A shale gas fractured horizontal well production prediction method based on a double medium model

Abstract

The present application provides a shale gas fractured horizontal well production prediction method based on a dual medium model, relates to the technical field of oil and gas engineering, and comprises the following steps: S1, constructing a dynamic coupling model of a dual medium considering adsorption and desorption, stress sensitivity and CO2 displacement; the present application provides a shale gas fractured horizontal well production prediction method based on a dual medium model, improves the production prediction accuracy: by constructing a dynamic coupling model, using a hybrid deep learning model to extract features, comprehensively considering various influencing factors, the production prediction is more accurate, providing a reliable basis for development decision, optimizing the development plan: by optimizing the fracturing parameters and CO2 injection scheme through an intelligent algorithm, the production is improved, and better carbon sequestration effect is realized, taking into account economic benefits and environmental benefits, improving the calculation efficiency: adopting parallel computing and model reduction technology, accelerating the model solving speed, quickly responding to actual production demand, facilitating real-time adjustment of production strategy.

Description

Technical Field

[0001] This invention relates to the field of oil and gas engineering technology, and in particular to a method for predicting the production of shale gas fractured horizontal wells based on a dual-medium model. Background Technology

[0002] Shale gas, as an important unconventional natural gas resource, has become one of the core areas of global energy development due to its abundant reserves and clean, efficient characteristics. Shale gas reservoirs are characterized by "low porosity, low permeability, and heterogeneity," with uneven development of natural fracture systems and matrix pores mostly at the nanoscale. Gas exists in both adsorbed and free states, resulting in complex flow mechanisms. To achieve economical and efficient shale gas extraction, horizontal well staged fracturing technology is typically employed. This involves artificially creating fractures to connect natural fractures with matrix pores, forming a complex fracture network system that provides channels for gas seepage. In this process, accurately predicting the production dynamics of fracturing horizontal wells is not only the core basis for reservoir evaluation and development scheme optimization but also a key prerequisite for reducing development risks and improving resource recovery. Currently, production prediction methods for shale gas fractured horizontal wells mainly include empirical formula methods, physical simulation methods, numerical simulation methods, and data-driven methods. Among them, the numerical simulation method based on the dual-medium model has become the mainstream technique because it can better characterize the coupled flow characteristics of the matrix and fracture systems. The dual-medium model abstracts the shale reservoir into a coupled system of a matrix system (providing the gas source) and a fracture system (the main seepage channel). By establishing a mass conservation equation to describe the migration law of gas between the two systems, production dynamics can be predicted. Existing production prediction methods based on dual-media models have several drawbacks. Firstly, the coupling mechanism is simplified, and the dynamic response is not adequately characterized. For example, they do not fully consider the matrix shrinkage effect caused by gas desorption, the attenuation of fracture conductivity due to changes in effective stress (stress sensitivity), and the dynamic impact of fracturing fluid flowback on fracture permeability. This leads to errors in the calculation of the crossflow coefficient between the matrix and fracture, resulting in decreased accuracy in medium- and long-term production predictions (often exceeding 20%). Secondly, the multi-scale flow mechanism is not fully characterized. For instance, the description of gas slippage effects in nanopores often uses empirical correction coefficients without establishing quantitative correlations with pore structure parameters (such as pore size distribution and specific surface area), leading to errors in matrix permeability calculations. Summary of the Invention

[0003] Based on the technical problems existing in the background art, the present invention proposes a method for predicting the production of shale gas fracturing horizontal wells based on a dual-medium model.

[0004] This invention proposes a method for predicting the production of shale gas fractured horizontal wells based on a dual-medium model, comprising the following steps: S1. Construct a dual-medium dynamic coupling model that considers adsorption and desorption, stress sensitivity and CO2 displacement; S2. A 3D-CNN+LSTM hybrid model is used for spatiotemporal feature extraction and yield prediction. S3. Combine strategy gradient-genetic-particle swarm optimization algorithm to optimize fracturing parameters and CO2 injection scheme.

