Shale oil three-dimensional well network well location multi-objective optimization method

Abstract

The invention belongs to the technical field of shale oil horizontal well three-dimensional development, and particularly discloses a shale oil three-dimensional well network well location multi-objective optimization method which is used for solving the problem of how to achieve multi-objective collaborative optimization of the recovery ratio, the economic benefit and the inter-well interference degree. Comprising the following steps: (1) constructing a multi-field coupling dynamic oil drainage volume characterization model; (2) establishing an inter-well interference degree quantitative evaluation system based on the dynamic oil drainage volume characterization model; (3) realizing multi-source data fusion and high-precision prediction on the dynamic oil drainage volume, the recovery ratio, the net present value and the inter-well interference degree by combining a CNN-LSTM deep learning prediction model; and (4) implementing multi-objective optimization based on an NSGA-III multi-objective optimization algorithm, and comprehensively balancing the recovery ratio, the net present value and the inter-well interference degree on the premise of meeting engineering and economic constraints. Reliable technical support is provided for decision deployment of oil reservoir engineers under complex development conditions.

Description

Technical Field

[0001] The invention belongs to the technical field of three-dimensional development of shale oil horizontal wells, and in particular relates to a multi-objective optimization method for well locations of a three-dimensional well network of shale oil. Background Art

[0002] As unconventional oil reservoirs are increasingly urgent to increase reserves and production in deep and complex formations, the relationship between formation pressure, fracture network and well interference is becoming more and more complex. If the well interference characteristics cannot be accurately assessed and the dynamic seepage law cannot be mastered, it is easy to lead to insufficient reservoir utilization or excessive density of ineffective wells, resulting in problems such as aggravated production decline and poor economic benefits.

[0003] At present, the optimization of well pattern layout usually relies on numerical simulation methods, which establish geological models and simulate and predict pressure fields and seepage fields. Although it has certain effects in conventional oil fields, it is still insufficient in dealing with microseismic fracture monitoring data of shale reservoirs, heterogeneous reservoirs, and coupled analysis of real-time production data. Some methods for manually evaluating the degree of interference between wells (such as pressure field reconstruction or fluid saturation tracking) may be time-consuming and laborious, and are difficult to meet the needs of fast and large-scale well pattern combination optimization.

[0004] In recent years, the rapid development of artificial intelligence has introduced new ideas for reservoir engineering. Combining artificial intelligence methods with numerical simulation or fracture mechanics models can dynamically predict and evaluate different well locations and construction parameters in a relatively short period of time, providing a new technical route for the optimization of multi-well locations and multi-objective well networks. However, how to balance reservoir recovery, economic benefits, and inter-well interference in high-dimensional, nonlinear, and multi-constrained optimization problems, while taking into account geological uncertainties, is still a difficult problem that needs to be tackled. Summary of the invention

[0005] The purpose of the present invention is to provide a multi-objective optimization method for well locations in a shale oil three-dimensional well network, which effectively solves the key technical problems in the development of a three-dimensional well network in shale oil reservoirs: how to establish an accurate well interference evaluation method under the conditions of strong reservoir heterogeneity and complex mutual influence between wells, and on this basis achieve multi-objective collaborative optimization of recovery rate, economic benefit and well interference degree.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A multi-objective optimization method for well locations in a shale oil three-dimensional well network comprises the following steps: S1, constructing a multi-field coupled dynamic oil leakage volume characterization model based on reservoir pressure field evolution, fluid seepage characteristics and fracture network response to achieve quantitative and dynamic evaluation of well-to-well interference in shale oil reservoirs.

[0008] S2. Based on the multi-field coupling dynamic oil drainage volume characterization model, establish a quantitative evaluation system for well interference degree. Use the quantitative evaluation system for well interference degree to measure the spatial overlap degree of the oil drainage volume of each well and the pressure field difference, providing a quantitative basis for well pattern layout and interference control.

[0009] S3. Combine the CNN-LSTM deep learning prediction model to achieve multi-source data fusion and high-precision prediction of dynamic oil drainage volume, recovery rate, net present value, and well interference degree.

[0010] S4. Implement multi-objective optimization based on the NSGA-III multi-objective optimization algorithm. Under the premise of meeting engineering and economic constraints, comprehensively balance the recovery rate, net present value, and well interference degree. Through the organic combination of the CNN-LSTM deep learning prediction model and multi-objective optimization, finally obtain the optimal well spacing and layer spacing layout that takes into account the recovery rate, net present value, and well interference degree.

