A fracturing reconstruction effect prediction and decision method and system

Abstract

The embodiment of the present application relates to the technical field of intelligent oil and gas exploitation, and discloses a fracturing reconstruction effect prediction and decision method and system, which comprises the following steps: collecting key parameters about microseismic event points and construction pressure during fracturing operation; constructing a coupling relationship model between the microseismic event points and the construction pressure, and obtaining a prediction model after parameter optimization training of the coupling relationship model based on the constructed minimum loss function; inputting the key parameters into the prediction model obtained after training, and outputting fracturing reconstruction prediction effect by the prediction model; and constructing a decision function based on the fracturing reconstruction prediction effect, and outputting an optimized fracturing reconstruction decision. The fracturing reconstruction effect prediction and decision method and system disclosed in the present application solve the problems of insufficient prediction accuracy of traditional fracturing effect and inability to adapt to geological conditions, thereby realizing accurate prediction of fracturing reconstruction effect.

Description

Technical Field

[0001] The embodiments of this invention relate to the field of intelligent oil and gas extraction technology, and in particular to a method and system for predicting and making decisions on fracturing effects. Background Technology

[0002] Hydraulic fracturing is one of the key technologies for improving the production rate of low-permeability oil and gas reservoirs, playing a crucial role, especially in shale gas development. However, due to the complexity and uncertainty of underground geological structures, the fracturing effect is often difficult to predict accurately, leading to difficulties in optimizing construction parameters and sometimes even causing resource waste or environmental damage.

[0003] Traditional methods for predicting fracturing effectiveness primarily rely on empirical formulas, laboratory tests, and finite element analysis (FEA). While these methods offer some predictive capability, they often lack sufficient accuracy in complex geological environments and cannot adjust construction strategies in real time to adapt to changing geological conditions. Furthermore, these traditional methods typically lack sensitivity and predictive ability for microseismic events, which are crucial indicators for assessing fracture propagation and fracturing effectiveness. Summary of the Invention

[0004] The purpose of this invention is to provide at least one method and system for predicting and deciding on the effect of fracturing stimulation, which can at least solve the problems of insufficient accuracy in predicting the effect of fracturing and inability to adapt to geological conditions in current traditional methods, and can at least accurately predict the effect of fracturing.

[0005] To address the aforementioned technical problems, at least one embodiment of this application provides a method for predicting and deciding on the effect of fracturing stimulation, comprising: collecting key parameters regarding microseismic event points and construction pressure during fracturing operations; constructing a coupling relationship model between the microseismic event points and the construction pressure, and obtaining a prediction model by optimizing the parameters of the coupling relationship model based on a constructed minimum loss function; inputting the key parameters into the prediction model obtained after training, and the prediction model outputting the predicted effect of fracturing stimulation; constructing a decision function based on the predicted effect of fracturing stimulation, and outputting an optimized fracturing stimulation decision.

[0006] At least one embodiment of this application also provides a fracturing stimulation effect prediction and decision-making system, including:

[0007] The fracturing parameter acquisition module is used to collect key parameters related to microseismic event points and construction pressure during fracturing operations;

[0008] The prediction model building module is used to construct a coupling relationship model between the microseismic event points and the construction pressure, and to obtain the prediction model by performing parameter optimization training on the coupling relationship model based on the constructed minimum loss function;

[0009] The prediction effect module includes the prediction model obtained after training, which is used to input the key parameters and output the predicted effect of fracturing transformation.

[0010] The decision optimization module is used to construct a decision function based on the predicted fracturing stimulation effect and output an optimized fracturing stimulation decision.

[0011] The embodiments of this application provide a method and system for predicting and deciding on the fracturing effect. This method constructs a coupling relationship model between microseismic event points and the construction pressure. After training the relationship model based on the constructed minimum loss function, a prediction model is obtained. By inputting key parameters into the prediction model, the predicted fracturing effect is output. This solves the problems of insufficient accuracy and inability to adapt to geological conditions in traditional fracturing effect prediction, thereby achieving accurate prediction of fracturing effect.

