Single-well water control measure decision-making method and system based on mechanical subdivision

Abstract

The invention relates to a single-well water control measure decision-making method and system based on mechanical subdivision. The method comprises the following steps: S1, constructing a training sample library; and S2, based on the training sample library, screening main control factors influencing the water control measure effect by using a random forest algorithm. And S3, based on the screened main control factors, respectively constructing an XGBoost oil increase rate prediction model and an XGBoost water content decrease amplitude prediction model, and analyzing the contribution degree of each main control factor to the water control measure effect in combination with an SHAP value. And S4, inputting target single well parameters to the trained XGBoost prediction model based on the trained XGBoost prediction model, automatically evaluating implementation effects of three water control measure schemes including ICD + subdivision, AICD + subdivision and subdivision, and recommending an optimal decision scheme according to evaluation results. The method can effectively improve the oil increasing rate, reduce the water content, finally achieve single-well oil stabilization and water control, and improve the overall recovery rate of an oil field.

Description

Technical Field

[0001] This invention relates to the field of oilfield development technology, and in particular to a decision-making method and system for single-well water control measures based on mechanical compartmentation. Background Technology

[0002] As oilfield development enters its mid-to-late stages, waterflooding typically faces challenges such as rising water cut and declining production. The rational selection of single-well water control measures is crucial for maintaining stable oil well production and extending the effective development period of the oilfield. Traditional single-well water control decisions rely primarily on experience or statistical analysis, failing to comprehensively consider multi-dimensional geological, reservoir, and engineering factors, resulting in unsatisfactory water control effects.

[0003] Existing technologies have significant limitations in single-well water control decision-making. First, traditional methods often rely on single or a few indicators for judgment, ignoring the combined influence of reservoir geological characteristics (such as permeability and porosity) and engineering factors (such as production systems and timing of measures), resulting in a lack of scientific basis for decision-making [5-6]. Although intelligent methods have been introduced in existing technologies, they are mostly concentrated on a single link (such as intelligent control or data analysis), lacking a systematic solution covering the entire process from single-well water control measure characterization and sample generation to measure recommendation. In addition, there are currently few recommended methods for single-well water control measures. ICD (Inflow Control Device), AICD (Adaptive Inflow Control Device), and compartmentalized technology have been widely used for water control, but how to recommend the optimal measures based on specific single-well parameters (such as geological reservoir characteristics, fluid properties, and engineering parameters) remains a technical problem that urgently needs to be solved. With the increasing complexity of oilfield development, single technical means are difficult to meet the needs of multi-objective optimization. To improve the oil production rate while reducing the water cut, it is necessary to comprehensively consider the balance between multiple objectives. Summary of the Invention

[0004] The technical problem to be solved by this invention is to address the shortcomings of existing single-well water control measures decision-making, including the lack of systematicity, difficulty in comprehensively considering multi-dimensional geological and engineering factors, and inability to recommend the optimal measures based on specific well conditions and single-well data. This invention provides a single-well water control measures decision-making method and system based on mechanical compartmentation.

[0005] The technical solution adopted by this invention to solve its technical problem is: a decision-making method for single-well water control measures based on mechanical compartmentation, which includes the following steps: S1. Construct a training sample library; S2. Based on the training sample library, use the random forest algorithm to screen the main controlling factors that affect the effectiveness of water control measures; S3. Based on the selected main control factors, construct the XGBoost oil increase rate prediction model and the XGBoost water content decrease rate prediction model respectively, and analyze the contribution of each main control factor to the water control measures in combination with the SHAP value. S4. Based on the trained XGBoost prediction model, input the target single well parameters into it, automatically evaluate the implementation effect of three water control measures: ICD+compartment, AICD+compartment, and compartment, and recommend the optimal decision scheme based on the evaluation results.

[0006] Furthermore, in the single-well water control measure decision-making method based on mechanical compartmentation described in this invention, the method further includes the following step before step S1: Establish an evaluation index system for the effectiveness of water control measures, with the reduction in oil content and water content as the core.

