Method and system for quantitatively representing connectivity between injection and production wells by using machine learning

Abstract

The invention discloses a method and system for quantitatively representing connectivity between injection wells and production wells through machine learning, and belongs to the technical field of heavy oil reservoir fireflooding reservoir dynamic monitoring. Comparing and analyzing the logging curves to determine the structural similarity and the trend similarity of the logging curves among the fireflooding injection and production wells, and determining a comprehensive static similarity coefficient among the fireflooding injection and production wells according to the structural similarity and the trend similarity; according to the fireflooding development block engineering data, main control factors influencing the yield of the production well are determined, a main feature-multi-element injection and production correlation model is constructed in combination with the main control factors and the comprehensive static similarity coefficient, and the inter-well connectivity coefficient of the fireflooding injection and production well is determined by adopting the main feature-multi-element injection and production correlation model; according to the method, the dynamic monitoring and analysis of the communication condition between the injection well and the production well in the fireflooding development process can be efficiently achieved, the dynamic monitoring and adjustment of heavy oil reservoir fireflooding can be optimized, and therefore the recovery efficiency and the economic benefits are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of dynamic monitoring of fire-driven oil reservoirs in heavy oil reservoirs, and in particular to a method and system for quantitatively characterizing connectivity between injection and production wells using machine learning. Background Art

[0002] Fire flooding is an effective method to enhance the recovery of heavy and thick oil. The technology ignites part of the crude oil in the reservoir to form a combustion front, which transfers heat energy to the deep reservoir, thereby reducing the viscosity of the crude oil and improving its fluidity. Fire flooding technology can significantly improve the ultimate recovery of the reservoir, especially in high-viscosity reservoirs that cannot be effectively exploited by traditional water injection or steam flooding. During the fire flooding process, the connectivity between the injection well and the production well is one of the key factors determining the recovery rate. Good inter-well connectivity can ensure uniform advancement of the combustion front, improve thermal efficiency, and reduce ineffective areas.

[0003] However, the geological complexity and heterogeneity in actual reservoirs make it very difficult to quantify well connectivity. Although traditional geological modeling and numerical simulation methods can provide certain prediction capabilities, they often require a large amount of geological data and computing resources, and the accuracy of the results depends on the quality of the input data and the assumptions of the model. Summary of the invention

[0004] In view of the fact that the existing technology cannot quickly and accurately quantify the connectivity status between injection and production wells in fire drive of heavy oil reservoirs, the purpose of the present invention is to provide a method and system for quantitatively characterizing the connectivity between injection and production wells by using machine learning. The connectivity status between injection and production wells in fire drive is calculated by static geological data and dynamic engineering data, which can efficiently solve the dynamic monitoring and analysis of the connectivity status between injection and production wells during fire drive development, help optimize the dynamic monitoring and adjustment of fire drive in heavy oil reservoirs, thereby improving the recovery rate and economic benefits.

[0005] The present invention is achieved through the following technical solutions: In a first aspect, the present application provides a method for quantitatively characterizing the connectivity between injection and production wells using machine learning, including the following process: Step 1, drawing a logging curve based on the logging data of the well group in the fire-driven development block, comparing and analyzing the logging curves to determine the structural similarity and trend similarity of the logging curves between the fire-driven injection and production wells, and determining the comprehensive static similarity coefficient between the fire-driven injection and production wells based on the structural similarity and trend similarity; Step 2: Determine the main controlling factors affecting the production of production wells based on the engineering data of the fire-driven development block, and construct a main feature-multivariate injection-production association model by combining the main controlling factors and the comprehensive static similarity coefficient. Use the main feature-multivariate injection-production association model to determine the well connectivity coefficient of the fire-driven injection-production wells.

[0006] Preferably, drawing the logging curve in step 1 includes: Obtain gamma logging data, sonic logging data and resistance logging data of the well group in the fire drive development block, pre-process the logging data, and draw logging curves based on the pre-processed logging data.

