Drilling fluid reservoir protection performance optimization method and system based on data correlation analysis

Abstract

The application provides a drilling fluid reservoir protection performance optimization method and system based on data correlation analysis, which comprises the following steps: performing attribution analysis on the correlation between initial drilling fluid parameters and target permeability recovery values; constructing a multiple regression model between all target drilling fluid controllable parameters and target permeability recovery values; calculating an optimal multiple regression equation based on the multiple regression model; generating parameter graph nodes and edges for each target drilling fluid controllable parameter to obtain a parameter graph structure of the target drilling fluid controllable parameters; combining the parameter graph structure and the optimal multiple regression equation to construct a parameter optimization graph network; inputting the initial drilling fluid parameters into the trained parameter optimization graph network to obtain optimal drilling fluid parameters output by the parameter optimization graph network; and adjusting the initial drilling fluid formula to an optimal drilling fluid formula according to the optimal drilling fluid parameters. The application has the effect of efficiently and accurately optimizing the reservoir protection performance of drilling fluid.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of reservoir protection in the petroleum industry, and particularly relates to a drilling fluid reservoir protection performance optimization method and system based on data correlation analysis. BACKGROUND

[0002] As a key fluid in drilling engineering, the performance of drilling fluid directly affects the efficiency, safety and economy of drilling. In the whole process of oil and gas drilling and well completion, the invasion of solid and liquid phases in the working fluid into the reservoir destroys the physical-chemical balance of the reservoir, damages the reservoir, and blocks the oil and gas flow channel, thereby making it difficult to discover or even impossible to discover oil and gas reservoirs and severely reducing the production of discovered oil and gas reservoirs. Therefore, reservoir protection has become a key to fully release the productivity of gas fields. However, when drilling fluid invades, it is easy to cause solid phase invasion and damage to the sensitivity of the reservoir. With the increase of drilling fluid solid content and the change of drilling fluid performance, the damage to the reservoir is aggravated.

[0003] In traditional drilling operations, the design and optimization of drilling fluid formulations mainly rely on the experience of field engineers and simple test methods. Engineers usually select a basic formulation according to the formation characteristics, drilling depth, temperature and pressure conditions, and then adjust the performance of the drilling fluid by adding various additives. This process often requires multiple tests and adjustments, which is time-consuming and costly. With the extension of oil and gas exploration and development to complex geological conditions and deep layers, the traditional method is increasingly difficult to meet the efficient and precise drilling needs. SUMMARY

[0004] The application provides a drilling fluid reservoir protection performance optimization method and system based on data correlation analysis to solve the problems of high cost and long time consumption in the drilling fluid reservoir protection performance optimization process.

[0005] In a first aspect, the application provides a drilling fluid reservoir protection performance optimization method based on data correlation analysis, which comprises the following steps:

[0006] An initial drilling fluid formulation and initial drilling fluid parameters of a target drilling fluid are obtained, and a target permeability recovery value corresponding to the target drilling fluid is determined according to the initial drilling fluid formulation and the initial drilling fluid parameters;

[0007] An attribution analysis is performed on the correlation between the initial drilling fluid parameters and the target permeability recovery value based on a preset parameter data set, and a target drilling fluid controllable parameter having a correlation impact on the target permeability recovery value is selected from all the initial drilling fluid parameters according to the attribution analysis result, the parameter data set containing multiple groups of drilling fluid parameters and multiple groups of permeability recovery values corresponding to the drilling fluid parameters;

[0008] regression model between all the target drilling fluid controllable parameters and the target permeability recovery value is constructed by taking the target permeability recovery value as the dependent variable and taking all the target drilling fluid controllable parameters as the independent variables;

[0009] An optimal multiple regression equation between all the target drilling fluid controllable parameters and the target permeability recovery value is calculated based on the multiple regression model and by using the parameter data set.

[0010] A parameter graph structure of the target drilling fluid controllable parameters is obtained by generating parameter graph nodes respectively for each of the target drilling fluid controllable parameters and by generating node features of the parameter graph nodes and node edges between the parameter graph nodes in combination with the attribution analysis result and the optimal multiple regression equation.

[0011] A parameter optimization graph network is constructed in combination with the parameter graph structure and the optimal multiple regression equation, and the parameter optimization graph network is trained by using the parameter data set until a difference between a permeability recovery value predicted by the parameter optimization graph network and an actual permeability recovery value reaches a minimum value.

[0012] The initial drilling fluid parameters are input into the trained parameter optimization graph network to obtain optimal drilling fluid parameters output by the parameter optimization graph network, and the initial drilling fluid formula is adjusted into an optimal drilling fluid formula according to the optimal drilling fluid parameters.

[0013] Optionally, the attribution analysis is performed on a correlation between the initial drilling fluid parameters and the target permeability recovery value based on a preset parameter data set, and target drilling fluid controllable parameters having a correlation with the target permeability recovery value are selected from all the initial drilling fluid parameters according to an attribution analysis result.

[0014] Drilling fluid controllable parameters associated with the target permeability recovery value are preliminarily screened from the initial drilling fluid parameters through a core flow experiment.

[0015] The data correlation degrees between all the drilling fluid controllable parameters and the target permeability recovery value are analyzed based on a preset parameter data set and in combination with a Pearson linear correlation analysis method and a Shapley nonlinear correlation analysis method, and the drilling fluid controllable parameters having a data correlation degree greater than a preset threshold value are taken as target drilling fluid controllable parameters.

[0016] Optionally, the target controllable drilling fluid parameter includes a basic drilling fluid parameter, a rheological parameter, a solid phase parameter, and a liquid phase parameter, the basic drilling fluid parameter includes a drilling fluid density and a linear expansion height, the rheological parameter includes an apparent viscosity, a plastic viscosity, a dynamic shear force, and a dynamic plasticity ratio, the solid phase parameter includes a solid phase content and a solid phase particle size, and the liquid phase parameter includes an HTHP fluid loss or an API fluid loss, and the liquid phase parameter further includes a pH value, a surface tension, and a cation concentration.

[0017] Optionally, the data correlation degrees between all the controllable drilling fluid parameters and the target permeability recovery value are analyzed based on the preset parameter data set and in combination with a Pearson linear correlation analysis method and a Shapley nonlinear correlation analysis method, and the controllable drilling fluid parameter with a data correlation degree greater than a preset threshold value is taken as a target controllable drilling fluid parameter, including the following steps:

[0018] A first Pearson correlation coefficient between each controllable drilling fluid parameter and the target permeability recovery value is calculated based on the preset parameter data set and by using the Pearson linear correlation analysis method.

[0019] If there are multiple linear correlation parameters with the same first Pearson correlation coefficient in the controllable drilling fluid parameters, then any one of the linear correlation parameters is retained and all other linear correlation parameters are excluded.

[0020] A Shapley value between each controllable drilling fluid parameter after the exclusion and the target permeability recovery value is calculated by using the Shapley nonlinear correlation analysis method.

[0021] If the Shapley value is greater than a preset Shapley threshold value, then the corresponding controllable drilling fluid parameter is taken as a target controllable drilling fluid parameter.

[0022] If the Shapley value is less than the Shapley threshold value, then the corresponding controllable drilling fluid parameter is excluded.

[0023] Optionally, for any one of the controllable drilling fluid parameters, a calculation formula of the first Pearson correlation coefficient is as follows:

[0024]

[0025] In the formula, R1 represents the first Pearson correlation coefficient, n represents a parameter data amount of the controllable drilling fluid parameter, xi represents a value of the i-th controllable drilling fluid parameter, i xi represents an average value of all the controllable drilling fluid parameters, yi represents the i-th permeability recovery value, i and y represents an average value of the permeability recovery values.