[0007] Furthermore, in the single-well water control measure decision-making method based on mechanical compartmentation described in this invention, step S1 includes: Determine the input parameters and perform Latin hypercube sampling on the input parameters to generate a combination of sample points covering the parameter space; By combining the implementation effects of three water control measures—ICD+compartmental, AICD+compartmental, and compartmentalized—in reservoir numerical simulation, the corresponding oil increase rate and water cut decrease rate are calculated, and these are combined with the generated parameter sample points to construct a training sample library.

[0008] Furthermore, in the single-well water control measure decision-making method based on mechanical compartmentation described in this invention, the step of performing Latin hypercube sampling on the input parameters to generate a combination of sample points covering the parameter space includes: Divide the range of values ​​for each input parameter variable into m equally probable intervals, and generate sample points: Where, x ij F represents the value of the j-th sample point of the i-th variable; i-1 π is the inverse function of the cumulative distribution function of the i-th variable; j Let j be the j-th permutation position, and j = 1, ..., m; u ij The number of uniformly random numbers in the interval [0,1] is m; m is the number of samples. For each input parameter variable, the sample points are randomly arranged to construct an m×p sample matrix, where p is the number of variables. The independence between variables is ensured by random arrangement. The uniformity of the distribution of the sample matrix and the independence of the variables are verified to obtain m sets of parameter schemes.

[0009] Furthermore, in the single-well water control measure decision-making method based on mechanical compartmentation described in this invention, step S2 includes: The training sample library is preprocessed; A random forest model is constructed using the input parameters in the preprocessed training sample library as feature variables and the decrease in oil content and water content in the preprocessed training sample library as target variables. Based on the random forest model, multiple decision trees are constructed through Bootstrap sampling and random feature subset selection, and the feature importance score of each input parameter is calculated. Based on the ranking results of feature importance scores, an importance threshold is set to screen out the main controlling factors affecting the effectiveness of water control measures.

[0010] Furthermore, in the single-well water control measure decision-making method based on mechanical compartmentation described in this invention, the step of preprocessing the training sample library includes: Standardize continuous variables: Where x i Let be the original value of the i-th feature, μ be the feature mean, and σ be the standard deviation. These are the standardized eigenvalues; The categorical variables are one-hot encoded to convert them into a numerical form that the model can recognize, generating a standardized input feature vector.

[0011] Furthermore, in the single-well water control measure decision-making method based on mechanical compartmentation described in this invention, before performing SHAP value analysis in step S3, the method further includes: Hyperparameter tuning steps: Adaptive tuning of key hyperparameters of the constructed XGBoost prediction model is performed using the Bayesian optimization algorithm; a mapping relationship between hyperparameters and model performance is established through a Gaussian process surrogate model, and the expected improvement criterion is used to guide the parameter search direction; optimized parameters include learning rate and maximum depth; Model evaluation steps: The performance of the constructed XGBoost prediction model is evaluated using a k-fold cross-validation strategy. The mean squared error and coefficient of determination are calculated as evaluation metrics, and the model complexity is controlled by regularization techniques.

[0012] Furthermore, in the single-well water control measure decision-making method based on mechanical compartmentation described in this invention, the contribution of each main controlling factor to the effectiveness of the water control measure, as described in step S3, includes: Based on the optimal XGBoost prediction model after model evaluation, the SHAP values ​​of each controlling factor are calculated using Shapley value theory to quantify the contribution of each controlling factor to the effectiveness of water control measures. Among them, ϕ iThe SHAP value of the i-th feature is given by f; S is a subset of features; N(i) is the set of all features except for feature i; n is the total number of features; f S These are the model predictions based on the feature subset S.