[0007] Preferably, the method for determining the comprehensive static similarity coefficient between the fire flooding injection and production wells in step 1 is as follows: The brightness, contrast and structure of the logging curves of each well are compared to determine the structural similarity between the logging curves; the fluctuation variation of the logging curves of each well within the set range is compared to determine the trends between the logging curves; the comprehensive static similarity coefficient is determined based on the average value of the structural similarity and the trend similarity.

[0008] Preferably, the method for determining the main controlling factors in step 2 includes: The machine learning method is used to determine the importance score of each characteristic variable in the engineering data, the importance scores are sorted, and the first several characteristic variables in the sequence are taken as the main controlling factors.

[0009] Preferably, in step 2, multiple machine learning methods are used to respectively determine the importance score of each feature variable in the engineering data, as well as the precision score of the feature variable importance score. The optimal machine learning method is determined according to the highest precision score, the feature variable importance scores of the optimal machine learning outputs are ranked, and the feature variables with the top three feature variable importance scores are used as the main controlling features.

[0010] Preferably, the multiple machine learning methods include a decision tree algorithm, a random forest algorithm and an AdaBoost algorithm.

[0011] Preferably, the engineering data include oil production, water content, gas injection volume, wellhead temperature, production days, oil pressure, well spacing and open / closed well conditions; Taking oil production as the target variable, water cut, monthly gas injection volume, wellhead temperature, production days, oil pressure, well spacing and open / closed well conditions as characteristic variables, the importance score of each characteristic variable was calculated.

[0012] Preferably, the method for constructing the main feature-multivariate injection-production association model is as follows: The comprehensive static similarity coefficient and the main characteristics of each target production well were taken as independent variables, and the monthly oil production of the target production well was taken as the dependent variable. The main characteristics-multivariate injection-production association model suitable for fire flooding was constructed by multivariate linear regression.

[0013] Preferably, the expression of the main feature-multivariate injection-production association model is as follows:

[0014] in, For thei The injection well is at the time step n The injection volume; is a constant term, injection-production imbalance coefficient; is the comprehensive static similarity coefficient between the j-th production well and the i-th injection well obtained in step 4; For the j Production well and i Dynamic correlation weights between injection wells; For the j The liquid production of a production well at time step n.

[0015] In a second aspect, the present application provides a system for quantitatively characterizing connectivity between injection and production wells using machine learning, comprising: The static coefficient module is used to draw the logging curve according to the logging data of the well group in the fire-driven development block, compare and analyze the logging curve to determine the structural similarity and trend similarity of the logging curves between the fire-driven injection and production wells, and determine the comprehensive static similarity coefficient between the fire-driven injection and production wells according to the structural similarity and trend similarity; The well-to-well connectivity characterization module is used to determine the main controlling factors affecting the production of production wells based on the engineering data of the fire-driven development block, and to construct a main feature-multivariate injection-production association model by combining the main controlling factors and the comprehensive static similarity coefficient. The main feature-multivariate injection-production association model is used to determine the well-to-well connectivity coefficient of the fire-driven injection-production wells.

[0016] Compared with the prior art, the present invention has the following beneficial technical effects: The present application provides a method for quantitatively characterizing the connectivity between injection and production wells using machine learning, using machine learning methods to select the main features associated with fire drive production, including static geological feature parameters and dynamic engineering feature parameters, to construct a main feature-multivariate injection and production association model suitable for fire drive, which can analyze the influence of geological features and engineering features on the connectivity between injection and production wells in fire drive, and can analyze the main features associated with production between injection and production wells in actual fire drive production process, quantitatively characterize the connectivity between injection and production wells in fire drive by calculating model parameters, and avoid the influence of redundant features on the accuracy of machine learning process through feature selection, reduce the workload of data analysis, adapt to the actual production of specific well groups in the process of fire drive in specific oil reservoirs, and efficiently and accurately quantify the connectivity between injection and production wells in fire drive. The present invention solves the problem that the connectivity between injection and production wells in the process of fire drive is difficult to quickly and accurately quantify, and can add various data through the actual production process to monitor the connectivity between injection and production wells in fire drive in real time, providing a decision-making basis for dynamic monitoring, dynamic adjustment and safe production of fire drive.