[0026] ​​Optionally, the calculation formula of the Shapley value is as follows:

[0027]

[0028] In the formula, φ j (val) represents the Shapley value of the drilling fluid controllable parameter x j , x j ∈X, X={x1, x2,..., x p}, S represents a subset formed by removing the drilling fluid controllable parameter x j from all the drilling fluid controllable parameters, |S| represents the number of parameters in the subset S, p represents the number of all the drilling fluid controllable parameters, val(S∪{x j}) represents the contribution generated by the alliance of the subset S and the drilling fluid controllable parameter x j , and val(S) represents the contribution generated by the subset S.

[0029] Optionally, the generation of the parameter graph structure of the target drilling fluid controllable parameter by respectively generating parameter graph nodes according to each target drilling fluid controllable parameter, generating node features of the parameter graph nodes in combination with the attribution analysis result and the optimal multiple regression equation, and generating node edges between the parameter graph nodes comprises the following steps:

[0030] Respectively generating parameter graph nodes according to each target drilling fluid controllable parameter, and generating node features of the parameter graph nodes according to target correlation factors between the target drilling fluid controllable parameter and the target permeability recovery value, the target correlation factors including an independent variable coefficient of the optimal multiple regression equation and a Shapley value calculated by the Shapley nonlinear correlation analysis method;

[0031] Calculating mutual information values between all the target drilling fluid controllable parameters based on the parameter data set;

[0032] Based on the Pearson linear correlation analysis method and according to the parameter data set, constructing node association edges between each parameter graph node, and determining node edge weights of the node association edges in combination with corresponding mutual information values, to obtain the parameter graph structure of the target drilling fluid controllable parameter.

[0033] Optionally, the generation of the parameter graph structure of the target drilling fluid controllable parameter by respectively generating parameter graph nodes according to each target drilling fluid controllable parameter, generating node features of the parameter graph nodes in combination with the attribution analysis result and the optimal multiple regression equation, and generating node edges between the parameter graph nodes comprises the following steps:

[0034] For any two target parameter graph nodes in all the parameter graph nodes, a second Pearson correlation coefficient between the two target parameter graph nodes is calculated according to drilling fluid parameter data corresponding to the two target parameter graph nodes in the parameter data set;

[0035] If an absolute value of the second Pearson correlation coefficient is less than a preset edge generation threshold, a node association edge between the two target parameter graph nodes is not constructed;

[0036] If the absolute value of the second Pearson correlation coefficient is greater than the edge generation threshold, the node association edge between the two target parameter graph nodes is constructed;

[0037] For each node association edge, a node edge weight of the node association edge is calculated according to the second Pearson correlation coefficient and the mutual information value of the two parameter graph nodes corresponding to the node association edge;

[0038] The parameter graph structure of the target drilling fluid controllable parameter is constructed in combination with all the parameter graph nodes and all the node association edges.

[0039] Optionally, the calculation of the node edge weight of the node association edge according to the second Pearson correlation coefficient and the mutual information value of the two parameter graph nodes corresponding to the node association edge includes the following steps:

[0040] The mutual information value of the two parameter graph nodes corresponding to the node association edge is normalized;

[0041] The absolute value of the second Pearson correlation coefficient and the normalized mutual information value are weighted and averaged to obtain the node edge weight of the node association edge.

[0042] In a second aspect, the present application further provides a drilling fluid reservoir protection performance optimization system based on data association analysis, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the drilling fluid reservoir protection performance optimization method based on data association analysis as described in the first aspect when executing the computer program.

[0043] The present application has the following advantages:

[0044] Compared with the prior art, the technical scheme of the present application has many advantages and significantly improves the efficiency and accuracy of drilling fluid formula optimization. Specifically, the advantages are as follows:

[0045] 1、The present application introduces a data-driven and machine learning method, which fully utilizes historical data and experience, greatly improving the scientificity and reliability of the optimization process. By establishing a parameter data set and conducting attribution analysis, this method can quickly identify key parameters that have a significant impact on permeability recovery value, avoiding the risk of ignoring some important factors in traditional methods. This data-based screening process not only improves the targeting of optimization, but also discovers some potential, non-intuitive influencing factors, providing a new approach for drilling fluid formulation optimization.

[0046] 2、The present application adopts an innovative method combining multiple regression models and graph networks, which can more comprehensively capture the complex relationships between parameters. Multiple regression models can reflect the quantitative influence of each parameter on permeability recovery value, while parameter graph structure further describes the interaction between parameters. This combination makes the optimization process no longer limited to the adjustment of a single parameter, but can consider the synergistic effect between parameters, resulting in more comprehensive and accurate optimization results.

[0047] 3、The present application realizes a complete closed loop from data to model to optimization by constructing and training parameter optimization graph networks. This method not only accurately predicts permeability recovery values under different parameter combinations, but also quickly recommends the optimal parameter configuration based on initial conditions. This greatly shortens the optimization cycle, improves efficiency, and reduces dependence on human experience, making the optimization process more objective and repeatable.

[0048] 4、This method has strong adaptability and scalability. As new data accumulates, the model can continuously learn and update, and its prediction and optimization capabilities will also continuously improve. This means that the method can adapt to different geological conditions and drilling requirements, providing strong support for drilling fluid formulation optimization in various complex situations.

[0049] In summary, this data-driven and machine learning-based drilling fluid formulation optimization method not only improves the efficiency and accuracy of optimization, but also provides a more intelligent and personalized solution for drilling engineering, which is expected to bring significant economic and technical benefits in improving drilling efficiency, reducing costs, and protecting oil and gas reservoirs. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 The flowchart of the drilling fluid reservoir protection performance optimization method based on data correlation analysis in one of the embodiments of the present application.

[0051] Figure 2 The Pearson correlation coefficient matrix of drilling fluid reservoir pollution associated factors in one of the embodiments of the present application.

[0052] Figure 3A Shapley value ranking diagram of controllable parameters of a drilling fluid in one embodiment of the present application. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present application will be clearly described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all of them. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art belong to the scope of protection of the present application.

[0054] The terms "first", "second", and the like in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally a category and do not limit the number of objects, for example, the first object can be one or more. In addition, "and / or" in the specification and claims means at least one of the connected objects, and the character " / " generally represents an "or" relationship between the front and rear associated objects.

[0055] Figure 1 A flowchart of a drilling fluid reservoir protection performance optimization method based on data correlation analysis in one embodiment. It should be understood that, although Figure 1 The steps in the flowchart are displayed in sequence according to the direction of the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps in the flowchart can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily sequential, but can be executed in rotation or alternation with other steps or at least part of the sub-steps or stages of other steps. As Figure 1 As shown in the flowchart, the drilling fluid reservoir protection performance optimization method based on data correlation analysis specifically includes the following steps:

[0056] S101. Obtain the initial drilling fluid formula and the initial drilling fluid parameters of the target drilling fluid, and determine the target permeability recovery value corresponding to the target drilling fluid according to the initial drilling fluid formula and the initial drilling fluid parameters.

[0057] The initial drilling fluid formulation typically includes base fluid type (e.g., water-based, oil-based, or synthetic-based), various additives (e.g., thickening agents, fluid loss additives, lubricants, etc.), and their proportions. The initial drilling fluid parameters include density, viscosity, water loss, pH value, and other physical and chemical properties. These data can be obtained through laboratory tests or field sampling analysis. For example, density can be measured with a mud balance, viscosity can be determined with a Marsh funnel or a rotational viscometer, and water loss can be determined through an API filtration test.