[0012] In some optional embodiments, the method further includes: comparing the actual fracturing effect data collected with the predicted effect to obtain the prediction error, adjusting the model parameters of the prediction model based on the principle of minimizing the prediction error, and adjusting the weight coefficients of the decision function based on the principle of the importance of each factor in the decision-making process. This embodiment efficiently explores the high-dimensional space of model parameters and decision function weights to find the optimal configuration, achieving adaptive optimization of fracturing transformation effect prediction and decision-making.

[0013] In some optional embodiments, the model for constructing the coupling relationship between the microseismic event points and the construction pressure is expressed by the following formula:

[0014]

[0015] Where J(θ) is the coupling relationship model function between microseismic event points and the construction pressure, N is the total number of microseismic event points, and h θ (L (i) ) represents the model's position L of the i-th microseismic event point. (i) The predicted construction pressure value, P (i) This represents the actual construction pressure at the i-th microseismic event point. This embodiment constructs a coupling relationship model function between the microseismic event point and the construction pressure, laying the foundation for parameter update training to obtain the prediction model.

[0016] In some optional embodiments, the prediction model obtained by training the coupling relationship model based on the constructed minimum loss function includes:

[0017] Construct a loss function that minimizes the loss function;

[0018] By updating the parameters of the coupling relationship model along the negative direction of the loss function's gradient, the minimum value of the loss function is gradually approximated, resulting in a trained prediction model. This embodiment improves the accuracy of the prediction model by optimizing the training of the coupling relationship model parameters through minimizing the loss function.

[0019] In some optional embodiments, the construction of the minimization loss function is expressed as follows:

[0020]

[0021] Wherein, the loss function L(θ) is a log-likelihood loss, m represents the number of training samples, and y (i) x is the true label of the i-th sample, with a value of 0 or 1. (i) h is the feature vector of the i-th sample. θ (x (i) ) is the model for the i-th sample x. (i) The predicted probability is given by θ, where θ is the set of model parameters. This embodiment constructs a loss function that minimizes the model parameters to achieve model parameter updates.

[0022] In some optional embodiments, updating the parameters of the coupling model by means of the negative direction of the gradient of the loss function includes:

[0023] The gradient descent algorithm is used to calculate the gradient of the loss function with respect to the parameters θ of the coupling model.

[0024] Update the coupling model parameters θ according to the following formula:

[0025]

[0026] Here, α is the learning rate, which controls the magnitude of each parameter update. In this embodiment, the choice of learning rate can reduce prediction error and improve the prediction model's ability to predict future fracturing operations.

[0027] In some optional embodiments, the step of inputting the key parameters into the prediction model obtained after training, and the prediction model outputting the predicted effect of fracturing modification, includes:

[0028] The key parameters are input into the multi-input multi-output prediction function constructed in the prediction model: The prediction results, including the morphological feature matrix and spatial distribution of microseismic event points, are obtained.

[0029] in, θ represents the key parameters of the microseismic event point and construction pressure, respectively.* These are the optimal model parameters obtained through the training process, where M and D represent the morphological feature matrix and spatial distribution of microseismic event points, respectively.

[0030] Based on the prediction results, the mean and standard deviation of the morphological feature matrix M and spatial distribution D of the microseismic event points are calculated respectively.

[0031] The confidence interval of the prediction result is calculated based on the mean and standard deviation of the morphological feature matrix M and spatial distribution D of the microseismic event points.

[0032] The predicted fracturing effect is output based on the confidence interval of the predicted results. This embodiment estimates the confidence interval of the predicted results using Monte Carlo simulation and Bayesian optimization methods, thereby achieving accurate prediction of the fracturing effect.

[0033] In some optional embodiments, the decision function constructed based on the predicted fracturing effect is expressed by the following formula:

[0034]

[0035] Where G() is the decision function, C = [C1, C2, ..., C m [E1, E2, ..., E2] represents the cost vector of a series of possible construction schemes. m [P1, P2, ..., P] represents the expected economic benefit vector of the corresponding construction plan. m ] is the predicted effect vector of fracturing stimulation, λ c , λ e , λ p The weighting coefficients are, in order, cost, expected economic benefits, and predicted fracturing effect. The objective of the decision function G in this embodiment is to find a construction plan i, assessing the actual importance of factors such as cost, economic benefits, fracturing effect, environmental risks, and social impacts in the decision-making process, maximizing the weighted sum of cost-benefit ratio, expected economic benefits, and predicted fracturing effect, and using analysis based on actual results to more accurately reflect the optimal decision under actual conditions.