[0011] Furthermore, in step S1, based on the multi-phase seepage theory, establish a dynamic oil drainage volume calculation equation:

[0012] V leak ′(t) = φ(p)·[S o,l,m,n (t + 1) - S o,l,m,n (t)]·V eff (p)·△p eff (t);

[0013] Among them, V leak ′(t) represents the dynamic oil drainage volume at time t, φ(p) represents the pressure-sensitive function, p represents the formation pressure, S o,l,m,n is the oil-phase saturation of the corresponding grid block under the (l, m, n) spatial coordinates, V eff (p) represents the effective stimulation volume varying with pressure, and △p eff (t) is the production pressure difference at time t.

[0014] The numerical simulation process of the multi-field coupling dynamic oil drainage volume characterization model is as follows: First, collect rock physical parameters, including but not limited to porosity, permeability, and elastic modulus, and combine the stress field distribution to build a three-dimensional geomechanics model; Second, use microseismic monitoring data and fracturing pumping data to correct the fracture network and generate a fracture geometry close to the actual field; Then, use numerical simulation to obtain the dynamic distribution of pressure and saturation and the fracture stimulation volume of different grids; Finally, substitute the porosity, permeability, formation pressure, dynamic distribution of saturation, and fracture stimulation volume of different grids into the dynamic oil drainage volume calculation equation to calculate the spatio-temporal distribution characteristics and evolution laws of the dynamic oil drainage volume at each time point.

[0015] Furthermore, in step S2, the quantitative evaluation system for well interference degree includes the overlapping degree of oil drainage volume, the pressure field difference coefficient, and the time-weighted function.

[0016] The calculation formula for the overlapping degree of oil drainage volume is:

[0017] V overlap (t) = V A (t) ∩ V B (t);

[0018]

[0019] Among them, V overlap represents the overlapping area of the dynamic oil drainage volume, V A (t) and V B (t) respectively represent the dynamic oil drainage volumes of Well A and Well B at time t, and Roverlap represents the proportion of the overlapping degree of oil drainage volume.

[0020] The calculation formula for the pressure field difference coefficient is:

[0021]

[0022] Among them, D pressure represents the pressure field difference coefficient, N is the total number of pressure field grid cells, p A (k, t) and p B (k, t) are the pressure values of the k-th grid cell of Well A and Well B at time t respectively.

[0023] The calculation formula for the time-weighted function is:

[0024] w(t) = e -αt ;

[0025] Among them, w represents the time-weighted function, and α is the attenuation coefficient.

[0026] The comprehensive evaluation index I total for the well interference degree is calculated as:

[0027]

[0028] Among them, T is the total time step, and both β and γ are weight coefficients.

[0029] When 0 ≤ I total < 0.3, the well interference level is weak interference; when 0.3 ≤ I total < 0.6, the well interference level is medium interference; when 0.6 ≤ I total ≤ 1, the well interference level is strong interference.

[0030] Further, in step S3, the architecture of the CNN-LSTM deep learning prediction model includes a CNN module and an LSTM module.

[0031] Among them, the CNN module uses a convolutional neural network to process geological parameters through multi-layer convolution and pooling operations to extract the spatial distribution characteristics of the parameter field; the geological parameters include porosity, permeability, oil saturation, formation pressure, and fracture stimulation volume.

[0032] The LSTM module uses a long short-term memory network to capture the time-dependent relationship of production dynamics for time series data including historical production and pressure changes, and predicts the evolution trend of dynamic drainage volume and well interference degree over time through a two-layer LSTM structure.

[0033] The CNN-LSTM deep learning prediction model deeply fuses the spatial features extracted by the CNN module, the temporal features captured by the LSTM module, and tabular data, makes full use of multi-source information, and realizes collaborative prediction. Finally, it predicts the dynamic drainage volume, recovery rate, net present value, and well interference degree.

[0034] Further, in step S4, a three-dimensional well pattern optimization model based on the NSGA-III multi-objective optimization algorithm is constructed. The construction method is as follows: S41. Determine the objective function: The objective is to maximize the recovery rate max f1(x), maximize the net present value max f2(x), and minimize the well interference degree min f3(x).

[0035] S42. Determine the decision variables: The horizontal well spacing and vertical well spacing are used as decision variables to optimize the variable set x:

[0036] x = {(X1, Y1), (X2, Y2), …, (X n , Y n ), d h ), Δh}};

[0037] Among them, (X i , Y i ) is the well location coordinate of the i-th well, i = 1, 2, …, n; d h is the horizontal spacing between adjacent wells; Δh is the vertical layer spacing between adjacent wells.

[0038] S43. Determine the constraint conditions:

[0039] Reservoir and geological constraints, the well positions are not deployed in areas with porosity lower than 4% or inaccessible areas:

[0040]

[0041] Among them, is the minimum allowable horizontal well spacing, is the maximum allowable horizontal well spacing, is the minimum allowable horizontal well interval, is the maximum allowable horizontal well interval.