[0058] After obtaining these data, a permeability recovery experiment is conducted. The specific operation is to first saturate the core sample with crude oil and measure its initial permeability K1; then immerse the core in the target drilling fluid for a period of time (e.g., 24 hours) to simulate the process of drilling fluid contacting the reservoir rock; finally, reverse-displace the drilling fluid with crude oil and measure the terminal permeability K2 of the core. The permeability recovery value calculation formula is: permeability recovery value = (K2 / K1) x 100%. For example, the initial permeability K1 of a core sample is 10 mD, after drilling fluid immersion and reverse displacement, the terminal permeability K2 is 8 mD, then the permeability recovery value is (8 / 10) x 100% = 80%. The purpose of this step is to establish the relationship between the initial drilling fluid formulation and its reservoir protection performance (represented by the permeability recovery value). Through this method, the degree of influence of the current drilling fluid formulation on the reservoir can be quantitatively evaluated, providing a benchmark for subsequent optimization.

[0059] S102. Perform attribution analysis on the correlation between the initial drilling fluid parameters and the target permeability recovery value based on the preset parameter data set, and select the target drilling fluid controllable parameters that have a correlation impact on the target permeability recovery value from all the initial drilling fluid parameters according to the attribution analysis result.

[0060] In this step, first, a preset parameter data set is needed. This data set should contain multiple groups of drilling fluid parameters and their corresponding permeability recovery values. For example, there can be 100 groups of drilling fluid with different formulations, each group containing parameters such as density, viscosity, pH value, and solid content, as well as the corresponding permeability recovery values obtained through experiments. This data set can come from historical data, laboratory test results, or field application data. Next, use this data set to perform attribution analysis on the correlation between the initial drilling fluid parameters and the target permeability recovery value. The purpose of attribution analysis is to identify which parameters have a significant impact on the permeability recovery value. Here, various statistical methods can be used, such as correlation analysis, principal component analysis (PCA), or random forest feature importance.

[0061] For example, using correlation analysis, the Pearson correlation coefficient between each parameter and the permeability recovery value can be calculated. The correlation coefficient ranges from -1 to 1, with an absolute value closer to 1 indicating a stronger correlation. For example, suppose the correlation coefficients of some parameters with the permeability recovery value are as follows:

[0062] Correlation coefficient of density and permeability recovery value: -0.75

[0063] Correlation coefficient of viscosity and permeability recovery value: -0.60

[0064] Correlation coefficient of pH value and permeability recovery value: 0.30

[0065] Correlation coefficient of solid content and permeability recovery value: -0.80

[0066] According to the above correlation coefficients, it can be preliminarily judged that the solid content and density have a greater impact on the permeability recovery value.

[0067] To consider the possible nonlinear relationship, nonlinear correlation indicators such as mutual information can also be used. The greater the mutual information value, the higher the degree of mutual dependence between the two variables. Based on the attribution analysis, a threshold is set to filter out parameters with significant correlations. For example, parameters with an absolute correlation coefficient greater than 0.5 or a mutual information value in the top 50% can be selected as target drilling fluid controllable parameters. The implementation effect of this step is to filter out the key parameters that have a significant impact on the permeability recovery value from the numerous drilling fluid parameters. This not only simplifies the subsequent optimization process, but also helps engineers better understand the relationship between drilling fluid performance and reservoir protection. For example, if it is found that the solid content is the most critical factor, then in the subsequent optimization, attention can be focused on how to control and adjust the solid content. Through this data-driven method, the bias caused by subjective judgment can be avoided, and the pertinence and efficiency of optimization can be improved. At the same time, this method also has strong adaptability and can be flexibly adjusted according to different reservoir characteristics and drilling fluid systems.

[0068] S103. The target permeability recovery value is taken as the dependent variable and all target drilling fluid controllable parameters are taken as the independent variables, and a multiple regression model between all target drilling fluid controllable parameters and the target permeability recovery value is constructed.

[0069] where the target permeability recovery value is taken as the dependent variable and all selected target drilling fluid controllable parameters are taken as the independent variables, a multiple regression model is constructed. This model aims to quantify the influence of each parameter on the permeability recovery value and establish a mathematical expression that can predict the permeability recovery value. The general form of the multiple regression model is:

[0070] Y = β0 + β1X1 + β2X2 + … + βnXn + ε

[0071] where Y is the dependent variable (permeability recovery value), X1, X2, …, Xn are the independent variables (each target drilling fluid controllable parameter), β0 is the intercept, β1, β2, …, βn are the regression coefficients of each independent variable, and ε is the error term.

[0072] The construction process of the above multiple regression model usually includes the following steps:

[0073] Data preparation: The selected target drilling fluid controllable parameters and corresponding permeability recovery values are arranged into a format suitable for regression analysis.

[0074] Model hypothesis testing: Check if the data meets the basic assumptions of multiple regression, such as linear relationship, independence of error terms, homoscedasticity and normal distribution, etc.

[0075] Multiple collinearity check: Calculate the correlation coefficient or variance inflation factor (VIF) between independent variables to ensure that there is no serious multiple collinearity between independent variables.

[0076] Model fitting: Estimate the regression coefficients using the least squares method or other appropriate methods.

[0077] Model evaluation: Evaluate the goodness of fit and significance of the model through indicators such as the coefficient of determination (R 2 ), adjusted R 2 , F test, etc.

[0078] Residual analysis: Check the distribution of residuals to ensure that the model meets the basic assumptions.

[0079] The implementation effect of this step is to obtain a mathematical model that can quantitatively describe the relationship between drilling fluid parameters and permeability recovery values. This model not only can be used to predict the permeability recovery value under different parameter combinations, but also can intuitively reflect the influence direction and degree of each parameter through the size and sign of the regression coefficient.

[0080] For example, if the obtained multiple regression model is:

[0081] Permeability recovery value = 120 - 0.05 × density - 0.02 × viscosity - 0.1 × solid content + 0.03 × pH value;

[0082] It indicates that the increase of density, viscosity and solid content will lead to the decrease of permeability recovery value, while the increase of pH value will slightly improve the permeability recovery value. At the same time, from the absolute value of the coefficient, it can be seen that the influence of solid content is the largest, followed by density.

[0083] S104. Calculate the optimal multiple regression equation between all target drilling fluid controllable parameters and target permeability recovery value based on the multiple regression model and using the parameter dataset.

[0084] In this step, the optimal multiple regression equation is calculated based on the previously constructed multiple regression model and using the parameter dataset. This process is essentially a further optimization and refinement of the initial regression model, aiming to obtain a mathematical expression that most accurately describes the relationship between drilling fluid parameters and permeability recovery value. To evaluate the performance of this optimal equation, indicators such as the determination coefficient (R 2 ), root mean square error (RMSE), etc. can be calculated. For example: the implementation effect of this step is to obtain an optimized and more accurate mathematical model, so as to more accurately predict the permeability recovery value under given drilling fluid parameters. And it can quickly evaluate the potential impact of different drilling fluid formulations, identify the most effective parameter adjustment direction to improve permeability recovery value, and find a balance point among multiple targets, such as ensuring sufficient permeability recovery value while minimizing the use of some expensive additives.

[0085] S105. Generate parameter graph nodes for each target drilling fluid controllable parameter, and generate node features of parameter graph nodes and node edges between parameter graph nodes based on the attribution analysis results and the optimal multiple regression equation, to obtain the parameter graph structure of the target drilling fluid controllable parameters.