[0013] Furthermore, in the single-well water control measure decision-making method based on mechanical compartmentation described in this invention, step S4, which recommends the optimal decision scheme based on the evaluation results, includes: The overall score for each scheme is calculated using a weighted scoring mechanism: Among them, S i Let i be the comprehensive score of the i-th scheme; , These are the weighting coefficients; Let T be the oil increase rate during the effective period of the measures of the i-th scheme; The percentage decrease in moisture content during the effective period of the measures for the i-th scheme; Compare the overall scores of each option and recommend the optimal water control measure.

[0014] In addition, the present invention also provides a single-well water control measure decision system based on mechanical compartmentation, comprising: The sample library construction module is used to build the training sample library; The main control factor screening module is used to screen the main control factors affecting the effectiveness of water control measures based on the training sample library and using the random forest algorithm. The XGBoost prediction model training module is used to construct XGBoost oil increase rate prediction model and XGBoost water cut decrease rate prediction model based on the selected main control factors, and combine the SHAP value to analyze the contribution of each main control factor to the effect of water control measures. The recommendation module is used to automatically evaluate the implementation effects of three water control measures—ICD+compartmental, AICD+compartmental, and compartmental—based on the trained XGBoost prediction model, by inputting target single-well parameters, and recommending the optimal decision-making scheme based on the evaluation results.

[0015] The single-well water control measure decision-making method and system based on mechanical compartmentation of the present invention has the following beneficial effects: The present invention can solve the technical problems of lack of systematicness, difficulty in comprehensively considering multi-dimensional factors, and inability to automatically recommend the optimal measures based on single-well parameters in the existing single-well water control measure decision-making. It can effectively improve the oil production rate, reduce the water cut, and ultimately achieve stable oil production and water control in single wells, thereby improving the overall oilfield recovery rate. Attached Figure Description

[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1This is a flowchart illustrating the decision-making method for single-well water control measures based on mechanical compartmentation provided in an embodiment of the present invention. Figure 2 This is a ranking diagram of the importance of various influencing factors evaluated by the random forest algorithm in some embodiments of the present invention; Figure 3 These are training effect diagrams of the XGBoost fuel efficiency prediction model in some embodiments of the present invention; Figure 4 These are training results of the XGBoost moisture content decrease prediction model in some embodiments of the present invention; Figure 5 This is an example diagram showing the calculation results display and user interface of the single-well water control measure decision system of the present invention; Figure 6 This is a schematic diagram of the structure of the single-well water control measure decision system based on mechanical compartmentation provided in an embodiment of the present invention. Detailed Implementation

[0017] To provide a clearer understanding of the technical features, objectives, and effects of the present invention, specific embodiments of the invention are now described in detail with reference to the accompanying drawings. In the following description, specific details such as particular system structures and techniques are set forth for illustrative purposes rather than for limiting the scope of the invention, in order to provide a thorough understanding of the embodiments. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.

[0018] In a preferred embodiment, reference Figure 1 The decision-making method for single-well water control measures based on mechanical compartmentation in this embodiment includes the following steps: S1. Constructing a training sample library. It can be understood that before step S1, the following steps are included: establishing an evaluation index system for the effectiveness of water control measures, with the oil increase rate and the decrease in water cut as the core. In other words, this embodiment establishes an evaluation index system with the oil increase rate and the decrease in water cut as the core, and constructs a training sample library to provide high-quality training data for subsequent machine learning models.

[0019] This embodiment establishes evaluation indicators with oil increase rate and water cut decrease rate as the core, combines oil increase rate to quantify production increase effect, and water cut decrease rate to evaluate water control effect, so as to achieve a comprehensive evaluation of the overall effect of water control measures for single wells.

[0020] Specifically, the calculation method for the fuel efficiency ratio is as follows: Among them, R o The oil production increase rate during the effective period T of the measure is expressed in %; Qe (t) represents the daily oil production under the implemented measures at time t, in m³. 3 / d;Q b (t) represents the daily oil production at time t under the scenario where no measures are implemented, in m³. 3 / d; T represents the effective period of the measure, in days.

