitoring wells, and the corresponding number of monitoring wells is k = 5. Cluster analysis is also performed on the 19-dimensional feature space. The results are Figure 4 According to the distance of each well from the cluster center, the regularly monitored wells are selected as 1, 7, 10, 17, and 18.

[0095] In some embodiments, the method further includes step S600, repeating steps S100-S500 within a preset time interval to update the minimum number of monitoring wells and the optimal number of monitoring wells to achieve dynamic adjustment of the monitoring well network.

[0096] The embodiment of the present invention further provides a device for intelligent optimization of reservoir dynamic monitoring well patterns based on machine learning, the device comprising a processor, wherein the processor is configured to:

[0097] Acquiring production monitoring data of the well, wherein the production monitoring data includes data of multiple monitoring items;

[0098] Taking each monitoring project data as a dimension, the coordinates of each well in the feature space are obtained;

[0099] The production monitoring data of all wells are spliced ​​as a whole, and the dimensions are filtered according to the overall statistical feature contribution rate based on principal component analysis to obtain the optimized feature space;

[0100] Based on the optimized feature space, K objects are randomly selected as initial cluster centers, and k-means clustering is performed. When the sum of squared errors is minimized, the cluster center is obtained, and the cluster center is used as the well location for regular monitoring;

[0101] By changing the value of K, the minimum number of monitoring wells and the optimal number of monitoring wells can be determined.

[0102] In this embodiment, the monitoring project data include but are not limited to wellhead temperature, bottom hole temperature, oil nozzle, wellhead pressure, bottom hole flow pressure, oil production, gas production, water production, liquid production, water content, pump frequency, pump current, pump voltage, pump leakage current, pump vibration, pump inlet temperature, pump motor temperature, pump inlet pressure and pump outlet pressure.

[0103] In this embodiment, the processor is further configured to:

[0104] Determine the principal components based on the cumulative variance contribution rate;

[0105] If the weight index of a dimension in each principal component is 0, the dimension will be eliminated.

[0106] In this embodiment, the processor is further configured to:

[0107] The components with cumulative variance contribution rate above 80% were selected as principal components;

[0108] There are five principal components, which are represented by Y1, Y2, Y3, Y4, and Y5 respectively; among which:

[0109] The principal component Y1 comprehensively reflects the influence of wellhead temperature, bottom hole temperature, nozzle, wellhead pressure, bottom hole flowing pressure, oil production, water production, liquid production, water content and pump parameters on the development dynamics of the block;

[0110] The principal component Y2 reflects the negative effects of the wellhead-bottomhole temperature difference, the wellhead-bottomhole pressure difference, the pump leakage current, the pump vibration, the pump inlet temperature, the pump motor temperature, the pump inlet pressure, and the pump outlet pressure;

[0111] The principal component Y3 reflects the negative influence of oil production and the positive influence of water production and water content;

[0112] The main components Y4 and Y5 highlight the influence of gas production.

[0113] In this embodiment, the processor is further configured to:

[0114] The sum of squared errors is calculated using the following formula:

[0115]

[0116] Among them, C i is the i-th class, p is C i Well in, m i It is C i center.

[0117] In some embodiments, the processor is further configured to:

[0118] Change the value of K and determine the sum of squared errors corresponding to the K value;

[0119] Cluster analysis is performed using the K value corresponding to the minimum sum of squared errors as the number of clusters to determine the minimum number of monitoring wells and the corresponding well locations.

[0120] In this embodiment, the processor is further configured to:

[0121] Change the value of K and determine the sum of squared errors corresponding to the K value;

[0122] Cluster analysis was performed using the K value corresponding to the error square sum less than 0.5 as the number of clusters to determine the optimal number of monitoring wells and the corresponding well locations.

[0123] In this embodiment, the processor is further configured to:

[0124] Within a preset time interval, the above steps are repeated to update the minimum number of monitoring wells and the optimal number of monitoring wells to achieve dynamic adjustment of the monitoring well network.

[0125] This embodiment further provides a readable storage medium, which stores one or more programs. The one or more programs can be executed by one or more processors to implement the method described above.

[0126] Example 2

[0127] In the specific embodiment 2 of the present invention, the Figure 1 The method flow chart is shown.

[0128] Step S100, reacquiring production monitoring data of 20 wells, wherein the production monitoring data includes data of multiple monitoring items.

[0129] The monitoring project data include but are not limited to wellhead temperature, bottomhole temperature, oil nozzle, wellhead pressure, bottomhole flow pressure, oil production, gas production, water production, liquid production, water content, pump frequency, pump current, pump voltage, pump leakage current, pump vibration, pump inlet temperature, pump motor temperature, pump inlet pressure and pump outlet pressure.

[0130] Step S200, taking each monitoring project data as a dimension, obtaining the coordinates of each well in the feature space.

[0131] Step S300, the production monitoring data of all wells are spliced ​​into a whole, and the dimensions are filtered according to the overall statistical feature contribution rate based on principal component analysis to obtain an optimized feature space.