[0017] The present application also proposes a system for quantitatively characterizing connectivity between injection and production wells in fire-driven oil reservoirs based on machine learning, an electronic device and a computer storage medium, which have all the advantages of the above-mentioned method for quantitatively characterizing connectivity between injection and production wells in fire-driven oil reservoirs based on machine learning. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0019] Figure 1 A flow chart of a method for quantitatively characterizing connectivity between injection and production wells using machine learning according to the present invention; Figure 2 The figure is a graph of the feature importance calculation results calculated by three methods in the present invention; Figure 3 This is the accuracy diagram of the results calculated by the three methods used in the present invention; Figure 4 Schematic diagram of well-to-well connectivity of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.

[0021] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for which protection is sought, but merely represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0022] In order to efficiently clarify the connectivity status between fire-driven injection and production wells and help dynamic monitoring and adjustment during fire-driven development, the present invention proposes a quantitative characterization method for connectivity between injection and production wells in fire-driven oil reservoirs based on machine learning. The method uses machine learning data analysis methods, comprehensively considers geological factors and engineering development factors, combines multiple disciplines such as computer science and reservoir development, determines the main engineering factors associated with production between fire-driven injection and production wells, establishes a characteristic association model between injection and production wells suitable for fire-driven, and calculates and quantifies the connectivity status between fire-driven injection and production wells. The method is described in detail below.

[0023] A method for quantitatively characterizing connectivity between injection and production wells using machine learning comprises the following steps: Step 1: Obtain the well logging data of the well group in the fire drive development block, draw the well logging curve according to the well logging data, compare and analyze the well logging curve to determine the well logging curve structure similarity and trend similarity between the fire drive injection and production wells, and obtain the comprehensive static similarity coefficient between the fire drive injection and production wells, as follows: S1.1. Obtain logging data, including gamma logging data, sonic logging data and electrical resistance logging data corresponding to the formation depth of the target oil layer of each well.

[0024] S1.2. Preprocess the logging data, including outlier processing and data noise reduction.

[0025] S1.3. Draw the gamma logging curve, sonic logging curve and resistance logging curve respectively according to the pre-processed logging data.

[0026] S1.4. Compare and analyze the gamma logging curves, sonic logging curves and electrical resistance logging curves to determine the structural similarity and trend similarity of the logging curves between the fire drive injection and production wells.

[0027] 1) Structural similarity of logging images (SSIM) By comparing the brightness, contrast and structure of logging images, the larger the value, the more similar the structures between the logging curves of the two wells are:

[0028] In the formula, and is the local window of the two image curves; and is the average of the two windows; and is the standard deviation of the two windows; is the covariance of the two windows; and It is a constant term added to stabilize the calculation.

[0029] 2) Trend similarity of logging curve images (TSIM) is achieved by comparing the fluctuation of the logging curve images within a certain distance. The larger the value, the more similar the trends between the logging curves of the two wells are.

[0030]

[0031] Among them, MSE is the mean square error, which refers to the average value of the square of the difference between the predicted value and the true value, and is calculated using formula (3):

[0032] In the formula, is the value of a signal of the curve in the figure; is the value of a signal for the curve in another graph; is the sample size.

[0033] S1.5. Determine the comprehensive static similarity coefficient based on structural similarity and trend similarity.

[0034] Comprehensive static similarity coefficient It is the average value of the sum of the structural similarity coefficient and the trend similarity coefficient. A larger value indicates that the geological conditions between the logging curves of the two wells are similar.