[0021] The calculation method for the decrease in moisture content is as follows: in, The percentage decrease in moisture content during the effective period of the measures is expressed in % (%). This represents the average moisture content under the implemented measures during the effective period of the measures, expressed in % (%). The average moisture content under the scenario where no measures were implemented during the effective period of the measures, expressed in percentage.

[0022] In some embodiments, step S1 includes: S11, determining the input parameters and performing Latin hypercube sampling on the input parameters to generate a combination of sample points covering the parameter space. This step is based on actual oilfield data and selects key parameters affecting the effectiveness of water control measures. These parameters include, but are not limited to: ① reservoir parameters: effective thickness (unit: m, range 3-25m), porosity (unit: %, range 13-30%), permeability (unit: mD, range 50-4000mD), oil saturation (dimensionless, range 0.4-0.8), permeability variation coefficient (dimensionless, range 0-3), permeability gradient (dimensionless, range 1-25), rhythmicity (categorical variable, including positive rhythmicity, anti-rhythmicity, and homogeneous types), and original formation pressure (unit: MPa, 12-35). ① Development degree of interlayers (categorical variable, including three types: none, partially developed, and fully developed), water-sheltered height (unit: m, 3-15); ② Fluid parameters: formation crude oil viscosity (unit: mPa·s, range 1-50 mPa·s), surface crude oil density (unit: g / cm³, range 0.75-0.95 g / cm³); ③ Engineering parameters: initial production capacity - liquid (unit: m³ / d, range 30-2000 m³ / d), type of measure (categorical variable, including three types: ICD + compartment, AICD + compartment, and compartment), water-bearing stage of the measure (unit: %, range 40-95%).

[0023] Specifically, in this step, Latin hypercube sampling is performed on the input parameters to generate a combination of sample points covering the parameter space. This can be achieved by dividing the range of values ​​for each input parameter variable into m equally probable intervals, thus generating sample points. Where, x ij Let F be the value of the j-th sample point of the i-th variable.i-1 π is the inverse function of the cumulative distribution function of the i-th variable. j Let u be the j-th permutation position, where j = 1, ..., m. ij is a uniformly random number within the interval [0,1]. m is the number of samples.

[0024] For each input parameter variable, the sample points are randomly arranged to construct an m×p sample matrix, where p is the number of variables. The independence between variables is ensured by random arrangement. The uniformity of the distribution of the sample matrix and the independence of the variables are verified to obtain m sets of parameter schemes.

[0025] For example, in this step, the Latin hypercube sampling (LHS) method is used to obtain diverse schemes, generating a combination of sample points covering the parameter space. The LHS method ensures a uniform distribution of sample points in the 18-dimensional parameter space through stratified sampling and random combination, avoiding the sample clustering problem of traditional random sampling. Sample points are generated using the above formula, and a 748×18 sample matrix is ​​constructed by random arrangement, ultimately obtaining 748 sets of parameter schemes.

[0026] It should be noted that the method of constructing a training sample library based on Latin hypercube sampling combined with reservoir numerical simulation can be replaced by using orthogonal design or uniform design methods to generate parameter combinations, and then combining this with reservoir numerical simulation to calculate the effect of water control measures. This alternative approach can also generate representative samples in a multidimensional parameter space, providing a data foundation for subsequent modeling.

[0027] S12. Combining the implementation effects of three water control measures—ICD+compartmental, AICD+compartmental, and compartmentalized—in reservoir numerical simulation, the corresponding oil increase rate and water cut decrease are calculated, and these are combined with the generated parameter sample points to construct a training sample library. This step, based on parameter schemes obtained through Latin hypercube sampling, utilizes reservoir numerical simulation software to establish a reservoir numerical model for numerical simulation. The reservoir numerical simulation model considers the lateral heterogeneity of reservoir permeability. For example, a horizontal well is set within the formation, traversing different permeability distribution areas in the reservoir. The implementation effects of the three water control measures—ICD+compartmental, AICD+compartmental, and compartmentalized—are simulated respectively, and the corresponding oil increase rate and water cut decrease are calculated, which are then combined with the parameter scheme data to construct a training sample library.