[0132] In this embodiment, there are 19 monitoring project data, so the dimension of the feature space is 19, and the monitoring data comes from the production wells in the block. The data of the same indicator of all wells are spliced ​​into the same dimension as the data of the entire block.

[0133] In this embodiment, the main influencing factors of block development dynamics are analyzed based on the principal component expression and cumulative variance contribution rate. The cumulative variance contribution rate of 6 principal components is more than 80%, which reflects the main statistical characteristics of the block. The expression of 6 principal components is shown in Table 2.

[0134] Table 2 Coefficient table of principal component expression of monitoring data

[0135]

[0136] Step S400, based on the optimized feature space, randomly select K objects as initial cluster centers, perform k-means clustering, and obtain the cluster center when the error square sum is minimized. The cluster center is used as the regular monitoring well location.

[0137] Based on the data of 19 monitoring indicators, k-means clustering is performed. As the number of clusters k increases, the corresponding sum of squared errors (SSE) gradually decreases. As the k value continues to increase, the change of SSE tends to be flat, that is, the relationship between SSE and k is an elbow shape ( Figure 2 ), and the k value corresponding to this elbow is the actual number of clusters of the data. Core indicator error sum of squares

[0138]

[0139] Among them, C i is the i-th class, p is C i Well in, m i It is C i SSE is the clustering error of all well locations, which represents the quality of clustering effect.

[0140] Finally, in step S500, the value of K is changed to determine the minimum number of monitoring wells and the optimal number of monitoring wells.

[0141] In this embodiment, according to Figure 5 The minimum number of clusters is selected as 3, and cluster analysis is performed on the 19-dimensional feature space. The results are as follows Figure 6 According to the distance of each well from the cluster center, the regularly monitored wells 1, 12, and 13 are selected.

[0142] Example 3

[0143] In the specific embodiment 3 of the present invention, the Figure 1 The method flow chart is shown.

[0144] Step S100, reacquiring production monitoring data of 30 wells, wherein the production monitoring data includes data of multiple monitoring items.

[0145] The monitoring project data include but are not limited to wellhead temperature, bottomhole temperature, oil nozzle, wellhead pressure, bottomhole flow pressure, oil production, gas production, water production, liquid production, water content, pump frequency, pump current, pump voltage, pump leakage current, pump vibration, pump inlet temperature, pump motor temperature, pump inlet pressure and pump outlet pressure.

[0146] Step S200, taking each monitoring project data as a dimension, obtaining the coordinates of each well in the feature space.

[0147] Step S300, the production monitoring data of all wells are spliced ​​into a whole, and the dimensions are filtered according to the overall statistical feature contribution rate based on principal component analysis to obtain an optimized feature space.

[0148] In this embodiment, there are 19 monitoring project data, so the dimension of the feature space is 19, and the monitoring data comes from the production wells in the block. The data of the same indicator of all wells are spliced ​​into the same dimension as the data of the entire block.

[0149] In this embodiment, the main influencing factors of block development dynamics are analyzed based on the principal component expression and cumulative variance contribution rate. The cumulative variance contribution rate of the five principal components is more than 80%, which reflects the main statistical characteristics of the block. The expression of the five principal components is shown in Table 3.

[0150] Table 3 Coefficient table of principal component expression of monitoring data

[0151]

[0152] Step S400, based on the optimized feature space, randomly select K objects as initial cluster centers, perform k-means clustering, and obtain the cluster center when the error square sum is minimized. The cluster center is used as the regular monitoring well location.

[0153] Based on the data of 19 monitoring indicators, k-means clustering is performed. As the number of clusters k increases, the corresponding sum of squared errors (SSE) gradually decreases. As the k value continues to increase, the change of SSE tends to be flat, that is, the relationship between SSE and k is an elbow shape ( Figure 2 ), and the k value corresponding to this elbow is the actual number of clusters of the data. Core indicator error sum of squares

[0154]

[0155] Among them, C i is the i-th class, p is C i Well in, m i It is C i SSE is the clustering error of all well locations, which represents the quality of clustering effect.

[0156] Finally, in step S500, the value of K is changed to determine the minimum number of monitoring wells and the optimal number of monitoring wells.

[0157] In this embodiment, according to Figure 7 It is found that the minimum number of monitoring wells is not obvious. SSE < 0.5 is set as the optimal number of monitoring wells, and the corresponding number of monitoring wells is k = 6. Cluster analysis is performed on the 19-dimensional feature space. The results Figure 8 According to the distance of each well from the cluster center, the regular monitoring wells are selected as 7, 10, 11, 18, 28, and 29.

[0158] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions recorded in the aforementioned embodiments or replace some of the technical features therein with equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0159] Except for the technical features described in the specification, all other technical features are known technologies to those skilled in the art.