[0035]

[0036] Step 2: Obtain engineering data of the fire-driven development block, determine the main controlling factors affecting the production of production wells based on the engineering data, and build a main feature-multivariate injection-production association model by combining the main controlling factors and the comprehensive static similarity coefficient. Use the main feature-multivariate injection-production association model to determine the engineering factor influence weights between fire-driven injection-production wells, and determine the connectivity coefficient between injection-production wells based on the engineering factor influence weights, as follows: S2.1. Obtain engineering data of fire drive development blocks, including monthly oil production, water cut, monthly gas injection volume, wellhead temperature, production days, oil pressure, well spacing and open / closed well status.

[0037] S2.2. Match the response of the gas injection wells with the target production wells and pre-process the engineering data of the fire drive development block.

[0038] S2.3, constructing an initial injection and production data sample based on the preprocessed engineering data and performing normalization processing to obtain a standard sample data set, and dividing the standard sample data set into a training set and a test set in a ratio of 8:2; S2.4. Take monthly oil production as the target variable, and water content, monthly gas injection volume, wellhead temperature, production days, oil pressure, well spacing and opening / closing well conditions as characteristic variables; use a variety of machine learning methods to determine the importance score of each characteristic variable and the precision score of the characteristic variable importance score. Determine the optimal machine learning method based on the highest precision score, rank the characteristic variable importance scores of the optimal machine learning output, and take the characteristic variables with the top three characteristic variable importance scores as the main controlling factors.

[0039] A variety of machine learning methods including decision tree algorithm, random forest algorithm and AdaBoost algorithm are used for classification algorithm, and the importance scores of feature variables and the precision scores of the importance scores of feature variables are calculated.

[0040] 1) Use decision tree to calculate the importance score of each feature variable as follows:

[0041] In the formula, For all nodes of the decision tree; For the Use features in nodes The information gain when splitting is as follows:

[0042] In the formula, is the entropy of the current node, which can be calculated by formula (7); Features Each value The corresponding subset; is the number of subset samples; is the total number of node samples.

[0043]

[0044] In the formula, is the number of categories; For category probability.

[0045] 2) Use the decision tree to calculate the result accuracy score as follows:

[0046] In the formula, For true positive examples, the number of samples predicted or classified as positive and actually positive; True negative examples are the number of samples predicted or classified as negative and actually negative; False positives are the number of samples that are predicted or classified as positive but are actually negative. False negatives are the number of samples that are predicted or classified as negative but are actually positive.

[0047] 3) Use random forest to calculate the importance score of each feature variable as follows:

[0048] In the formula, is the number of trees in the random forest; For the Using features in a tree Information gain when performing a split.

[0049] The method for calculating the accuracy of the results of random forests is the same as the method for calculating the accuracy score of the results of decision trees. See step 2 for details).

[0050] 4) The importance score of each feature variable is calculated using the AdaBoost algorithm, where the algorithm uses a weak classifier and a decision tree as follows:

[0051] In the formula, is the indicator function, if the characteristic In the classifier If it is used, it takes 1, otherwise it takes 0; is the total number of weak classifiers.

[0052] The calculation method of the result accuracy of the AdaBoost algorithm is consistent with the calculation method of the result accuracy of the decision tree, see step 2 for details).

[0053] S2.5. The comprehensive static similarity coefficient and the main characteristics of each target production well are taken as independent variables, and the monthly oil production of the target production well is taken as the dependent variable. The main characteristics-multivariate injection-production association model suitable for fire flooding is constructed by multivariate linear regression. The expression is as follows:

[0054] in, For the i The injection well is at the time step n The injection volume; is a constant term, injection-production imbalance coefficient; is the comprehensive static similarity coefficient between the j-th production well and the i-th injection well obtained in step 4; For the j Production well and i Dynamic correlation weights between injection wells; For the j The liquid production of a production well at time step n.

[0055] S2.6. Calculate the multivariate regression weight coefficient of the main feature-multivariate injection-production association model The obtained weight coefficient is the obtained correlation coefficient between the fire-driven injection-production wells, which can also be understood as the well-to-well connectivity coefficient of the fire-driven injection-production well group.