[0086] First, generate the corresponding parameter graph nodes for each target drilling fluid controllable parameter. These parameters include key indicators such as density, viscosity, fluid loss, pH value, etc. When generating nodes, the characteristics and influence range of each parameter need to be considered. For example, the density node may contain information such as numerical range, unit, etc.; the viscosity node may contain viscosity values at different shear rates, etc. Next, combine the attribution analysis results conducted previously to enrich the features of these nodes. Attribution analysis may reveal the contribution of each parameter to drilling fluid performance, and this information can be used as an important attribute of the node. For example, if attribution analysis shows that density has the greatest impact on permeability recovery, then the density node may be assigned a higher weight or priority.

[0087] Meanwhile, the optimal multiple regression equation is used to further define the node features. Specifically, the coefficient of each parameter in the equation can serve as an important feature of the parameter node, reflecting the degree of influence of the parameter on the target variable. After defining the node features, the next step is to determine the edges between nodes. These edges represent the mutual relationship and influence between parameters. The properties of the edges can be defined by analyzing the correlation and interaction between parameters. For example, if a strong correlation is found between density and viscosity, the edge connecting these two nodes may be assigned a higher weight. The direction of the edge can be determined according to the causal relationship. If it can be determined that the change of a parameter will lead to the change of another parameter, a directed edge can be used to represent it.

[0088] Finally, all this information is integrated to form a complete parameter graph structure. This graph structure not only contains the features of each parameter, but also embodies the complex relationship network between parameters. For example, an adjacency matrix can be used to represent this graph structure, and the elements in the matrix represent the weights or other properties of the edges between nodes. The parameter graph structure constructed by this method provides a strong foundation for the subsequent optimization process. It not only captures the independent characteristics of each parameter, but also reflects the interaction between parameters, thus being able to more comprehensively describe the complexity of the drilling fluid system. This structure enables the subsequent optimization algorithm to better understand and utilize the relationships between parameters, thus finding a more optimal drilling fluid formulation.

[0089] S106. Construct a parameter optimization graph network in combination with the parameter graph structure and the optimal multiple regression equation, train the parameter optimization graph network with the parameter dataset until the difference between the permeability recovery value predicted by the parameter optimization graph network and the actual permeability recovery value reaches a minimum value.

[0090] Wherein, after obtaining the parameter graph structure, the next step is to construct a parameter optimization graph network in combination with this structure and the optimal multiple regression equation. The goal of this network is to accurately predict the permeability recovery value and ultimately be used to optimize the drilling fluid parameters. The construction process first needs to convert the parameter graph structure into a trainable neural network model. Each parameter node can be regarded as a neuron in the network, and the edges between nodes correspond to the connections between neurons. The features of the nodes can be used as the initial weights or biases of the neurons, and the properties of the edges can be used to initialize the connection weights. The optimal multiple regression equation plays a key role in this process, which can be used to design the loss function of the network, ensuring that the output of the network is consistent with the prediction result of the regression equation. Specifically, the last layer of the network can be designed as a linear activation function, so that its output form is consistent with the regression equation.

[0091] The training process of the network follows. The network is trained using a dataset of parameter combinations and their corresponding actual permeability recovery values. The training process employs a backpropagation algorithm that continuously adjusts the weights and biases in the network to minimize the difference between the predicted and actual values.

[0092] Specifically, the Mean Squared Error (MSE) can be used as the loss function:

[0093]

[0094] where y i is the actual permeability recovery value, is the predicted permeability recovery value by the network, and n is the number of samples.

[0095] During the training process, optimization algorithms such as gradient descent can be used to update the network parameters. For example, using the Stochastic Gradient Descent (SGD) algorithm, the formula for updating parameters at each iteration is:

[0096]

[0097] where θ represents the network parameters, α is the learning rate, is the gradient of the loss function with respect to the parameters.

[0098] The training process needs to be iterated continuously until the difference between the predicted and actual values of the network reaches a pre-set threshold. The choice of this threshold needs to balance the accuracy of the model and the risk of overfitting. For example, when the validation set loss does not decrease for consecutive epochs, training can be stopped. To improve the generalization ability of the model, some regularization techniques can be used, such as L1 / L2 regularization, Dropout, etc. At the same time, cross-validation is used to evaluate the performance of the model to ensure that the model also performs well on unseen data.

[0099] S107. The initial drilling fluid parameters are input into the trained parameter optimization graph network to obtain the optimal drilling fluid parameters output by the parameter optimization graph network, and the initial drilling fluid formula is adjusted to the optimal drilling fluid formula according to the optimal drilling fluid parameters.

[0100] where, after the training of the parameter optimization graph network is completed, the next step is to use this network to optimize the drilling fluid formula. First, the initial drilling fluid parameters are converted into a format suitable for network input. This first requires standardization or normalization of the data to ensure that all input parameters are on the same scale. For example, min-max normalization can be used:

[0101]

[0102] where X is the normalized value, x is the original parameter value, x min and x max are the minimum and maximum values of the parameter in the training data, respectively.

[0103] Next, the normalized parameters are input into the trained parameter optimization graph network. The network, based on these inputs, calculates a series of forward propagations and finally outputs the predicted permeability recovery value and the corresponding optimal parameter combination. To find the optimal parameter combination, the gradient ascent method can be used. Specifically, the weights of the network are fixed, but the input parameters are allowed to change, and then the gradient of the output (permeability recovery value) with respect to the input parameters is calculated. According to this gradient, the input parameters are gradually adjusted to maximize the output value. During the optimization process, physical constraints of the parameters need to be considered. For example, density cannot be negative, and pH value is usually within a certain range. Constraints can be added or appropriate activation functions can be used to ensure that the optimization results conform to physical reality.

[0104] The optimization process may require multiple iterations until the permeability recovery value no longer increases significantly or reaches a preset number of iterations. After each iteration, check whether the new parameter combination is better than the previous one. A threshold can be used to determine whether the improvement is significant, for example:

[0105]

[0106] where ∈ is a small positive number, such as 0.001.

[0107] After obtaining the optimal parameters, they need to be converted back to the original scale. Finally, adjust the initial drilling fluid formulation based on the obtained optimal parameters. This process needs to consider the feasibility and economy of actual operation. For example, if the optimal density is higher than the initial density, you can increase the weight material such as barite to increase the density; if the optimal pH value is lower than the initial value, you can add an appropriate amount of acidic substances to adjust. During the adjustment process, the mutual influence between parameters needs to be considered. For example, increasing a certain additive may affect both density and viscosity. Therefore, when adjusting the formulation, multiple fine-tuning may be required, and after each adjustment, key parameters need to be tested to ensure they meet the optimization results. Through this process, the initial drilling fluid formulation is optimized to maximize the permeability recovery. This method combines data-driven machine learning techniques and traditional drilling fluid formulation adjustment experience, which can more accurately and efficiently find the optimal formulation, thereby improving drilling efficiency and reducing drilling risks.

[0108] In one embodiment, the correlation between the initial drilling fluid parameters and the target permeability recovery value is attributed based on a preset parameter data set, and the target drilling fluid controllable parameters having a correlation with the target permeability recovery value are screened from all the initial drilling fluid parameters according to the attribution analysis result, including the following steps:

[0109] The drilling fluid controllable parameters associated with the target permeability recovery value are initially screened from the initial drilling fluid parameters through a core flow experiment;

[0110] Based on the preset parameter data set and in combination with the Pearson linear correlation analysis method and the Shapley nonlinear correlation analysis method, the data correlation degrees between all the drilling fluid controllable parameters and the target permeability recovery value are analyzed, and the drilling fluid controllable parameters having a data correlation degree greater than a preset threshold are taken as the target drilling fluid controllable parameters.