[0028] S2. Based on the training sample library, the random forest algorithm is used to screen the main controlling factors affecting the effectiveness of water control measures. That is, the random forest method is used to screen the main controlling factors for the oil increase rate and the decrease in water content.

[0029] Specifically, step S2 includes: S21. Preprocess the training sample database. Specifically, standardize continuous variables: Where x i Let be the original value of the i-th feature, μ be the feature mean, and σ be the standard deviation. These are the standardized feature values.

[0030] One-hot encoding is performed on categorical variables to convert them into numerical forms recognizable by the model, generating standardized input feature vectors. This step standardizes continuous variables such as permeability, effective thickness, and porosity to ensure data quality and scale consistency, while one-hot encoding is performed on categorical variables such as rhythmicity, measure type, and interlayer development degree to convert them into numerical forms recognizable by the model.

[0031] S22. Using the input parameters in the preprocessed training sample library as feature variables and the decrease in oil content and water content in the preprocessed training sample library as target variables, construct a random forest model.

[0032] S23. Based on the random forest model, multiple decision trees are constructed through Bootstrap sampling and random feature subset selection, and the feature importance score of each input parameter is calculated.

[0033] S24. Based on the ranking results of feature importance scores, set importance thresholds to screen out the main controlling factors affecting the effectiveness of water control measures.

[0034] S3. Based on the selected main control factors, construct XGBoost models for predicting oil yield increase and water cut decrease, respectively, and analyze the contribution of each main control factor to the effectiveness of water control measures using SHAP values. This step involves using the main control factors as input features and the oil yield increase and water cut decrease as target variables to construct two independent XGBoost regression models, utilizing a gradient boosting framework to achieve high-precision prediction.

[0035] Specifically, refer to Figure 3 and Figure 4 In step S3, before parsing the SHAP value, the following steps are also included: Hyperparameter tuning steps: The key hyperparameters of the constructed XGBoost prediction model are adaptively tuned using a Bayesian optimization algorithm. A Gaussian process surrogate model is used to establish a mapping relationship between hyperparameters and model performance. The expected improvement criterion guides the parameter search direction, iterating until convergence or the maximum number of iterations is reached. Optimized parameters include learning rate, maximum depth, etc., such as maximum depth (2-4), learning rate (0.01-0.2), number of trees (100-300), minimum child node weight (5-20), gamma parameter (0.5-2.0), subsample ratio (0.6-0.9), feature sampling ratio (0.6-0.9), L1 regularization parameter (0.5-5.0), and L2 regularization parameter (1.0-5.0). For example, the number of trees can be set to 100, the maximum depth to 10, and the minimum number of split samples to 2. Multiple decision trees are constructed using Bootstrap sampling and random feature subset selection. Feature importance scores for each input parameter are calculated to identify the major controlling factors that significantly impact water control effectiveness, such as... Figure 2 As shown, Figure 2 The chart shows the order of importance of the main controlling factors for the decrease in oil content and water content.

[0036] Model evaluation steps: The performance of the constructed XGBoost prediction model is evaluated using a k-fold cross-validation strategy. The mean squared error and coefficient of determination are calculated as evaluation metrics, and the model complexity is controlled by regularization techniques.

[0037] Specifically, model performance can be evaluated using the coefficient of determination: Where y i This is the actual value. For predicted values, This is the mean of the actual values. This step prevents the model from overfitting and ensures that the model has good generalization ability and predictive stability on unknown data.

[0038] It is understandable that this method uses the established evaluation index system and the constructed training sample library to screen the main control factors using the random forest method, and inputs the screening results into the XGBoost prediction model for training. The model is then improved and perfected through Bayesian hyperparameter optimization to enhance prediction accuracy.