[0005] Preferably, the dual-medium dynamic coupling model includes: Matrix system equations: ; Crack system equations: ; in, C represents matrix porosity. tm P is the total compressibility coefficient of the matrix. m t is the matrix pressure; k is the time; t is the matrix pressure; t is the time; k is the matrix pressure. m For matrix permeability; This refers to the viscosity of the gas. p is the crosstalk coefficient; f For crack pressure; q des q represents the desorption volume; CO2 The amount of desorbed gas is replaced by CO2; For crack porosity; c tf k is the total compressibility coefficient of the crack. f q represents the crack permeability. co2 The amount of desorbed gas replaced by CO2; q inj This refers to the CO2 injection rate.

[0006] Preferably, the 3D-CNN+LSTM hybrid model includes: 3D convolutional layers: extracting spatial features of reservoir heterogeneity; LSTM layer: Captures the temporal characteristics of production dynamics; Fully connected layer: fuses spatiotemporal features and outputs yield prediction results.

[0007] Preferably, the parameter optimization for CO2 displacement includes: The replacement efficiency of CO2 on methane was determined by isothermal adsorption experiments. Optimization of CO2 injection rate and slug combination based on true triaxial fracturing experiments.

[0008] Preferably, the optimization of fracturing parameters includes: Using net present value (NPV) as the objective function, the crack complexity index is constrained to be ≥1.5, and the carbon sequestration efficiency is constrained to be ≥60%. Multi-objective optimization is achieved using a policy gradient-genetic-particle swarm optimization algorithm.

[0009] Preferably, dynamic penetration tracking includes: A stress-sensitive model was constructed based on digital core technology; The stress-percolation coupling effect was simulated by combining the lattice Boltzmann method.

[0010] Preferably, the computational efficiency improvement of the dual-medium dynamic coupling model includes: The discrete crack network model is solved using a hybrid parallel mode of MPI+OpenMP. Model order reduction is achieved by combining orthogonal decomposition and Kriging interpolation.

[0011] Preferably, the digital twin system for shale gas fracturing horizontal wells includes: Real-time data acquisition module: acquires wellhead pressure and injection volume parameters; Kalman filter module: updates model parameters and predicts production dynamics; Closed-loop optimization module: Adjusts fracturing parameters and injection strategy.

[0012] Preferably, it also includes a data closed-loop update step: by collecting production data such as wellhead pressure, CO2 injection volume, and gas production in real time, the key parameters of matrix permeability, fracture conductivity, and crossflow coefficient in the dual-medium model are dynamically corrected using the Kalman filter algorithm, so as to achieve continuous matching between the prediction model and the actual production status and improve the stability of medium- and long-term production prediction.

[0013] Preferably, it also includes a multi-scenario prediction and scheme optimization step: based on the optimized dual-medium model, different combinations of fracturing parameters and CO2 injection strategies are set to generate at least 3 sets of production dynamic prediction curves, and the optimal development scheme is output through a comprehensive evaluation of net present value, carbon sequestration efficiency and recovery rate indicators.

[0014] The shale gas fracturing horizontal well production prediction method proposed in this invention has the following beneficial effects: improved production prediction accuracy: by constructing a dynamic coupling model and extracting features using a hybrid deep learning model, and by comprehensively considering multiple influencing factors, the production prediction is more accurate, providing a reliable basis for development decisions.

[0015] Optimized development plan: By optimizing fracturing parameters and CO2 injection scheme through intelligent algorithms, the production capacity can be increased while achieving better carbon sequestration, thus balancing economic and environmental benefits.

[0016] Improve computational efficiency: By adopting parallel computing and model reduction techniques, the model solving speed is accelerated, enabling rapid response to actual production needs and facilitating real-time adjustment of production strategies.

[0017] Enhance forecast stability: By updating data in a closed loop, model parameters are continuously adapted to changes in actual production, improving the stability of medium- and long-term output forecasts and reducing development risks.