[0056] Step 3, repeating steps 1 and 2, obtaining the injection-production connectivity coefficients of each production well and gas injection well in the required fire-driven oil well group, drawing an injection-production connectivity diagram, and clarifying the connectivity status between the fire-driven injection-production wells.

[0057] The method for quantitatively characterizing the connectivity between injection and production wells in fire-driven oil reservoirs based on machine learning uses geological data and engineering data in the process of oil reservoir development, and compares and analyzes well logging curves to obtain a comprehensive static similarity coefficient, and uses machine learning to screen the main features associated with production between injection and production wells in fire-driven oil reservoirs, and uses the comprehensive static similarity coefficient and main features to construct a main feature-multivariate injection and production association model suitable for injection and production wells in fire-driven oil reservoirs, calculates the connectivity coefficient between injection and production wells in fire-driven oil reservoirs, and accurately quantifies and characterizes the connectivity between injection and production wells. The present invention solves the problem of difficulty in determining the connectivity judgment between gas injection wells and corresponding production wells during the fire-driven development of heavy oil reservoirs, can accurately and quickly identify the connectivity between injection and production wells, and can also judge the connectivity between injection and production wells based on real-time updates of actual production, providing assistance for the regulation and control of fire-driven development of heavy oil reservoirs.

[0058] Example 1 Taking a fire drive development well group in a heavy oil block of a certain oil field as an example, a method of using machine learning to quantitatively characterize the connectivity between injection and production wells in this application is described.

[0059] S100, collecting well logging data and engineering data of the oil field block to be analyzed, the well logging data and engineering data of the well group include those shown in Table 1.

[0060] Table 1

[0061] S200: Draw logging curves using the logging data of the target oil layer steam injection well and the target production well collected by S100.

[0062] S300: performing noise reduction processing on the well logging curve obtained in S200.

[0063] S400: Calculate the structural similarity and trend similarity between the target oil layer gas injection well logging curve and the target production well logging curve image, and obtain the comprehensive static similarity coefficient between the gas injection well and the target oil well, as shown in Table 2.

[0064] Table 2

[0065] S500: Match the response of the gas injection well with the target production well, perform data preprocessing on the engineering data obtained in S100, and the preprocessed data constitute the initial injection and production data sample. The initial injection sample data is normalized to form a standard sample data set, and the standard sample data set is divided into a training set and a test set in a ratio of 8:2.

[0066] S600: Based on the training set, monthly oil production is used as the target variable, water content, monthly gas injection volume, wellhead temperature, production days, oil pressure, well spacing and open / closed well conditions are used as feature variables, and feature importance score calculation and result accuracy calculation are performed based on decision tree, random forest and AdaBoost algorithms respectively. The feature importance score calculation results are shown in Figure 2. Figure 2 As shown, the result accuracy calculation results are as follows Figure 3 shown.

[0067] S700: Compare and analyze the accuracy of the results of S600, select the algorithm with the highest result accuracy as the algorithm for calculating the feature importance score, and get the feature importance ranking. The top three feature importance rankings (sorted from large to small according to the feature importance scores) are taken as the main features affecting the monthly oil production of the target production wells. Figure 2 , Figure 3 The main characteristics can be selected as: water content, monthly gas injection and oil pressure.

[0068] S800: Select the comprehensive static similarity coefficient obtained in S400 and the main characteristics of each target production well in S700 as independent variables, and the monthly oil production of the target production well as the dependent variable to construct a main characteristic-multivariate injection-production association model suitable for fire flooding; S900: Calculate the multivariate regression weight coefficient of the main feature-multivariate injection-production association model. The obtained weight coefficient is the obtained association coefficient between the fire-driven injection-production wells, and is also the connectivity coefficient between the fire-driven injection-production wells. The results are shown in Table 3.