[0111] In the present embodiment, the core flow experiment is an important method for simulating the interaction between the drilling fluid and the rock under formation conditions, and is used for initially screening the drilling fluid controllable parameters associated with the target permeability recovery value. The experimental process first needs to prepare representative core samples, which are usually obtained from the target formation or artificially made to be similar to the target formation in properties. The core samples need to be pretreated, including steps of cleaning, drying, and saturation, to ensure the consistency and repeatability of the experimental conditions.

[0112] The experimental device usually includes a core holder, a pressure vessel, a pump system, a pressure sensor, and a flow meter, etc. The core is placed in the pressure vessel simulating the formation pressure and temperature, and the drilling fluid of different formulations is injected into the core at a specific flow rate through the pump system. During the injection process, the pressure difference and the flow rate change at the inlet and outlet are continuously monitored, and these data are used to calculate the permeability change of the core. The experimental process usually includes the following stages: first, the initial permeability of the core is measured, then the drilling fluid is injected to simulate the drilling process, after that, the core is flushed with a cleaning fluid (such as brine) to simulate the production process, and finally, the core permeability is measured again.

[0113] In the experiment, the effects of different parameters of the drilling fluid (such as density, viscosity, pH value, and solid content) on the permeability recovery value are observed by changing them. For example, other parameters can be kept unchanged, and only the drilling fluid density is changed from 1.0 g / cm 3 to 1.5 g / cm 3 , with an increase of 0.1 g / cm 3The permeability recovery value changes under different densities are observed. Similarly, the viscosity can be changed from 10 mPa·s to 50 mPa·s, the pH value can be changed from 7 to 11, etc., and the changes in the permeability recovery value are recorded respectively. Through the parameter change experiment of this system, it can be preliminarily identified which parameters have a significant impact on the permeability recovery value. For example, if it is found that the permeability recovery value presents a significant downward trend with the increase of the density, it can be preliminarily judged that the density is an important related parameter. Conversely, if the change of a certain parameter has little effect on the permeability recovery value, this parameter may not be a key factor.

[0114] Through the core flow experiment, a group of parameters that may be related to the permeability recovery value, i.e., the controllable parameters of the drilling fluid associated with the target permeability recovery value, can be screened from the initial numerous drilling fluid parameters. After completing the preliminary screening of the core flow experiment, the next step is to conduct a more in-depth and systematic data correlation analysis on all possible related drilling fluid controllable parameters. Two complementary analysis methods, Pearson linear correlation analysis and Spearman nonlinear correlation analysis, are used in this step to comprehensively evaluate the relationship between the parameters and the target permeability recovery value. Before this, analysis needs to be based on a preset parameter data set. This data set usually contains a large amount of historical data or experimental data, covering various combinations of drilling fluid parameters and their corresponding permeability recovery values. The quality of the data set directly affects the reliability of the analysis results, so the data needs to be preprocessed, including removing outliers, handling missing data, standardization, etc.

[0115] Pearson linear correlation analysis is used to evaluate the strength of the linear relationship between the parameters and the permeability recovery value, and the value of the correlation coefficient r is between -1 and 1. The closer the absolute value is to 1, the stronger the linear correlation is, and the positive value indicates a positive correlation and the negative value indicates a negative correlation. However, in practice, there may be complex nonlinear relationships between parameters and permeability recovery values. Therefore, Spearman nonlinear correlation analysis is introduced to supplement. The higher the Spearman value, the greater the contribution of the parameter to the prediction of the permeability recovery value. Combining the results of the two analysis methods, the correlation between each parameter and the permeability recovery value can be comprehensively evaluated. A preset threshold is usually set, for example, the absolute value of the Pearson correlation coefficient is greater than 0.5 or the Spearman value is greater than 0.1, and the parameters that meet the conditions are identified as the target drilling fluid controllable parameters. The advantage of this double analysis method is that it considers both simple linear relationships and does not ignore possible complex nonlinear relationships. Through this method, the truly important drilling fluid parameters can be more accurately identified, providing a reliable basis for the subsequent optimization process.

[0116] In one embodiment, based on the preset parameter data set, in combination with Pearson linear correlation analysis method and Shapley nonlinear correlation analysis method, data correlation between all controllable drilling fluid parameters and target permeability recovery value is analyzed, and the controllable drilling fluid parameter with data correlation greater than a preset threshold is taken as the target controllable drilling fluid parameter, including the following steps:

[0117] Based on the preset parameter data set, the first Pearson correlation coefficient between each controllable drilling fluid parameter and the target permeability recovery value is calculated by using the Pearson linear correlation analysis method.

[0118] If there are multiple linear correlation parameters with the same first Pearson correlation coefficient among the controllable drilling fluid parameters, any one linear correlation parameter is retained and all other linear correlation parameters are excluded.

[0119] The Shapley value between each controllable drilling fluid parameter after the exclusion and the target permeability recovery value is calculated by using the Shapley nonlinear correlation analysis method.

[0120] If the Shapley value is greater than a preset Shapley threshold, the corresponding controllable drilling fluid parameter is taken as the target controllable drilling fluid parameter.

[0121] If the Shapley value is less than the Shapley threshold, the corresponding controllable drilling fluid parameter is excluded.

[0122] In this embodiment, for any controllable drilling fluid parameter, the calculation formula of the first Pearson correlation coefficient is as follows:

[0123]

[0124] In the formula, R1 represents the first Pearson correlation coefficient, n represents the parameter data amount of the controllable drilling fluid parameter, xi represents the value of the i th controllable drilling fluid parameter, x represents the average value of all controllable drilling fluid parameters, yi represents the i th permeability recovery value, and y represents the average value of the permeability recovery value. i i

[0125] In this embodiment, the calculation formula of the Shapley value is as follows:

[0126]

[0127] In the formula, φ (val) represents the Shapley value of the controllable drilling fluid parameter x, x ∈ X, X = {x1, x2,..., x j j j p}, and S represents the set of all controllable drilling fluid parameters except x j ​​​​​​​|S| represents the number of parameters in the subset S, p represents the number of all controllable drilling fluid parameters, val(S∪{x j}) represents the contribution of the subset S and the controllable drilling fluid parameter x j to form the alliance, and val(S) represents the contribution of the subset S.

[0128] In one embodiment, all the above embodiments are based on and exemplified for deep water water-based drilling fluid reservoir contamination attribution analysis:

[0129] The controllable drilling fluid parameters are selected from the initial numerous drilling fluid parameters through core flow experiments, including drilling fluid basic parameters, rheological parameters, solid phase parameters and liquid phase parameters. The drilling fluid basic parameters include drilling fluid density and linear expansion height, the rheological parameters include apparent viscosity, plastic viscosity, dynamic shear force, dynamic plasticity ratio and static shear force, the solid phase parameters include solid phase content and solid phase particle size, and the liquid phase parameters include API fluid loss, HTHP fluid loss, pH value, surface tension and cation concentration.

[0130] The parameter data set used in this embodiment is the performance parameters of drilling fluids with a density of 1.5 g / cm3 to 2.0 g / cm3. All parameters are standardized to eliminate the influence of original data types on the analysis process. The solid phase content and cation concentration are directly taken as actual values represented by decimals. The permeability recovery value, fluid loss, inhibition, rheological performance, etc. are processed by linear normalization method. The processed data set data is shown in Table 1 below.