[0036] In some optional implementations, the system further includes an adaptive optimization module, used to compare the actual fracturing effect data collected with the predicted effect to obtain the prediction error, and adjust the model parameters of the prediction model based on the principle of minimizing the prediction error, while adjusting the weight coefficients of the decision function based on the principle of the importance of each factor in the decision-making process. This embodiment can achieve adaptive optimization for predicting and deciding on the fracturing transformation effect. Attached Figure Description

[0037] One or more embodiments are illustrated by way of example with reference to the accompanying drawings, and these illustrative descriptions do not constitute a limitation on the embodiments.

[0038] Figure 1 This is a flowchart of a fracturing stimulation effect prediction and decision-making method provided in one embodiment of this application;

[0039] Figure 2 This is a schematic diagram of the structure of a fracturing effect prediction and decision-making system provided in another embodiment of this application. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the various embodiments of this application will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details have been provided in the various embodiments of this application to help readers better understand this application. However, the technical solutions claimed in this application can be implemented even without these technical details and various changes and modifications based on the following embodiments. The division of the various embodiments below is for the convenience of description and should not constitute any limitation on the specific implementation of this application. The various embodiments can be combined with and referenced by each other without contradiction.

[0041] To facilitate understanding of the embodiments of this application, relevant content about machine learning models will be introduced first.

[0042] In recent years, the development of machine learning technology, especially the advancements in deep learning and big data analytics, has brought new possibilities for predicting the effects of fracturing stimulation. Machine learning models can handle large amounts of nonlinear, high-dimensional data, and by learning from historical fracturing operation data, they can predict the effects of future operations, providing a scientific basis for optimizing construction parameters. However, applying machine learning technology to predict fracturing effects still presents challenges, including how to effectively integrate physical models with data-driven models, and how to achieve model generalization and robustness under multivariate conditions.

[0043] Therefore, there is an urgent need for a new prediction method that combines physical model constraints with machine learning technology to improve the accuracy and reliability of fracturing effect prediction, while also being able to intelligently recommend the optimal construction plan and have adaptive adjustment capabilities to cope with dynamic changes in the geological environment, thereby realizing intelligent and efficient fracturing operations.

[0044] To address the technical problems of insufficient accuracy in predicting fracturing effects and inability to adapt to geological conditions in traditional methods, this invention proposes a machine learning-based method and system for predicting fracturing effects. The implementation details of the machine learning-based fracturing effect prediction method in this embodiment are described below. The following content is only for ease of understanding and is not necessary for implementing this solution.

[0045] This invention discloses a machine learning-based method for predicting the effects of fracturing operations. It aims to achieve accurate prediction of fracturing operation effects and intelligent optimization of construction parameters by integrating multiple technologies and algorithms. The method includes the following steps: data acquisition and preprocessing, comprehensively collecting key parameters of the fracturing operation, including detailed information on microseismic event points and dynamic changes in construction pressure; constructing a physically constrained microseismic-construction pressure correlation model, using finite element analysis technology based on the principle of fracture propagation dynamics to establish a coupling relationship model between construction pressure and microseismic event points, accurately describing the stress distribution of rock strata, fracture propagation paths, and microseismic activity, minimizing prediction errors through parameter optimization; designing and training a machine learning model, employing supervised learning methods and using the gradient descent algorithm to minimize the loss function, predicting the morphology and distribution of microseismic events; and optimizing construction parameters. This invention, along with fracturing effect prediction, uses machine learning models to predict the morphological characteristics and spatial distribution of microseismic event points. Monte Carlo simulation and Bayesian optimization methods are employed to evaluate the confidence intervals of the prediction results, providing decision support for construction. The invention also features intelligent recommendation and adaptive adjustment of fracturing stimulation schemes. A decision function comprehensively considers cost, economic benefits, fracturing prediction effects, environmental risks, and social impacts to output the optimal construction scheme and dynamically adjusts model parameters to cope with changes in the geological environment. The adaptive optimization module utilizes Bayesian optimization technology, based on actual fracturing effect feedback, to continuously optimize model parameters and decision function weights, ensuring the accuracy and practicality of the prediction model and decision-making mechanism. This invention combines data-driven approaches with physical modeling to achieve accurate prediction of fracturing stimulation effects and intelligent optimization of construction schemes, providing a scientific basis and technical support for efficient oil and gas extraction.