[0018] Scientific planning: Through multi-scenario prediction and scheme optimization, the advantages and disadvantages of different schemes are comprehensively evaluated, the optimal development scheme is selected, and the resource recovery rate and the scientific nature of development are improved. Attached Figure Description

[0019] Figure 1 This is a flowchart of a method for predicting the production of shale gas fracturing horizontal wells based on a dual-medium model, as proposed in this invention. Detailed Implementation

[0020] Reference Figure 1 This invention proposes a method for predicting the production of shale gas fractured horizontal wells based on a dual-medium model, comprising the following steps: S1. A dual-medium dynamic coupling model considering adsorption and desorption, stress sensitivity and CO2 displacement is constructed, which can simultaneously reflect the flow and interaction of gases in the matrix and fracture systems, and incorporates a variety of dynamic influencing factors. S2. A hybrid 3D-CNN+LSTM model is adopted. 3D-CNN extracts the spatial heterogeneity features of the reservoir, while LSTM captures the dynamic features of the production process over time. The combination of the two enables accurate prediction of production. S3. Combine strategy gradient-genetic-particle swarm optimization algorithm to optimize fracturing parameters and CO2 injection scheme in order to achieve the goals of maximizing production and carbon sequestration.

[0021] The specific working principle is to construct a dynamically coupled dual-medium model, combine it with a deep learning model to extract spatiotemporal features, and then use intelligent optimization algorithms to optimize relevant parameters and schemes, forming a complete process from model construction to prediction to scheme optimization. This enables accurate prediction of shale gas fracturing horizontal well production and optimization of development schemes, forming a complete system from model construction to prediction to scheme optimization. This avoids the problem of disconnect between the model and prediction / optimization links in traditional methods, and can more systematically realize production prediction and development scheme optimization, significantly reducing the blindness of development decisions and improving the scientific and economic efficiency of shale gas development.

[0022] The matrix system equations and fracture system equations in the dual-medium dynamic coupling model describe the gas flow patterns in the matrix and fractures, respectively. The matrix system equations consider gas seepage in the matrix, cross-flow between the matrix and fractures, gas desorption, and the amount of gas generated by CO2 displacement desorption; the fracture system equations consider gas seepage in the fractures, cross-flow with the matrix, and the amount of CO2 injected.

[0023] The dual-medium dynamic coupling model includes: Matrix system equations: ; Crack system equations: ; in: C represents matrix porosity. tm P is the total compressibility coefficient of the matrix. m t is the matrix pressure; k is the time; t is the matrix pressure; t is the time; k is the matrix pressure. m For matrix permeability; This refers to the viscosity of the gas. p is the crosstalk coefficient; f For crack pressure; q des q represents the desorption volume; CO2 The amount of desorbed gas is replaced by CO2; For crack porosity; c tf k is the total compressibility coefficient of the crack. f q represents the crack permeability. co2 The amount of desorbed gas replaced by CO2; q inj This refers to the CO2 injection rate.

[0024] By employing two interrelated equations, the gas flow and mass exchange in the matrix and fracture systems are characterized, respectively. A CO2-related source term is introduced to accurately describe the gas flow in the dual-medium system under CO2 displacement, providing a foundation for production forecasting. The matrix and fracture system equations are solved numerically using the finite element method (FEM) or finite difference method. By discretizing the spatial and temporal dimensions, the pressure distribution in the matrix and fracture at different times is obtained, accurately characterizing the gas flow in the dual-medium system under CO2 displacement. Compared to traditional models, this approach better reflects actual production conditions, providing a more reliable basis for production forecasting. Using the FEM or finite difference method to solve the equations efficiently obtains key data such as pressure distribution at different times, improving the model's practicality and computational efficiency.

[0025] The 3D-CNN+LSTM hybrid model includes (the training process of the 3D-CNN+LSTM hybrid model can be supplemented by collecting a large amount of historical production data, reservoir data, etc. as training samples to train the model, and adjusting the network parameters through the backpropagation algorithm to minimize the prediction error of the model): 3D convolutional layer: Extracting spatial features of reservoir heterogeneity. The 3D convolutional layer extracts spatial features of reservoir heterogeneity from the input three-dimensional data (such as spatial distribution data of reservoirs) through convolution operations, such as differences in porosity and permeability in different regions. LSTM layer: Capturing the temporal characteristics of production dynamics. The LSTM layer uses its unique gating mechanism to process time series data (such as the changes in output and pressure over time in the production process) and capture the temporal characteristics of production dynamics. Fully connected layer: fuses spatiotemporal features and outputs yield prediction results. The fully connected layer fuses the spatial features extracted by 3D-CNN and the temporal features captured by LSTM, and outputs yield prediction results through neural network calculation.