[0039] In some embodiments, step S3, which combines SHAP values ​​to analyze the contribution of each controlling factor to the effectiveness of water control measures, includes: Based on the optimal XGBoost prediction model after model evaluation, the SHAP values ​​of each controlling factor are calculated using Shapley value theory to quantify the contribution of each controlling factor to the effectiveness of water control measures. Among them, ϕ i f is the SHAP value of the i-th feature. S is a subset of features. N(i) is the set of all features except feature i. n is the total number of features. S These are model predictions based on a feature subset S. This embodiment calculates the SHAP values ​​of each controlling factor using Shapley value theory, providing interpretable support for decision-making regarding mechanical compartmentation measures.

[0040] S4. Based on the trained XGBoost prediction model, input the target single well parameters into it, automatically evaluate the implementation effect of three water control measures: ICD+compartment, AICD+compartment, and compartment, and recommend the optimal decision scheme based on the evaluation results.

[0041] In step S4, recommending the optimal decision-making scheme based on the evaluation results includes: The overall score for each scheme is calculated using a weighted scoring mechanism: Among them, S i This is the overall score for the i-th scheme. , These are the weighting coefficients. Let be the oil increase rate during the effective period T of the i-th scheme. Let represent the rate of decrease in moisture content during the effective period of the measures for the i-th scheme.

[0042] Compare the overall scores of each option and recommend the optimal water control measure.

[0043] In some embodiments, such as Figure 5 As shown, a user interface can be designed to receive 15 key parameters input by the user. The system automatically verifies the validity of these parameters to ensure the input data meets the model requirements. The user-input data is standardized to ensure consistency with the training data and prevent scale differences from affecting prediction accuracy. Using the trained XGBoost model, the system predicts the oil yield and water cut reduction for three water control measures: ICD+compartmentalization, AICD+compartmentalization, and compartmentalization. Based on the prediction results, a weighted scoring method is used to calculate the comprehensive score for each measure. The comprehensive scores of each measure are compared, and the measure with the highest score is selected as the recommended option. For example, for a target well implementing water control measures at a water cut of 90%, the system predicts the effects of the three options, setting weighting coefficients: oil yield weight... =0.7, weight of water content decrease =0.3, and the weighted scoring method was used to calculate the comprehensive score: cabin partitioning measure S1=1; ICD+cabin partitioning measure S2=0; AICD+cabin partitioning measure S3=0.84. After comparing the comprehensive scores, the cabin partitioning measure scored the highest, and the system recommends the cabin partitioning measure as the optimal solution.

[0044] Compared with existing single-well water control decision-making methods, this embodiment can fully utilize multi-dimensional geological reservoir and engineering data to accurately predict and recommend water control effects based on input single-well parameters. It solves the technical problem of lacking quantitative basis in traditional water control measure selection, providing a systematic scientific and technological means for oilfield stabilization, water control, and improved recovery. This method establishes a dual-objective evaluation index system with oil increase rate and water cut reduction rate as the core. It uses random forest to screen the main controlling factors, employs the XGBoost model to build a prediction model, and uses Bayesian optimization algorithm for hyperparameter tuning. Combined with SHAP value analysis, it improves model interpretability and constructs a fully intelligent decision-making system from sample generation and model building to measure recommendation, significantly improving the scientific nature and decision-making efficiency of water control measure selection.

[0045] In other embodiments, the method of using the random forest algorithm to screen the main controlling factors can be replaced by feature selection methods based on correlation analysis or principal component analysis. By calculating the correlation between each parameter and the rate of increase in oil content and the decrease in water cut, or by using principal component analysis to reduce dimensionality and identify the main influencing factors, it is also possible to screen out the key factors that significantly affect the water control effect from the 18 parameters. The method of building a prediction model based on XGBoost can be replaced by machine learning methods such as support vector regression or neural networks. These methods can also learn the nonlinear mapping relationship between the main controlling factors and the water control effect, and achieve accurate prediction of the effects of different water control measures.