[0046] Compared with the prior art, the beneficial effects of the present invention are:

[0047] 1. This invention improves prediction accuracy: By utilizing a machine learning model and combining it with a microseismic-construction pressure correlation model under physical constraints, this invention can more accurately predict the morphology and distribution of microseismic event points. Compared with traditional methods that rely solely on empirical formulas or finite element analysis, the prediction accuracy is significantly improved.

[0048] 2. This invention can optimize construction parameters: Through machine learning prediction models, this invention can optimize fracturing construction schemes based on real-time construction pressure and other parameters, guiding construction personnel to perform more precise operations and reducing unnecessary resource consumption and environmental impact;

[0049] 3. This invention achieves adaptive optimization of the model and decision: This invention utilizes the Bayesian optimization method to achieve efficient exploration of the high-dimensional space of model parameters and decision function weights, and adjusts them based on feedback from actual results to continuously optimize prediction ability and decision quality;

[0050] 4. This invention improves decision support: the decision function comprehensively considers cost, economic benefits, environmental risks and social impacts, ensuring the comprehensive optimization of fracturing modification schemes and maximizing cost-effectiveness.

[0051] Example 1:

[0052] The fracturing stimulation effect prediction and decision-making method of this embodiment can be applied to electronic devices with communication, computing and data storage capabilities. Its specific process can be as follows: Figure 1 As shown, it includes:

[0053] Step 101: Collect key parameters regarding microseismic event points and construction pressure during the fracturing operation.

[0054] Specifically, for the parameter data acquisition and preprocessing of fracturing stimulation effect prediction, advanced sensors are used to comprehensively collect key parameters during fracturing operations, including but not limited to the number N, location L, and influence range R of microseismic events, as well as the dynamic changes P of the construction pressure. The data quality index is calculated to ensure optimal input data quality; the calculation formula is:

[0055]

[0056] Among them, ω1 is completeness, ω2 is accuracy, ω3 is consistency, ω4 is timeliness, and ω5 is richness; D is the data quality index.

[0057] Step 102: Construct a coupling relationship model between the microseismic event points and the construction pressure, and obtain a prediction model by optimizing the parameters of the coupling relationship model based on the constructed minimum loss function.

[0058] Specifically, a microseismic-construction pressure correlation model under physical constraints is constructed. Using finite element analysis (FEM) technology and based on the principles of crack propagation dynamics, a coupling relationship model between microseismic event points and construction pressure is established. A machine learning prediction model is designed and trained: a supervised learning method is used to develop and train a model predicting the morphology and distribution of microseismic events. The gradient descent algorithm is used to minimize the loss function, and construction parameters are optimized based on minimizing the loss function.

[0059] Step 103: Input the key parameters into the prediction model obtained after training, and the prediction model outputs the prediction effect of fracturing transformation.

[0060] Specifically, the prediction of fracturing stimulation effects utilizes machine learning prediction models combined with construction pressure data. Other construction parameter sets The morphology and distribution of microseismic event points are predicted, and confidence intervals are estimated for the predicted morphology and distribution of microseismic event points through Monte Carlo simulation and Bayesian optimization evaluation.

[0061] Step 104: Construct a decision function based on the predicted fracturing stimulation effect and output an optimized fracturing stimulation decision.

[0062] In practical implementation, the optimal construction plan is automatically output based on feedback from the prediction model, while the model parameters are adaptively updated to cope with dynamic changes in the geological environment. The prediction model selects the outcome with the highest cost-effectiveness ratio by constructing a decision function G.