Claims

1. An intelligent optimization method for reservoir dynamic monitoring well pattern based on machine learning, characterized in that: The intelligent optimization method for reservoir dynamic monitoring well network based on machine learning includes: Step 1, obtaining production monitoring data of the well; Step 2, taking each monitoring project data as a dimension, obtain the coordinates of each well in the feature space; Step 3, filter the dimensions to obtain an optimized feature space; Step 4: Perform k-means clustering and obtain the cluster center when the sum of squared errors is minimized; Step 5, change the value of K to determine the minimum number of monitoring wells and corresponding well locations as well as the optimal number of monitoring wells and corresponding well locations.

2. The intelligent optimization method for reservoir dynamic monitoring well pattern based on machine learning according to claim 1 is characterized in that: In step 1, production monitoring data of a well is obtained, and the production monitoring data includes data of multiple monitoring items.

3. The intelligent optimization method for reservoir dynamic monitoring well pattern based on machine learning according to claim 2 is characterized in that: In step 1, the monitoring project data include wellhead temperature, bottom hole temperature, oil nozzle, wellhead pressure, bottom hole flow pressure, oil production, gas production, water production, liquid production, water content, pump frequency, pump current, pump voltage, pump leakage current, pump vibration, pump inlet temperature, pump motor temperature, pump inlet pressure and pump outlet pressure.

4. The intelligent optimization method for reservoir dynamic monitoring well pattern based on machine learning according to claim 1 is characterized in that: In step 3, the production monitoring data of all wells are spliced ​​into a whole, and the dimensions are filtered according to the overall statistical feature contribution rate based on principal component analysis to obtain an optimized feature space.

5. The intelligent optimization method for reservoir dynamic monitoring well pattern based on machine learning according to claim 4 is characterized in that: In step 3, the dimensions are filtered based on the principal component analysis according to the overall statistical feature contribution rate, including: Step 31, determining the principal component according to the cumulative variance contribution rate; Step 32: If the weight index of a dimension in each principal component is 0, then the dimension is eliminated.

6. The intelligent optimization method for reservoir dynamic monitoring well pattern based on machine learning according to claim 5 is characterized in that: In step 31, the components with cumulative variance contribution rates of more than 80% are taken as principal components; there are five principal components, which are represented as Y1, Y2, Y3, Y4, and Y5 respectively.

7. The intelligent optimization method for reservoir dynamic monitoring well pattern based on machine learning according to claim 6 is characterized in that: In step 31, the principal component Y1 comprehensively reflects the influence of wellhead temperature, bottom hole temperature, oil nozzle, wellhead pressure, bottom hole flowing pressure, oil production, water production, liquid production, water content and pump parameters on the development dynamics of the block; The principal component Y2 reflects the negative effects of the wellhead-bottomhole temperature difference, the wellhead-bottomhole pressure difference, the pump leakage current, the pump vibration, the pump inlet temperature, the pump motor temperature, the pump inlet pressure, and the pump outlet pressure; The principal component Y3 reflects the negative influence of oil production and the positive influence of water production and water content; The main components Y4 and Y5 highlight the influence of gas production.

8. The intelligent optimization method for reservoir dynamic monitoring well pattern based on machine learning according to claim 1 is characterized in that: In step 4, based on the optimized feature space, K objects are randomly selected as initial cluster centers, and k-means clustering is performed. When the sum of squared errors is minimized, the cluster center is obtained and used as the well location for regular monitoring.

9. The intelligent optimization method for reservoir dynamic monitoring well pattern based on machine learning according to claim 8, characterized in that: In step 4, the sum of squared errors is calculated using the following formula: Among them, C i is the i-th class, p is C i Well in, m i It is C i center.

10. The intelligent optimization method for reservoir dynamic monitoring well pattern based on machine learning according to claim 1, characterized in that: In step 5, the value of K is changed to determine the minimum number of monitoring wells and the corresponding well locations, including: Change the value of K and determine the sum of squared errors corresponding to the K value; Cluster analysis is performed using the K value corresponding to the minimum sum of squared errors as the number of clusters to determine the minimum number of monitoring wells and the corresponding well locations.

11. The intelligent optimization method for reservoir dynamic monitoring well pattern based on machine learning according to claim 10, characterized in that: In step 5, the value of K is changed to determine the optimal number of monitoring wells and the corresponding well locations, which specifically includes: Change the value of K and determine the sum of squared errors corresponding to the K value; Cluster analysis was performed using the K value corresponding to the error square sum less than 0.5 as the number of clusters to determine the optimal number of monitoring wells and the corresponding well locations.

12. The intelligent optimization method for reservoir dynamic monitoring well pattern based on machine learning according to claim 1, characterized in that: The intelligent optimization method for reservoir dynamic monitoring well network based on machine learning also includes, after step 5, repeating steps 1-5 within a preset time interval to update the minimum number of monitoring wells and the optimal number of monitoring wells to achieve dynamic adjustment of the monitoring well network.

Images