[0046] In another preferred embodiment, reference Figure 6 The single-well water control measure decision system based on mechanical compartmentation in this embodiment includes: The sample library construction module is used to build the training sample library.

[0047] The main control factor screening module is used to screen the main control factors affecting the effectiveness of water control measures based on the training sample library and using the random forest algorithm.

[0048] The XGBoost prediction model training module is used to construct XGBoost oil increase rate prediction models and XGBoost water cut decrease rate prediction models based on the selected main control factors, and to analyze the contribution of each main control factor to the effectiveness of water control measures by combining SHAP values.

[0049] The recommendation module is used to automatically evaluate the implementation effects of three water control measures—ICD+compartmental, AICD+compartmental, and compartmental—based on the trained XGBoost prediction model, by inputting target single-well parameters, and recommending the optimal decision-making scheme based on the evaluation results.

[0050] This embodiment can solve the technical problems of existing single-well water control measures, such as lack of systematicness, difficulty in comprehensively considering multiple factors, and inability to automatically recommend the optimal measures based on single-well parameters. It can effectively improve the oil production rate, reduce the water cut, and ultimately achieve stable oil production and water control in single wells, thereby improving the overall oilfield recovery rate.

[0051] This method can also be integrated into a computer-readable storage medium. The computer-readable storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, external hard drive, read-only memory (ROM), magnetic disk, or optical disk.

[0052] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0053] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0054] It is understood that the above embodiments only illustrate preferred embodiments of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can freely combine the above technical features without departing from the concept of the present invention, and can also make several modifications and improvements, all of which fall within the protection scope of the present invention. Therefore, all equivalent transformations and modifications made with respect to the scope of the claims of the present invention should fall within the scope of the claims of the present invention.

Claims

1. A decision-making method for single-well water control measures based on mechanical compartmentation, characterized in that, The method includes the following steps: S1. Construct a training sample library; S2. Based on the training sample library, use the random forest algorithm to screen the main controlling factors that affect the effectiveness of water control measures; S3. Based on the selected main control factors, construct the XGBoost oil increase rate prediction model and the XGBoost water content decrease rate prediction model respectively, and analyze the contribution of each main control factor to the water control measures in combination with the SHAP value. S4. Based on the trained XGBoost prediction model, input the target single well parameters into it, automatically evaluate the implementation effects of three water control measures: ICD+compartment, AICD+compartment, and compartment, and recommend the optimal decision scheme based on the evaluation results.

2. The decision-making method for single-well water control measures based on mechanical compartmentation according to claim 1, characterized in that, The steps preceding step S1 also include: Establish an evaluation index system for the effectiveness of water control measures, with the reduction in oil content and water content as the core.

3. The decision-making method for single-well water control measures based on mechanical compartmentation according to claim 2, characterized in that, Step S1 includes: Determine the input parameters and perform Latin hypercube sampling on the input parameters to generate a combination of sample points covering the parameter space; By combining the implementation effects of three water control measures—ICD+compartmental, AICD+compartmental, and compartmentalized—in reservoir numerical simulation, the corresponding oil increase rate and water cut decrease rate are calculated, and these are combined with the generated parameter sample points to construct a training sample library.

4. The decision-making method for single-well water control measures based on mechanical compartmentation according to claim 3, characterized in that, The step of performing Latin hypercube sampling on the input parameters to generate a combination of sample points covering the parameter space includes: Divide the range of values ​​for each input parameter variable into m equally probable intervals, and generate sample points: Where, x ij Let F be the value of the j-th sample point of the i-th variable; i-1 π is the inverse function of the cumulative distribution function of the i-th variable; j Let j be the j-th permutation position, and j = 1, ..., m; u ij The number of uniformly random numbers in the interval [0,1] is m; m is the number of samples. For each input parameter variable, the sample points are randomly arranged to construct an m×p sample matrix, where p is the number of variables. The independence between variables is ensured by random arrangement. The uniformity of the distribution of the sample matrix and the independence of the variables are verified to obtain m sets of parameter schemes.