[0131] Table 1 Controllable drilling fluid parameters and permeability recovery values of drilling fluids with a density of 1.5 g / cm3 to 2.0 g / cm3

[0132]

[0133]

[0134] According to the data in Table 1 above, the attribution analysis is performed. First, the first Pearson correlation coefficient between each controllable drilling fluid parameter and the permeability recovery value is calculated, as shown in Table 2. Figure 2 Figure 2 The first column or the first row in Table 2 is the first Pearson correlation coefficient. Among them, there are 6 factors positively correlated with the reservoir protection performance (i.e. the permeability recovery value) (the first Pearson correlation coefficient is positive). Among the negative correlation factors (the first Pearson correlation coefficient is negative), the API fluid loss and the HTHP fluid loss are both -0.85, which can be regarded as linear correlation factors and contain repeated information. Therefore, the API fluid loss or the HTHP fluid loss is randomly removed. In this embodiment, the API fluid loss is removed.

[0135] ​Then the drilling fluid controllable parameters are subjected to a Shapley nonlinear correlation analysis, and the Shapley values of the drilling fluid controllable parameters are calculated, as shown in Table 1. Figure 3 In the present embodiment, the Shapley threshold value is preset to 0.2, and therefore the static shear force (GEL) with a Shapley value less than 0.2 is screened out. After screening, the target drilling fluid controllable parameters include drilling fluid basic parameters, rheological parameters, solid phase parameters and liquid phase parameters. The drilling fluid basic parameters include drilling fluid density and linear expansion height, the rheological parameters include apparent viscosity, plastic viscosity, dynamic shear force and dynamic plasticity ratio, the solid phase parameters include solid phase content and solid phase particle size, and the liquid phase parameters include HTHP filtration loss, pH value, surface tension and cation concentration.

[0136] A multivariate regression model is constructed with the target permeability recovery value as the dependent variable and all the screened target drilling fluid controllable parameters as the independent variables. The optimal multivariate regression equation between all the target drilling fluid controllable parameters and the target permeability recovery value is calculated using the parameter data set in Table 1 above:

[0137] Kr = 0.4367σ + 0.8799AV - 0.4594Gu - 0.8148HTHP + 0.6184Inh - 0.7325pH

[0138] + 0.6362Li + 0.5684DPR - 0.3219YP + 0.2384PV - 0.1580ρ + 0.1426Kation

[0139] In the formula, Kr is the target permeability recovery value; σ is the surface tension; AV is the apparent viscosity; Gu is the solid phase content; HTHP is the high temperature and high pressure filtration loss; Inh is the linear expansion height; pH is the pH value; Li is the solid phase average diameter; DPR is the dynamic plasticity ratio; YP is the dynamic shear force; PV is the plastic viscosity; ρ is the drilling fluid density; and Kation is the cation concentration.

[0140] In one embodiment, parameter graph nodes are generated for each target drilling fluid controllable parameter, and the node features of the parameter graph nodes and the node edges between the parameter graph nodes are generated in combination with the attribution analysis results and the optimal multivariate regression equation, so as to obtain the parameter graph structure of the target drilling fluid controllable parameters, including the following steps:

[0141] Parameter graph nodes are generated for each target drilling fluid controllable parameter, and the node features of the parameter graph nodes are generated according to the target correlation factors between the target drilling fluid controllable parameters and the target permeability recovery value. The target correlation factors include the independent variable coefficients of the optimal multivariate regression equation and the Shapley values calculated by the Shapley nonlinear correlation analysis method;

[0142] The mutual information values between all the target drilling fluid controllable parameters are calculated based on the parameter data set.

[0143] Based on the Pearson linear correlation analysis method and according to the parameter data set, node association edges between each parameter graph node are constructed, and the node edge weight of the node association edge is determined in combination with the corresponding mutual information value, to obtain the parameter graph structure of the target drilling fluid controllable parameter.

[0144] In the present embodiment, for each target drilling fluid controllable parameter, a corresponding parameter graph node is created. The characteristics of these nodes consist of two parts: the independent variable coefficient of the optimal multiple regression equation and the Sharpley value calculated by the Sharpley nonlinear correlation analysis method. The optimal multiple regression equation is obtained by analyzing the historical data to find the parameter combination that can best predict the target permeability recovery value. For example, suppose the obtained regression equation is Y = 0.5X1 + 0.3X2 - 0.2X3 + 0.1X4, where Y is the target permeability recovery value, X1, X2, X3, X4 represent different drilling fluid parameters, then these coefficients (0.5, 0.3, -0.2, 0.1) are used as a characteristic of the corresponding node. At the same time, the contribution of each parameter to the target permeability recovery value is calculated by using the Sharpley nonlinear correlation analysis method. The Sharpley value reflects the average marginal contribution of the parameter to the result in different combinations, and its calculation involves all possible parameter combinations. For example, for parameter X1, its Sharpley value may be 0.4, indicating that X1 contributes an average of 40% of the prediction ability when considering all possible parameter combinations. These Sharpley values are another characteristic of the node. In this way, each parameter graph node contains two key indicators reflecting its importance and influence, laying a foundation for subsequent graph structure analysis.

[0145] Next, the mutual information value between all target drilling fluid controllable parameters is calculated based on the parameter data set. Mutual information is a method to measure the degree of mutual dependence between two random variables, which can capture nonlinear relationships, so it is more comprehensive than simple correlation coefficients. For each pair of parameters, for example, density (X) and API fluid loss (Y), the calculation formula of its mutual information I(X;Y) is:

[0146] I(X;Y) = ΣΣp(x,y)*log(p(x,y) / (p(x)p(y))

[0147] where p(x, y) is the joint probability distribution of X and Y, and p(x) and p(y) are the marginal probability distributions of X and Y, respectively. In practical calculations, histogram methods or kernel density estimation can be used to approximate these probability distributions. For example, assuming there are 100 groups of data, the value range of each parameter is divided into 10 equal intervals, and then the frequency of each interval is counted to estimate the probability distribution. The larger the calculated mutual information value, the higher the degree of mutual dependence between the two parameters. For example, if the mutual information value between density and API fluid loss is 0.8, and the mutual information value between density and apparent viscosity is 0.3, it means that the relationship between density and API fluid loss is more closely related than that between density and apparent viscosity. These mutual information values will play an important role in subsequent construction of parameter graph edge weights, helping to identify important associations between parameters.

[0148] Finally, based on the Pearson linear correlation analysis method, the association edges between the nodes of the parameter graph are constructed, and the weights of the edges are determined in combination with the mutual information values, thereby obtaining the complete parameter graph structure. For each pair of parameters, if the absolute value of the Pearson correlation coefficient exceeds a predetermined threshold (e.g., 0.5), an edge is added between the nodes corresponding to the two parameters. The weight of the edge is determined by the absolute value of the correlation coefficient and the mutual information value calculated earlier, and a weighted average can be used, for example, if the correlation coefficient between density and API fluid loss is -0.7 and the normalized mutual information value is 0.8, and α = 0.6, then the weight of the edge between them is 0.6*0.7 + 0.4*0.8 = 0.74. In this way, the final parameter graph not only reflects the linear relationship between parameters, but also contains nonlinear mutual dependence information, providing a comprehensive description of parameter relationships for subsequent analysis and optimization. This parameter graph structure can intuitively display the complex relationship network between various drilling fluid parameters, helping to identify key parameters and parameter groups, and providing an important basis for drilling fluid formulation optimization and performance prediction.