[0026] In practical applications, the advantages of 3D-CNN and LSTM are leveraged to extract spatial and temporal features respectively. These features are then fused through fully connected layers for prediction, improving the ability to characterize complex reservoirs and dynamic production processes, thereby enhancing the accuracy of production prediction. 3D-CNN can effectively extract spatial features of reservoirs, while LSTM can accurately capture temporal features of production. The combination of the two significantly improves the ability to characterize complex reservoirs and dynamic production processes, resulting in a substantial increase in production prediction accuracy. Compared to single-model prediction, the prediction error is reduced, providing a more reliable basis for development decisions.

[0027] CO2 displacement parameter optimization includes: The replacement efficiency of CO2 on methane was determined by isothermal adsorption experiments. Under controlled temperature and other conditions in the laboratory, the adsorption amounts of CO2 and methane on shale samples under different pressures were measured to determine the replacement efficiency of CO2 on methane and understand the ability of CO2 to displace methane. Based on true triaxial fracturing experiments, the CO2 injection rate and slug combination were optimized. The in-situ stress conditions of the actual reservoir were simulated by the true triaxial fracturing experiments. By changing the CO2 injection rate and slug combination, the fracture propagation and gas production were observed. The optimal CO2 injection rate and slug combination that can fully propagate the fracture and improve the displacement effect was then determined.

[0028] In practice, by using experimental methods, key parameters and optimization criteria related to CO2 displacement can be obtained, providing scientific support for the formulation of CO2 displacement schemes in actual production, improving the efficiency and effectiveness of CO2 displacement. The key parameters and optimization criteria obtained through experiments make the formulation of CO2 displacement schemes more scientific, effectively improving the efficiency and effectiveness of CO2 displacement, increasing shale gas production while increasing CO2 sequestration, and achieving a win-win situation for both economic and environmental benefits.

[0029] Fracturing parameter optimization includes: Using net present value (NPV) as the objective function, and constraining a fracture complexity index ≥ 1.5 and a carbon sequestration efficiency ≥ 60%, the optimization of fracturing parameters uses NPV as the objective function. NPV considers both the project's revenue and costs, reflecting its economic value. Simultaneously, a fracture complexity index ≥ 1.5 is constrained to ensure sufficient fracture network complexity and improve gas conductivity; a carbon sequestration efficiency ≥ 60% is constrained to achieve better carbon sequestration results. A multi-objective optimization algorithm is adopted to achieve multi-objective optimization. This algorithm combines the advantages of policy gradient in continuous space optimization, the global search capability of genetic algorithm and the fast convergence of particle swarm algorithm to optimize fracturing parameters in multiple objectives and maximize net present value under the premise of satisfying constraints.

[0030] In practical operation, reasonable objective functions and constraints are set, and efficient intelligent optimization algorithms are used to optimize fracturing parameters. While ensuring the development effect and carbon sequestration effect, the economic benefits are maximized. Using net present value as the objective function and combined with reasonable constraints, the optimization of fracturing parameters ensures that the economic benefits are maximized while taking into account the development effect and carbon sequestration effect. The application of strategy gradient-genetic-particle swarm optimization algorithm improves the optimization efficiency and the accuracy of the optimization results, and can find the optimal solution faster than traditional optimization algorithms.

[0031] Dynamic penetration tracking includes: Based on digital core technology, a stress-sensitive model is constructed for dynamic permeability tracking. By performing CT scans on shale cores, three-dimensional pore structure data of the cores are obtained, and a digital core model is constructed. Then, combined with rock mechanics theory, a stress-sensitive model is established to describe the changes in matrix permeability under different stress conditions. By combining the lattice Boltzmann method to simulate the stress-permeability coupling effect, and by simulating the flow of fluid in complex porous media based on the kinetic theory of molecules, the permeability changes under the stress-permeability coupling effect can be obtained by simulating the flow of fluid in the matrix and cracks under different stresses.

[0032] In practical operation, digital core technology and the lattice Boltzmann method are used to dynamically track permeability under different stress conditions, providing more accurate permeability parameters for accurate prediction of gas flow and production. This achieves dynamic tracking of permeability under different stress conditions, and the obtained permeability parameters are more accurate. This overcomes the prediction bias caused by the fixed permeability parameters in traditional methods, making gas flow and production predictions more accurate and providing more reliable parameter support for adjusting development plans.