[0069] Table 3

[0070] S1000: Repeat steps S100 to S900 to obtain the injection-production connectivity coefficients of each production well and gas injection well in the required fire-driven oil well group, draw an injection-production connectivity diagram, and clarify the connectivity status between the fire-driven injection-production wells, such as Figure 4 shown.

[0071] Based on the above method for quantitatively characterizing the connectivity between injection and production wells using machine learning, the present application also provides a system for quantitatively characterizing the connectivity between injection and production wells using machine learning, including: The static coefficient module is used to draw the logging curve according to the logging data of the well group in the fire-driven development block, compare and analyze the logging curve to determine the structural similarity and trend similarity of the logging curves between the fire-driven injection and production wells, and determine the comprehensive static similarity coefficient between the fire-driven injection and production wells according to the structural similarity and trend similarity; The well-to-well connectivity characterization module is used to determine the main controlling factors affecting the production of production wells based on the engineering data of the fire-driven development block, and to construct a main feature-multivariate injection-production association model by combining the main controlling factors and the comprehensive static similarity coefficient. The main feature-multivariate injection-production association model is used to determine the well-to-well connectivity coefficient of the fire-driven injection-production wells.

[0072] It should be noted that in the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of each module is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules can be combined or integrated into another device, or some features can be ignored or not executed. The module described as a separate component may or may not be physically separated. The component displayed as a module may be a physical unit or multiple physical units, that is, it may be located in one place, or it may be distributed in multiple different places. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0073] In addition, each module in each embodiment of the present invention may be integrated into a processing unit, each module may exist physically separately, or two or more modules may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of a software functional unit.

[0074] An electronic device provided in an embodiment of the present application includes a memory and a processor. A computer program is stored in the memory. When the processor executes the computer program, the steps of the method for quantitatively characterizing connectivity between injection and production wells in a fire-driven oil reservoir based on machine learning as described in any of the above embodiments are implemented.

[0075] Another electronic device provided in the embodiment of the present application may also include: an input port connected to the processor, used to transmit multimodal data collected by an external acquisition device to the processor; and a display unit connected to the processor, used to display the processing results of the processor to the outside world; a communication module connected to the processor, used to realize the communication between the electronic device and the outside world. The display unit can be a display panel, a laser scanning display, etc.; the communication mode adopted by the communication module includes but is not limited to mobile high-definition link technology (HML), universal serial bus (USB), high-definition multimedia interface (HDMI), wireless connection (including wireless fidelity technology (WiFi), Bluetooth communication technology, low-power Bluetooth communication technology, and communication technology based on IEEE802.11s).

[0076] An embodiment of the present application provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the steps of the method for quantitatively characterizing connectivity between injection and production wells in a fire-drive oil reservoir based on machine learning as described in any of the above embodiments are implemented.

[0077] For the description of the relevant parts of the system for quantitative characterization of connectivity between injection and production wells in a fire-driven oil reservoir based on machine learning, the electronic device, and the computer-readable storage medium provided in the embodiments of the present application, please refer to the detailed description of the corresponding parts in the method for quantitative characterization of connectivity between injection and production wells in a fire-driven oil reservoir based on machine learning provided in the embodiments of the present application, which will not be repeated here. In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive elaboration.

[0078] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A method for quantitatively characterizing the connectivity between injection and production wells using machine learning, characterized in that: The process includes: Step 1, drawing a logging curve based on the logging data of the well group in the fire-driven development block, comparing and analyzing the logging curves to determine the structural similarity and trend similarity of the logging curves between the fire-driven injection and production wells, and determining the comprehensive static similarity coefficient between the fire-driven injection and production wells based on the structural similarity and trend similarity; Step 2: Determine the main controlling factors affecting the production of production wells based on the engineering data of the fire-driven development block, and construct a main feature-multivariate injection-production association model by combining the main controlling factors and the comprehensive static similarity coefficient. Use the main feature-multivariate injection-production association model to determine the well connectivity coefficient of the fire-driven injection-production wells.