5. The decision-making method for single-well water control measures based on mechanical compartmentation according to claim 2, characterized in that, Step S2 includes: The training sample library is preprocessed; A random forest model is constructed using the input parameters in the preprocessed training sample library as feature variables and the decrease in oil content and water content in the preprocessed training sample library as target variables. Based on the random forest model, multiple decision trees are constructed through Bootstrap sampling and random feature subset selection, and the feature importance score of each input parameter is calculated. Based on the ranking results of feature importance scores, an importance threshold is set to screen out the main controlling factors affecting the effectiveness of water control measures.

6. The decision-making method for single-well water control measures based on mechanical compartmentation according to claim 5, characterized in that, The preprocessing of the training sample library in the step includes: Standardize continuous variables: Where x i Let be the original value of the i-th feature, μ be the feature mean, and σ be the standard deviation. These are the standardized eigenvalues; The categorical variables are one-hot encoded to convert them into a numerical form that the model can recognize, generating a standardized input feature vector.

7. The decision-making method for single-well water control measures based on mechanical compartmentation according to claim 1, characterized in that, Step S3, before parsing the SHAP value, also includes: Hyperparameter tuning steps: Adaptive tuning of key hyperparameters of the constructed XGBoost prediction model is performed using the Bayesian optimization algorithm; a mapping relationship between hyperparameters and model performance is established through a Gaussian process surrogate model, and the expected improvement criterion is used to guide the parameter search direction; the optimized parameters include learning rate and maximum depth. Model evaluation steps: The performance of the constructed XGBoost prediction model is evaluated using a k-fold cross-validation strategy. The mean squared error and coefficient of determination are calculated as evaluation metrics, and the model complexity is controlled by regularization techniques.

8. The decision-making method for single-well water control measures based on mechanical compartmentation according to claim 7, characterized in that, In step S3, the contribution of each controlling factor to the effectiveness of water control measures is analyzed by combining SHAP values, including: Based on the optimal XGBoost prediction model after model evaluation, the SHAP values ​​of each controlling factor are calculated using Shapley value theory to quantify the contribution of each controlling factor to the effectiveness of water control measures. Among them, ϕ i The SHAP value of the i-th feature is given by f; S is a subset of features; N(i) is the set of all features except for feature i; n is the total number of features; f S These are the model predictions based on the feature subset S.

9. The decision-making method for single-well water control measures based on mechanical compartmentation according to claim 1, characterized in that, The step S4, which recommends the optimal decision-making scheme based on the evaluation results, includes: The overall score for each scheme is calculated using a weighted scoring mechanism: Among them, S i Let i be the comprehensive score of the i-th scheme; , These are the weighting coefficients; Let T be the oil increase rate during the effective period of the measures of the i-th scheme; The percentage decrease in moisture content during the effective period of the measures for the i-th scheme; Compare the overall scores of each option and recommend the optimal water control measure.

10. A decision-making system for single-well water control measures based on mechanical compartmentation, characterized in that, include: The sample library construction module is used to build the training sample library; The main control factor screening module is used to screen the main control factors affecting the effectiveness of water control measures based on the training sample library and using the random forest algorithm. The XGBoost prediction model training module is used to construct XGBoost oil increase rate prediction model and XGBoost water cut decrease rate prediction model based on the selected main control factors, and combine the SHAP value to analyze the contribution of each main control factor to the effect of water control measures. The recommendation module is used to automatically evaluate the implementation effects of three water control measures—ICD+compartmental, AICD+compartmental, and compartmental—based on the trained XGBoost prediction model, by inputting target single-well parameters, and recommending the optimal decision-making scheme based on the evaluation results.

Images