[0033] The computational efficiency improvements of the dual-medium dynamic coupling model include: A hybrid parallel mode of MPI+OpenMP is adopted to solve the discrete crack network model. In improving the computational efficiency of the dual-medium dynamic coupling model, the hybrid parallel mode of MPI+OpenMP is adopted. MPI is responsible for inter-process communication and task allocation, and the computational task of the discrete crack network model is distributed to multiple processes for parallel processing. OpenMP performs thread-level parallelism within each process, making full use of the computing power of multi-core processors and accelerating the solution speed of the discrete crack network model. Model reduction is achieved by combining orthogonal decomposition and Kriging interpolation. Orthogonal decomposition reduces the dimensionality of high-dimensional data and extracts key features, while Kriging interpolation uses known data to estimate unknown data. The model is then reconstructed in the reduced-dimensional space, reducing the computational load and improving computational efficiency.

[0034] In practice, parallel computing and model reduction techniques reduce computation time and resource consumption, improve the computational efficiency of the production forecasting model, and enable it to respond to actual production needs more quickly. The application of MPI+OpenMP hybrid parallel mode and model reduction techniques significantly reduces computation time and resource consumption, improves the computational efficiency of the production forecasting model, enables the model to respond quickly to actual production needs, facilitates timely adjustment of production strategies, and improves the timeliness and flexibility of development.

[0035] The digital twin system for shale gas fracturing horizontal wells includes: Real-time data acquisition module: Acquires parameters such as wellhead pressure and injection volume. In the digital twin system, the real-time data acquisition module acquires production parameters such as wellhead pressure and injection volume in real time through sensors installed at locations such as the wellhead, providing the system with real-time production data (the specific equipment and data transmission method of the real-time data acquisition module are as follows: data is collected using devices such as pressure sensors and flow sensors, and the data is transmitted to the system for processing through wireless communication technology (such as the Internet of Things). Kalman filter module: Updates model parameters and predicts production dynamics. The Kalman filter module uses this real-time data, combined with the output of the prediction model, to update and correct the model parameters, so that the model can more accurately predict production dynamics. Closed-loop optimization module: Adjusts fracturing parameters and injection strategy. Based on the updated model prediction results and actual production data, the closed-loop optimization module adjusts fracturing parameters and injection strategy to optimize the production process.

[0036] In actual operation, a closed-loop optimization system is formed through real-time data acquisition, model parameter updates, and strategy adjustments. This enables real-time monitoring and dynamic optimization of the production process, improving production efficiency and output. The digital twin system for shale gas fracturing horizontal wells forms a closed-loop optimization, which can monitor the production process in real time, update model parameters in a timely manner, and adjust strategies to keep the production process in an optimal state. Compared with the traditional ex-post adjustment mode, this significantly improves production efficiency and output and reduces production risks.

[0037] It also includes a data closed-loop update step: by collecting production data such as wellhead pressure, CO2 injection rate, and gas production in real time, the Kalman filter algorithm is used to dynamically correct key parameters in the dual-medium model, such as matrix permeability, fracture conductivity, and crossflow coefficient, to achieve continuous matching between the prediction model and the actual production status, thereby improving the stability of medium- and long-term production forecasts. In actual operation, the model parameters are dynamically corrected using real-time production data and the Kalman filter algorithm to keep the model consistent with actual production, thereby improving the reliability and stability of medium- and long-term production forecasts. Dynamically correcting the model parameters ensures that the prediction model continuously matches the actual production status, effectively improving the stability and reliability of medium- and long-term production forecasts, avoiding medium- and long-term forecast deviations caused by fixed parameters, and providing a more accurate basis for the formulation of long-term development plans.