[0149] In one embodiment, based on the Pearson linear correlation analysis method and according to the parameter data set, node association edges between the nodes of the parameter graph are constructed, and node edge weights of the node association edges are determined in combination with the corresponding mutual information values, to obtain the parameter graph structure of the controllable parameters of the target drilling fluid, including the following steps:

[0150] For any two target parameter graph nodes in all parameter graph nodes, according to the drilling fluid parameter data corresponding to the two target parameter graph nodes in the parameter data set, a second Pearson correlation coefficient between the two target parameter graph nodes is calculated;

[0151] If the absolute value of the second Pearson correlation coefficient is less than a predetermined edge generation threshold, no node association edge between the two target parameter graph nodes is constructed;

[0152] If the absolute value of the second Pearson correlation coefficient is greater than the edge generation threshold, a node association edge between the two target parameter graph nodes is constructed;

[0153] For each node association edge, a node edge weight of the node association edge is calculated according to the second Pearson correlation coefficient and the mutual information value of the two parameter graph nodes corresponding to the node association edge;

[0154] All parameter graph nodes and all node association edges are combined to construct a parameter graph structure of the target drilling fluid controllable parameter.

[0155] In the present embodiment, for any two target parameter graph nodes in the parameter graph, a second Pearson correlation coefficient between them needs to be calculated first. The calculation method of the second Pearson correlation coefficient is the same as that of the first Pearson correlation coefficient. This process can also be based on the parameter data set in Table 1 in the foregoing embodiment, and the calculated second Pearson correlation coefficient is as shown in Table 2. Figure 2

[0156] Next, according to the absolute value of the calculated second Pearson correlation coefficient, it is determined whether to construct a node association edge between the two target parameter graph nodes. A preset edge generation threshold is introduced in this step to filter important parameter associations. If the absolute value of the second Pearson correlation coefficient is less than the threshold, no association edge is constructed between the two nodes. On the contrary, if it is greater than the threshold, an association edge is constructed. For example, assuming that the set edge generation threshold is 0.5, if the absolute value of the second Pearson correlation coefficient of density and API fluid loss is 0.7, an association edge is constructed between the two parameter nodes. If the absolute value of the second Pearson correlation coefficient of density and apparent viscosity is 0.3, no association edge is constructed. This method can effectively filter out parameter pairs with weak correlation, simplify the parameter graph structure, and highlight important parameter relationships. The implementation effect of this step is to obtain a preliminary parameter relationship network, which only contains parameter pairs with strong correlation.

[0157] ​For each reserved node-associated edge, the next step is to calculate its node edge weight. This weight takes into account both the second Pearson correlation coefficient and the mutual information value to more comprehensively describe the strength of the relationship between parameters. This calculation method considers both linear correlation and nonlinear mutual information between parameters, resulting in a more comprehensive and accurate measure of relationship strength. Finally, combining all parameter graph nodes and calculated node-associated edges, the complete parameter graph structure of the target drilling fluid controllable parameters is constructed. This parameter graph is a complex network, where each node represents a drilling fluid parameter, and the edges between nodes represent the relationship between parameters, and the weight of the edge reflects the strength of the relationship. For example, density, API fluid loss, apparent viscosity, and other parameters can be used as nodes, and the associated edges between them are determined according to the results calculated above. This graph structure intuitively shows the mutual relationship and influence degree between various parameters. By analyzing this graph structure, key parameters and important parameter combinations can be identified, providing important reference for drilling fluid formulation optimization. For example, it can be found that there is a strong correlation between some parameters, so that when adjusting a parameter, the possible changes of other parameters can be predicted; it can also identify relatively independent parameter groups, which helps to simplify the control and prediction model of drilling fluid performance. This parameter graph structure construction method integrates various statistical and information theory tools, which can comprehensively capture the complex relationships between parameters, and provide a solid data foundation and analysis framework for subsequent drilling fluid optimization and performance prediction.

[0158] In one embodiment, the calculation of the node edge weight of the node-associated edge according to the second Pearson correlation coefficient and the mutual information value of the two parameter graph nodes corresponding to the node-associated edge includes the following steps:

[0159] The mutual information value of the node-associated edge corresponding to the two parameter graph nodes is normalized;

[0160] The absolute value of the second Pearson correlation coefficient and the normalized mutual information value are weighted and averaged to obtain the node edge weight of the node-associated edge.

[0161] In the embodiment, firstly, the mutual information value of the node associated edge corresponding to two parameter graph nodes is normalized. The purpose of normalization is to adjust the mutual information value to a unified scale, so as to facilitate comparison and combination with the second Pearson correlation coefficient. The normalization method can adopt the minimum-maximum normalization formula: normalized mutual information value=(original mutual information value-minimum mutual information value) / (maximum mutual information value-minimum mutual information value). For example, if the mutual information value of all parameter pairs ranges from 0.2 to 0.8, then for a parameter pair with a mutual information value of 0.5, the normalized value is (0.5-0.2) / (0.8-0.2)=0.5. Secondly, the absolute value of the second Pearson correlation coefficient is weighted and averaged with the normalized mutual information value to obtain the final node edge weight. The calculation formula can be represented as: node edge weight=α*|second Pearson correlation coefficient|+(1-α)*normalized mutual information value, wherein α is a weight coefficient between 0 and 1, used to adjust the proportion of the two indicators in the final weight. For example, if α is set to 0.6, the absolute value of the second Pearson correlation coefficient of a certain parameter pair is 0.7, and the normalized mutual information value is 0.5, then the node edge weight is 0.6*0.7+0.4*0.5=0.62.

[0162] The application further discloses a drilling fluid reservoir protection performance optimization system based on data correlation analysis, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the drilling fluid out-formation plate protection performance optimization method based on data correlation analysis described in any of the embodiments when executing the computer program.

[0163] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, and the application does not limit this.

[0164] The memory can be an internal storage unit of the computer device, such as a hard disk or a memory of the computer device, or an external storage device of the computer device, such as a plug-in hard disk, a smart memory card (SMC), a secure digital card (SD) or a flash memory card (FC) provided on the computer device, etc. The memory can also be a combination of the internal storage unit and the external storage device of the computer device. The memory is used to store the computer program and other programs and data required by the computer device. The memory can also be used to temporarily store data that has been output or will be output, and the application does not limit this.

[0165] Those skilled in the art will understand that the above discussion of any embodiment is merely exemplary in nature and is not intended to imply that the present application is limited to these examples; the above embodiments or technical features among different embodiments can also be combined, steps can be implemented in any order, and there are many other changes to different aspects of one or more embodiments of the present application as described above, which are not provided in detail for the sake of brevity.

[0166] One or more embodiments of the present application are intended to cover all such alternatives, modifications, and variations as fall within the broad scope of the present application. Accordingly, any omission, modification, equivalent replacement, improvement, etc. made within the spirit and principle of one or more embodiments of the present application should be included in the scope of the present application.

Claims

1. A method for optimizing drilling fluid reservoir protection performance based on data correlation analysis, characterized in that, The method comprises the following steps: obtaining an initial drilling fluid formula and initial drilling fluid parameters of a target drilling fluid, and determining a target permeability recovery value corresponding to the target drilling fluid according to the initial drilling fluid formula and the initial drilling fluid parameters; performing attribution analysis on a correlation between the initial drilling fluid parameters and the target permeability recovery value based on a preset parameter data set, and screening target drilling fluid controllable parameters having a correlation with the target permeability recovery value from all the initial drilling fluid parameters according to an attribution analysis result, wherein the parameter data set comprises multiple groups of drilling fluid parameters and multiple groups of permeability recovery values corresponding to the drilling fluid parameters; constructing a multiple regression model between all the target drilling fluid controllable parameters and the target permeability recovery value by taking the target permeability recovery value as a dependent variable and taking all the target drilling fluid controllable parameters as independent variables; calculating an optimal multiple regression equation between all the target drilling fluid controllable parameters and the target permeability recovery value based on the multiple regression model and by using the parameter data set; generating parameter graph nodes respectively according to each of the target drilling fluid controllable parameters, and generating node features of the parameter graph nodes and node edges between the parameter graph nodes in combination with the attribution analysis result and the optimal multiple regression equation, to obtain a parameter graph structure of the target drilling fluid controllable parameters; constructing a parameter optimization graph network in combination with the parameter graph structure and the optimal multiple regression equation, training the parameter optimization graph network by using the parameter data set until a difference between a predicted permeability recovery value and an actual permeability recovery value predicted by the parameter optimization graph network reaches a minimum value; inputting the initial drilling fluid parameters into the trained parameter optimization graph network to obtain optimal drilling fluid parameters output by the parameter optimization graph network, and adjusting the initial drilling fluid formula to an optimal drilling fluid formula according to the optimal drilling fluid parameters.