2. The method of using machine learning to quantitatively characterize the connectivity between injection and production wells according to claim 1, characterized in that: Drawing the logging curve in step 1 includes: Obtain gamma logging data, sonic logging data and resistance logging data of the well group in the fire drive development block, pre-process the logging data, and draw logging curves based on the pre-processed logging data.

3. The method of using machine learning to quantitatively characterize the connectivity between injection and production wells according to claim 1, characterized in that: The method for determining the comprehensive static similarity coefficient between the fire flooding injection and production wells described in step 1 is as follows: Compare the brightness, contrast and structure of the logging curves of each well to determine the structural similarity between the logging curves; compare the fluctuation variation of the logging curves of each well within the set range to determine the more similar the trends between the logging curves; The comprehensive static similarity coefficient is determined based on the average value of structural similarity and the more similar trends.

4. The method of using machine learning to quantitatively characterize the connectivity between injection and production wells according to claim 1, characterized in that: The method for determining the main controlling factors in step 2 includes: The machine learning method is used to determine the importance score of each characteristic variable in the engineering data, the importance scores are sorted, and the first several characteristic variables in the sequence are taken as the main controlling factors.

5. The method of using machine learning to quantitatively characterize the connectivity between injection and production wells according to claim 4, characterized in that: In step 2, multiple machine learning methods are used to determine the importance score of each feature variable in the engineering data, as well as the precision score of the feature variable importance score. The optimal machine learning method is determined based on the highest precision score, and the feature variable importance scores of the optimal machine learning output are ranked. The feature variables with the top three feature variable importance scores are used as the main controlling features.

6. The method of using machine learning to quantitatively characterize the connectivity between injection and production wells according to claim 4, characterized in that: The multiple machine learning methods include decision tree algorithm, random forest algorithm and AdaBoost algorithm.

7. The method of using machine learning to quantitatively characterize the connectivity between injection and production wells according to claim 4, characterized in that: The engineering data include oil production, water content, gas injection volume, wellhead temperature, production days, oil pressure, well spacing and well opening / closing conditions; Taking oil production as the target variable, water cut, monthly gas injection volume, wellhead temperature, production days, oil pressure, well spacing and open / closed well conditions as characteristic variables, the importance score of each characteristic variable was calculated.

8. The method of using machine learning to quantitatively characterize the connectivity between injection and production wells according to claim 1, characterized in that: The method for constructing the main feature-multivariate injection-production association model is as follows: The comprehensive static similarity coefficient and the main characteristics of each target production well were taken as independent variables, and the monthly oil production of the target production well was taken as the dependent variable. The main characteristics-multivariate injection-production association model suitable for fire flooding was constructed by multivariate linear regression.

9. The method of using machine learning to quantitatively characterize the connectivity between injection and production wells according to claim 8, characterized in that: The expression of the main feature-multivariate injection-production association model is as follows: in, For the i The injection well is at the time step n The injection volume; is a constant term, injection-production imbalance coefficient; is the comprehensive static similarity coefficient between the j-th production well and the i-th injection well obtained in step 4; For the j Production well and i Dynamic correlation weights between injection wells; For the j The liquid production of a production well at time step n.

10. A system for quantitatively characterizing connectivity between injection and production wells using machine learning, characterized in that: include: The static coefficient module is used to draw the logging curve according to the logging data of the well group in the fire-driven development block, compare and analyze the logging curve to determine the structural similarity and trend similarity of the logging curves between the fire-driven injection and production wells, and determine the comprehensive static similarity coefficient between the fire-driven injection and production wells according to the structural similarity and trend similarity; The well-to-well connectivity characterization module is used to determine the main controlling factors affecting the production of production wells based on the engineering data of the fire-driven development block, and to construct a main feature-multivariate injection-production association model by combining the main controlling factors and the comprehensive static similarity coefficient. The main feature-multivariate injection-production association model is used to determine the well-to-well connectivity coefficient of the fire-driven injection-production wells.

Images