[0038] It also includes multi-scenario prediction and scheme selection steps: Based on the optimized dual-medium model, different combinations of fracturing parameters and CO2 injection strategies are set to generate at least three sets of production dynamic prediction curves. Through comprehensive evaluation of net present value, carbon sequestration efficiency, and recovery rate indicators, the optimal development scheme is output. In actual operation, by setting multiple scenarios and conducting predictions and evaluations, the advantages and disadvantages of different schemes are comprehensively considered, and the optimal development scheme that performs well in terms of economy, environment, and resource utilization is selected, providing decision support for actual production. Multi-scenario prediction and comprehensive evaluation can comprehensively consider the advantages and disadvantages of different schemes. The selected optimal development scheme performs well in terms of economy, environment, and resource utilization, avoiding the limitations of a single scheme, improving the scientificity and rationality of the development scheme, and providing better decision support for actual production.

[0039] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for predicting the production of a shale gas fractured horizontal well based on a dual media model, characterized in that, Includes the following steps: S1. Construct a dual-medium dynamic coupling model that considers adsorption and desorption, stress sensitivity and CO2 displacement; S2. A 3D-CNN+LSTM hybrid model is used for spatiotemporal feature extraction and yield prediction. S3. Optimize fracturing parameters and CO2 injection scheme by combining strategy gradient-genetic-particle swarm optimization algorithm; The dual-medium dynamic coupling model includes: Matrix system equations: ； Crack system equations: ； in, C represents matrix porosity. tm P is the total compressibility coefficient of the matrix. m t is the matrix pressure; k is the time; t is the matrix pressure; t is the time; k is the matrix pressure. m For matrix permeability; This refers to the viscosity of the gas. p is the crosstalk coefficient; f For crack pressure; q des q represents the desorption volume; CO2 The amount of desorbed gas is replaced by CO2; For crack porosity; c tf k is the total compressibility coefficient of the crack. f q represents the crack permeability. inj This refers to the CO2 injection rate; The optimization of fracturing parameters includes: Using net present value as the objective function, the crack complexity index is constrained to be ≥1.5, and the carbon sequestration efficiency is constrained to be ≥60%. Multi-objective optimization is achieved using a policy gradient-genetic-particle swarm optimization algorithm.

2. The method for predicting shale gas fracturing horizontal well production based on a dual-medium model according to claim 1, characterized in that, The 3D-CNN+LSTM hybrid model includes: 3D convolutional layers: extracting spatial features of reservoir heterogeneity; LSTM layer: Captures the temporal characteristics of production dynamics; Fully connected layer: fuses spatiotemporal features and outputs yield prediction results.

3. The method for predicting shale gas fracturing horizontal well production based on a dual-medium model according to claim 1, characterized in that, Parameter optimization for CO2 displacement includes: The replacement efficiency of CO2 on methane was determined by isothermal adsorption experiments. Optimization of CO2 injection rate and slug combination based on true triaxial fracturing experiments.

4. The method for predicting shale gas fracturing horizontal well production based on a dual-medium model according to claim 1, characterized in that, The computational efficiency improvement of the dual-medium dynamic coupling model includes: The discrete crack network model is solved using a hybrid parallel mode of MPI+OpenMP. Model order reduction is achieved by combining orthogonal decomposition and Kriging interpolation.

5. The method for predicting shale gas fracturing horizontal well production based on a dual-medium model according to claim 1, characterized in that, The digital twin system for shale gas fracturing horizontal wells includes: Real-time data acquisition module: acquires wellhead pressure and injection volume parameters; Kalman filter module: updates model parameters and predicts production dynamics; Closed-loop optimization module: Adjusts fracturing parameters and injection strategy.

6. The method for predicting shale gas fracturing horizontal well production based on a dual-medium model according to claim 1, characterized in that, It also includes a data closed-loop update step: by collecting production data such as wellhead pressure, CO2 injection volume, and gas production in real time, the Kalman filter algorithm is used to dynamically correct key parameters such as matrix permeability, fracture conductivity, and crossflow coefficient in the dual-medium model, so as to achieve continuous matching between the prediction model and the actual production status and improve the stability of medium and long-term production prediction.

7. The method for predicting shale gas fracturing horizontal well production based on a dual-medium model according to claim 1, characterized in that, It also includes multi-scenario prediction and scheme optimization steps: based on the optimized dual-medium model, different combinations of fracturing parameters and CO2 injection strategies are set to generate at least 3 sets of production dynamic prediction curves. Through comprehensive evaluation of net present value, carbon sequestration efficiency and recovery rate indicators, the optimal development scheme is output.

Images