[0063] This embodiment constructs a coupling relationship model between microseismic event points and construction pressure. After training the relationship model based on the constructed minimum loss function, a prediction model is obtained. By inputting key parameters into the prediction model, the predicted effect of fracturing transformation is output, and optimized construction decisions are output based on the predicted effect. This solves the problems of insufficient accuracy of traditional fracturing effect prediction and inability to adapt to geological conditions, thereby achieving accurate prediction of fracturing effect.

[0064] In some embodiments, the fracturing effect prediction and decision-making method further includes: comparing the actual fracturing effect data collected with the predicted effect to obtain the prediction error, adjusting the model parameters of the prediction model based on the principle of minimizing the prediction error, and adjusting the weight coefficients of the decision function based on the principle of the importance of each factor in the decision-making process.

[0065] Specifically, Bayesian optimization methods can be used to efficiently explore the high-dimensional space of model parameters and decision function weights to find the optimal configuration. After the fracturing operation is completed, the adaptive optimization module evaluates the accuracy of the prediction model and the applicability of the decision function by comparing the actual fracturing effect with the predicted results. Actual data from the fracturing operation is collected, including the final shape of the fracture network, changes in production well output, and environmental monitoring data, and compared with the prediction results in step S103 to evaluate the accuracy of the prediction model. Using feedback from the actual results, the parameters θ of the machine learning prediction model in step S102 are adjusted through model inversion technology to reduce prediction errors and improve the model's ability to predict future fracturing operations. Based on the analysis of the actual results, the actual importance of factors such as cost, economic benefits, fracturing effect, environmental risks, and social impacts in the decision-making process is evaluated, and the weight coefficients λ of the decision function in step S104 are adjusted accordingly. c , λ e , λ p To more accurately reflect the optimal decision under actual conditions.

[0066] In some embodiments, the model for constructing the coupling relationship between the microseismic event points and the construction pressure is expressed by the following formula:

[0067]

[0068] Where J(θ) is the coupling relationship model function between microseismic event points and the construction pressure, N is the total number of microseismic event points, and h θ (L (i) ) represents the model's position L of the i-th microseismic event point. (i) The predicted construction pressure value, P (i) It represents the actual construction pressure at the i-th microseismic event point.

[0069] Specifically, by integrating geomechanical properties and fluid dynamics principles, and applying finite element analysis (FEA) technology, a coupling relationship model between microseismic event points and construction pressure is constructed based on the principle of fracture propagation dynamics. The specific steps and formulas are as follows:

[0070] Model Establishment: First, a three-dimensional rock strata model is established, incorporating the physical properties of the rock strata, such as elastic modulus E, Poisson's ratio v, and density ρ. The model considers the fracture network geometry and the stress intensity factor K at the fracture tips. I And the distribution of construction pressure. The equilibrium equation of the model is:

[0071]

[0072] Where σ is the stress tensor and b is the body force density vector. This represents the divergence operator.

[0073] Stress-strain relationship: The stress σ and strain ε in rock strata follow Hooke's law:

[0074] σ ij =C ijkl ε kl

[0075] For isotropic linear elastic materials, where C ijkl It is a fourth-order tensor of the elastic modulus.

[0076] Crack propagation: Introducing the stress intensity factor K at the crack tip I To determine whether a material will experience crack propagation, the calculation formula is:

[0077]

[0078] Where σ is the far-field stress, a is the crack half-length, W is the specimen width, and f(a / W) is a function related to the crack size and specimen geometry.

[0079] Model solution: The above equations are solved by the finite element method to obtain the stress distribution inside the rock strata, the crack propagation path, and the prediction of microseismic activity.

[0080] Objective function: The goal of building the model is to minimize the prediction error by optimizing the parameters θ. The objective function is:

[0081]

[0082] Where N is the total number of microseismic event points, h θ (L (i) ) represents the model's position L of the i-th microseismic event point. (i) The predicted construction pressure value, P (i) It represents the actual construction pressure at the i-th microseismic event point.

[0083] In some embodiments, the step of obtaining a prediction model by training the coupling relationship model with parameter optimization based on the constructed minimum loss function includes:

[0084] Construct a loss function that minimizes the loss function;

[0085] Specifically, a supervised learning method is used to train a machine learning prediction model, and a gradient descent algorithm is used to construct a loss function that minimizes the loss function.