2. The method for optimizing drilling fluid reservoir protection performance based on data correlation analysis according to claim 1, characterized in that, The attribution analysis on the correlation between the initial drilling fluid parameters and the target permeability recovery value based on the preset parameter data set, and the screening of the target drilling fluid controllable parameters having a correlation with the target permeability recovery value from all the initial drilling fluid parameters according to the attribution analysis result comprise the following steps: preliminarily screening drilling fluid controllable parameters associated with the target permeability recovery value from the initial drilling fluid parameters through a core flow experiment; analyzing data correlation degrees between all the drilling fluid controllable parameters and the target permeability recovery value based on the preset parameter data set and in combination with a Pearson linear correlation analysis method and a Shapley nonlinear correlation analysis method, and taking the drilling fluid controllable parameters having a data correlation degree greater than a preset threshold as target drilling fluid controllable parameters.

3. The method for optimizing drilling fluid reservoir protection performance based on data correlation analysis according to claim 1 or 2, characterized in that, The target controllable drilling fluid parameter includes a drilling fluid basic parameter, a rheological parameter, a solid phase parameter, and a liquid phase parameter, the drilling fluid basic parameter includes a drilling fluid density and a linear expansion height, the rheological parameter includes an apparent viscosity, a plastic viscosity, a dynamic shear force, and a dynamic plasticity ratio, the solid phase parameter includes a solid phase content and a solid phase particle size, and the liquid phase parameter includes an HTHP fluid loss or an API fluid loss, and the liquid phase parameter further includes a pH value, a surface tension, and a cation concentration.

4. The method for optimizing drilling fluid reservoir protection performance based on data correlation analysis according to claim 2, characterized in that, The data correlation degree between each of the controllable drilling fluid parameters and the target permeability recovery value is calculated based on the preset parameter data set and by using the Pearson linear correlation analysis method, if the data correlation degree of a controllable drilling fluid parameter is greater than a preset threshold, the controllable drilling fluid parameter is taken as a target controllable drilling fluid parameter, and the data correlation degree between each of the controllable drilling fluid parameters and the target permeability recovery value is calculated based on the preset parameter data set and by using the Pearson linear correlation analysis method, if the data correlation degree of a controllable drilling fluid parameter is greater than a preset threshold, the controllable drilling fluid parameter is taken as a target controllable drilling fluid parameter. The first Pearson correlation coefficient between each of the controllable drilling fluid parameters and the target permeability recovery value is calculated based on the preset parameter data set and by using the Pearson linear correlation analysis method, if there are multiple linear correlation parameters with the same first Pearson correlation coefficient among the controllable drilling fluid parameters, any one of the linear correlation parameters is retained and all other linear correlation parameters are excluded, the Sharpley value between each of the controllable drilling fluid parameters after the exclusion and the target permeability recovery value is calculated by using the Sharpley nonlinear correlation analysis method, if the Sharpley value is greater than a preset Sharpley threshold, the corresponding controllable drilling fluid parameter is taken as a target controllable drilling fluid parameter, and if the Sharpley value is less than the Sharpley threshold, the corresponding controllable drilling fluid parameter is excluded. The calculation formula of the first Pearson correlation coefficient is as follows: The calculation formula of the Sharpley value is as follows: The parameter graph structure of the target controllable drilling fluid parameter is obtained by generating a parameter graph node for each of the target controllable drilling fluid parameters, generating a node feature of the parameter graph node according to a target correlation factor between the target controllable drilling fluid parameter and the target permeability recovery value, and constructing a node association edge between each of the parameter graph nodes based on the Pearson linear correlation analysis method and according to the parameter data set, and determining a node edge weight of the node association edge in combination with the corresponding mutual information value. The mutual information value between all the target controllable drilling fluid parameters is calculated based on the parameter data set, a node association edge between each of the parameter graph nodes is constructed based on the Pearson linear correlation analysis method and according to the parameter data set, and a node edge weight of the node association edge is determined in combination with the corresponding mutual information value, to obtain the parameter graph structure of the target controllable drilling fluid parameter.

5. The method for optimizing drilling fluid reservoir protection performance based on data correlation analysis according to claim 4, characterized in that, ​ ， wherein: represents the first Pearson correlation coefficient, represents the parameter data volume of the drilling fluid controllable parameter, represents the value of the first drilling fluid controllable parameter, represents the average value of all the drilling fluid controllable parameters, represents the value of the first permeability recovery value, represents the average value of the permeability recovery value.

6. The method for optimizing drilling fluid reservoir protection performance based on data correlation analysis according to claim 4, characterized in that, ​ ， wherein: represents the Shapley value of the controllable parameter of the drilling fluid , , , represents the subset of all controllable parameters of the drilling fluid except the controllable parameter of the drilling fluid , , represents the number of parameters in the subset , , , represents the contribution of the subset .

7. The method for optimizing drilling fluid reservoir protection performance based on data correlation analysis according to claim 2, characterized in that, ​ ​ ​ ​ 8. The method for optimizing drilling fluid reservoir protection performance based on data correlation analysis according to claim 7, characterized in that, The node association edge between each of the parameter graph nodes is constructed based on the Pearson linear correlation analysis method and according to the parameter data set, and the node edge weight of the node association edge is determined in combination with the corresponding mutual information value, so as to obtain the parameter graph structure of the target drilling fluid controllable parameter. For any two target parameter graph nodes in all the parameter graph nodes, a second Pearson correlation coefficient between the two target parameter graph nodes is calculated according to the corresponding drilling fluid parameter data of the two target parameter graph nodes in the parameter data set. If the absolute value of the second Pearson correlation coefficient is less than a preset edge generation threshold, no node association edge between the two target parameter graph nodes is constructed. If the absolute value of the second Pearson correlation coefficient is greater than the edge generation threshold, a node association edge between the two target parameter graph nodes is constructed. For each node association edge, a node edge weight of the node association edge is calculated according to the second Pearson correlation coefficient and the mutual information value of the two parameter graph nodes corresponding to the node association edge. The parameter graph structure of the target drilling fluid controllable parameter is constructed in combination with all the parameter graph nodes and all the node association edges.

9. The method for optimizing drilling fluid reservoir protection performance based on data correlation analysis according to claim 8, characterized in that, The node edge weight of the node association edge is calculated according to the second Pearson correlation coefficient and the mutual information value of the two parameter graph nodes corresponding to the node association edge, including the following steps: The mutual information value of the two parameter graph nodes corresponding to the node association edge is normalized. The absolute value of the second Pearson correlation coefficient and the normalized mutual information value are weighted and averaged to obtain the node edge weight of the node association edge.

10. A data correlation analysis based drilling fluid reservoir protection performance optimization system comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein, The processor executes the computer program to implement the drilling fluid reservoir protection performance optimization method based on data association analysis according to any one of claims 1 to 9.

Images