[0086] By updating the parameters of the coupling relationship model along the negative direction of the gradient of the loss function, the minimum value of the loss function is gradually approximated, and the trained prediction model is obtained.

[0087] Specifically, gradient descent is an iterative optimization algorithm that gradually approaches the minimum value of the loss function by updating the model parameters along the negative direction of the loss function's gradient. In each iteration, gradient descent calculates the gradient of the loss function with respect to the model parameters. In some embodiments, the construction of the minimized loss function is expressed as follows:

[0088]

[0089] Wherein, the loss function L(θ) is a log-likelihood loss, m represents the number of training samples, and y (i) x is the true label of the i-th sample, with a value of 0 or 1. (i) h is the feature vector of the i-th sample. θ (x (i) ) is the model for the i-th sample x. (i) The predicted probability is θ, where θ is the set of parameters of the model.

[0090] In some embodiments, updating the parameters of the coupling model along the negative direction of the gradient of the loss function includes:

[0091] The gradient descent algorithm is used to calculate the gradient of the loss function with respect to the parameters θ of the coupling model.

[0092] Update the coupling model parameters θ according to the following formula:

[0093]

[0094] Here, α is the learning rate, which controls the magnitude of each parameter update.

[0095] It should be noted that the choice of learning rate is crucial. An excessively large learning rate may lead to excessively large parameter updates and failure to converge; while an excessively small learning rate will result in a slow convergence speed and increase training time.

[0096] In some embodiments, the step of inputting the key parameters into the prediction model obtained after training, and the prediction model outputting the fracturing modification prediction effect, includes:

[0097] The key parameters are input into the multi-input multi-output prediction function constructed in the prediction model: The prediction results, including the morphological feature matrix and spatial distribution of microseismic event points, are obtained.

[0098] in, θ represents the key parameters of the microseismic event point and construction pressure, respectively. * These are the optimal model parameters obtained through the training process, where M and D represent the morphological feature matrix and spatial distribution of microseismic event points, respectively.

[0099] Based on the prediction results, the mean and standard deviation of the morphological feature matrix M and spatial distribution D of the microseismic event points are calculated respectively.

[0100] The confidence interval of the prediction result is calculated based on the mean and standard deviation of the morphological feature matrix M and spatial distribution D of the microseismic event points.

[0101] The predicted fracturing effect is output based on the confidence interval of the predicted results.

[0102] Specifically, a prediction function F is constructed, which outputs the morphological feature matrix M and spatial distribution D of microseismic event points.

[0103] The morphological feature matrix M contains information such as crack length, width, opening size, and direction for each microseismic event point, and can be represented as a set of feature vectors for all event points:

[0104] M = [m1, m2, ..., m n ] T

[0105] Where, m i =[L i Wi O i D i The crack length, width, opening size, and direction correspond to the i-th microseismic event point.

[0106] The spatial distribution D describes the location of each microseismic event point in three-dimensional space, and is represented as the set of all event point locations:

[0107] D=[(x1,y1,z1),(x2,y2,z2),…,(x n ,y n ,z n )]

[0108] A method based on Monte Carlo simulation and Bayesian optimization is constructed to estimate the confidence interval of the prediction results.

[0109] Based on all predictions, calculate the mean and standard deviation of the predictions:

[0110]

[0111] The confidence interval is:

[0112]

[0113] Among them, M (j) D (j) Calculate the corresponding prediction result for the prediction function F. σ is the mean of the predicted results. M σ D This represents the standard deviation of the predicted results.

[0114] In some embodiments, the decision function constructed based on the predicted fracturing effect is expressed by the following formula:

[0115]

[0116] Where G() is the decision function, C = [C1, C2, ..., C m [E1, E2, ..., E2] represents the cost vector of a series of possible construction schemes. m [P1, P2, ..., P] represents the expected economic benefit vector of the corresponding construction plan. m ] is the predicted effect vector of fracturing stimulation, λ c , λ e , λ p The weighting coefficients are, in order, cost, expected economic benefits, and predicted effects of fracturing modification.

[0117] Specifically, based on an assessment of cost and economic benefits, environmental risks, and social impacts, a decision function G() is constructed to achieve comprehensive optimization of fracturing stimulation schemes. The goal of decision function G() is to find a construction scheme i that maximizes the weighted sum of cost-benefit ratio, expected economic benefits, and predicted fracturing effects.

[0118] Example 2:

[0119] Another embodiment of this application relates to a fracturing stimulation effect prediction and decision-making system. The implementation details of this fracturing stimulation effect prediction and decision-making system are described below. The following details are provided for ease of understanding and are not essential for implementing this solution. A schematic diagram of the fracturing stimulation effect prediction and decision-making system in this embodiment can be seen as follows: Figure 2 As shown, it includes a fracturing parameter acquisition module 201, a prediction model establishment module 202, a prediction effect module 203, and a decision optimization module 204.

[0120] The fracturing parameter acquisition module 201 is used to acquire key parameters about microseismic event points and construction pressure during fracturing operations.

[0121] The prediction model building module 202 is used to construct a coupling relationship model between the microseismic event points and the construction pressure, and to obtain a prediction model by performing parameter optimization training on the coupling relationship model based on the constructed minimum loss function.

[0122] The prediction effect module 203 is used to include the prediction model obtained after training, to input the key parameters, and to output the prediction effect of fracturing transformation.

[0123] The decision optimization module 204 is used to construct a decision function based on the predicted effect of the fracturing modification and output an optimized fracturing modification decision.

[0124] This embodiment constructs a coupling relationship model between microseismic event points and construction pressure through a prediction model building module. After training the relationship model based on the constructed minimum loss function, a prediction model is obtained. Key parameters are input into the prediction model through a fracturing parameter acquisition module, and the predicted fracturing transformation effect is output. The optimized construction decision is output based on the prediction effect through a decision optimization module. This solves the problems of insufficient accuracy and inability to adapt to geological conditions in traditional fracturing effect prediction, thereby achieving accurate prediction of fracturing effect.

[0125] In some embodiments, combined with Figure 2 The fracturing transformation effect prediction and decision-making system also includes an adaptive optimization module;

[0126] The adaptive optimization module 205 is used to compare the actual fracturing effect data collected with the predicted effect to obtain the prediction error, and adjust the model parameters of the prediction model based on the principle of minimizing the prediction error, while adjusting the weight coefficients of the decision function based on the principle of the importance of each factor in the decision-making process.

[0127] Specifically, the adaptive optimization module utilizes Bayesian optimization methods to efficiently explore the high-dimensional space of model parameters and decision function weights to find the optimal configuration. After the fracturing operation is completed, the adaptive optimization module evaluates the accuracy of the prediction model and the applicability of the decision function by comparing the actual fracturing effect with the predicted results.

[0128] It is worth mentioning that all modules involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed in this application; however, this does not mean that other units are absent in this embodiment.

[0129] Those skilled in the art will understand that the above embodiments are specific embodiments for implementing this application, and in practical applications, various changes can be made to them in form and detail without departing from the spirit and scope of this application.

Claims

1. A method for predicting and decision making of fracturing reconstruction effect, characterized in that, include: Collect key parameters regarding microseismic event points and construction pressure during fracturing operations; A coupling relationship model between the microseismic event points and the construction pressure is constructed, and a prediction model is obtained by optimizing the parameters of the coupling relationship model based on the constructed minimum loss function. The key parameters are input into the prediction model obtained after training, and the prediction model outputs the prediction effect of fracturing transformation. A decision function is constructed based on the predicted fracturing stimulation effect, and an optimized fracturing stimulation decision is output.

2. The method for predicting and deciding on the fracturing effect according to claim 1, characterized in that, The method further includes: comparing the actual fracturing effect data collected with the predicted effect to obtain the prediction error, adjusting the model parameters of the prediction model based on the principle of minimizing the prediction error, and adjusting the weight coefficients of the decision function based on the principle of the importance of each factor in the decision-making process.

3. The method for predicting and deciding on the fracturing effect according to claim 1, characterized in that, The coupling relationship model between the microseismic event points and the construction pressure is constructed, and its formula is expressed as follows: Wherein, J(θ) is the coupling relationship model function between the microseismic event point and the construction pressure, N is the total number of microseismic event points, h θ (L (i) ) represents the construction pressure predicted value of the model to the i th microseismic event point position L (i) , P (i) is the actual construction pressure of the i th microseismic event point.

4. The fracture reformation effect prediction and decision method according to claim 3, characterized in that, The prediction model is obtained by training the coupling relationship model based on the constructed minimum loss function and optimizing the parameters, including: Construct a loss function that minimizes the loss function; By updating the parameters of the coupling relationship model along the negative direction of the gradient of the loss function, the minimum value of the loss function is gradually approximated, thus obtaining the trained prediction model.

5. The fracture reformation effect prediction and decision method according to claim 4, characterized in that, The loss function is constructed to minimize the loss function, which is expressed by the following formula: where the loss function L(θ) is a log loss, m represents the number of training samples, y (i) is the true label of the i-th sample, taking values 0 or 1, x (i) is the feature vector of the i-th sample, h θ (x (i) ) is the predicted probability of the i-th sample x (i) by the model, and θ is the parameter set of the model.

6. The fracture reformation effect prediction and decision method according to claim 5, characterized in that, The step of updating the parameters of the coupling relationship model along the negative direction of the gradient of the loss function includes: computing a gradient of the loss function with respect to the coupling relationship model parameters θ using a gradient descent algorithm Update the coupling model parameters θ according to the following formula: Here, α is the learning rate, which controls the magnitude of each parameter update.

7. The fracture reformation effect prediction and decision method according to claim 1 or 4, characterized by, The prediction model obtained after training by inputting the key parameters, and the prediction model outputting the fracturing modification prediction effect, includes: inputting the key parameters into a multi-input multi-output prediction function constructed by the prediction model: obtaining a prediction result including a shape feature matrix and a spatial distribution of microseismic event points; in, θ represents the key parameters of the microseismic event point and construction pressure, respectively. * These are the optimal model parameters obtained through the training process, where M and D represent the morphological feature matrix and spatial distribution of microseismic event points, respectively. Based on the prediction results, the mean and standard deviation of the morphological feature matrix M and spatial distribution D of the microseismic event points are calculated respectively. The confidence interval of the prediction result is calculated based on the mean and standard deviation of the morphological feature matrix M and spatial distribution D of the microseismic event points. The predicted fracturing effect is output based on the confidence interval of the predicted results.

8. The fracture reformation effect prediction and decision method according to any one of claims 1-7, characterized in that, The decision function constructed based on the predicted fracturing effect is expressed by the following formula: Where G() is the decision function, C = [C1, C2, ..., C m [E1, E2, ..., E2] represents the cost vector of a series of possible construction schemes. m [P1, P2, ..., P] represents the expected economic benefit vector of the corresponding construction plan. m ] is the predicted effect vector of fracturing stimulation, λ c , λ e , λ p The weighting coefficients are, in order, cost, expected economic benefits, and predicted effects of fracturing modification.

9. A fracturing reformation effect prediction and decision system, characterized in that, include: The fracturing parameter acquisition module is used to collect key parameters related to microseismic event points and construction pressure during fracturing operations; The prediction model building module is used to construct a coupling relationship model between the microseismic event points and the construction pressure, and to obtain the prediction model by performing parameter optimization training on the coupling relationship model based on the constructed minimum loss function; The prediction effect module includes the prediction model obtained after training, which is used to input the key parameters and output the predicted effect of fracturing transformation. The decision optimization module is used to construct a decision function based on the predicted fracturing stimulation effect and output an optimized fracturing stimulation decision.

10. The fracture reformation effect prediction and decision system of claim 9, wherein, It also includes an adaptive optimization module, which is used to compare the actual fracturing effect data collected with the predicted effect to obtain the prediction error, and adjust the model parameters of the prediction model based on the principle of minimizing the prediction error, while adjusting the weight coefficients of the decision function based on the principle of the importance of each factor in the decision-making process.

Images