[0042] The well location and interval should avoid interbeds and fault fracture zones:

[0043] d h,min ≤d h ≤d h,max , Δh min ≤Δh≤Δh max ;

[0044] where d h,min is the minimum allowable horizontal well spacing, d h,max is the maximum allowable horizontal well spacing, Δh min is the minimum allowable horizontal well interval, Δh max is the maximum allowable horizontal well interval.

[0045] The total investment cost under economic constraints does not exceed the budget ceiling:

[0046]

[0047] where Cost(X i , Y i ) is the total cost of deploying wells at the coordinate position (X i , Y i ), and Budget max is the maximum investment budget ceiling allowed for the project.

[0048] Furthermore, in step S4, the design and solution method of the NSGA-III multi-objective optimization algorithm includes:

[0049] S51. Population initialization: Using the Latin hypercube sampling method, randomly generate an initial scheme of well spacing and interval that meets the constraint conditions, covering the design space.

[0050] S52. Fitness evaluation and reference point generation: Call the CNN-LSTM deep learning prediction model to predict the recovery factor, net present value, and well interference degree of each well pattern layout scheme, calculate the fitness of each individual according to the objective function value, and sort the individuals in combination with the effect of well spacing and interval adjustment.

[0051] S53. Solution set evolution and reference point distribution: Generate reference points in the "recovery factor - net present value - well interference degree" objective space to guide the distribution of Pareto front solutions and ensure a reasonable trade-off between well spacing and interval.

[0052] S53. Crossover Mutation and Iterative Update: Using simulated binary crossover and multi-point mutation operations, iteratively adjust the well spacing and layer spacing to optimize the objective value.

[0053] S54. Convergence Criterion: Set the maximum number of iterations or the convergence condition of the Pareto front solutions, and output a set of optimal well patterns that are balanced and compromised in terms of recovery factor, net present value, and well interference degree.

[0054] Furthermore, after solving the NSGA-III multi-objective optimization algorithm, a set of Pareto front solutions is obtained. Each solution corresponds to a set of well pattern deployment and engineering parameter combinations. The methods for analyzing and comparing and evaluating the optimization results include:

[0055] S61. The optimization results are presented in the three-dimensional objective space in the form of a Pareto front. Each solution in the Pareto front represents a well pattern optimization scheme, and the spatial distribution of the scheme reflects the trade-off relationship between different objectives in multi-objective optimization.

[0056] S62. Select several representative solutions from the Pareto optimal solution set for analysis. Each solution corresponds to a different well pattern layout scheme. The development decision-makers select the most suitable scheme from the complete Pareto optimal solution set based on actual needs and preferences.

[0057] S63. Sensitivity and Uncertainty Analysis: For the finally selected most suitable scheme, by changing the well spacing, layer spacing, number of wells, and horizontal section length, observe the influence degree on the recovery factor, net present value, and well interference degree.

[0058] S64. Field Test and Model Iteration: According to the comparison between the monitoring results and the predicted values, continuously correct the CNN-LSTM deep learning prediction model and the well interference degree evaluation parameters to form a closed loop of "numerical simulation - deep learning - multi-objective optimization - field feedback".

[0059] Compared with the prior art, the beneficial technical effects of the present invention are:

[0060] (1) The present invention integrates the dynamic drainage volume theory and the quantification of well interference degree: Traditional well pattern design is difficult to comprehensively measure the reservoir utilization degree and well pressure interference. The present invention tightly couples the dynamic seepage characteristics of shale oil reservoirs with the well-to-well competition relationship through the dynamic drainage volume characterization model and the comprehensive evaluation system of interference degree, greatly improving the accuracy of interference identification and avoiding the economic losses caused by the arrangement of ineffective well density.

[0061] (2) The present invention introduces a CNN-LSTM deep learning prediction model to improve the prediction efficiency and accuracy: converting three-dimensional geological attributes, microseismic data, and historical production data into trainable samples, integrating the spatial feature extraction of the CNN module and the temporal feature learning of the LSTM module, significantly shortening the evaluation time for multi-well layout; having better consistency and generalization ability compared with pure numerical simulation.

[0062] (3) The multi-objective optimization framework of the present invention is perfect, taking into account the recovery factor, NPV, and well interference degree: the present invention uses the NSGA-III multi-objective optimization algorithm to jointly optimize the three major objectives, not only paying attention to the final oil reservoir production (recovery factor) and economic benefits (net present value), but also effectively controlling the well interference degree, ensuring the balance between the engineering feasibility and economic benefits of the development plan. Description of the Drawings

[0063] Figure 1 It is a schematic diagram of the simulated results of the dynamic drainage volume of the block in Example 1 at the initial production stage and currently.

[0064] Figure 2 It is a comparison effect diagram of the predicted values and true values of the CNN-LSTM deep learning prediction model in Example 1 on different targets.

[0065] Figure 3 It is a schematic diagram of the Pareto front of the optimization results in Example 1.

[0066] Figure 4 It is a comparison effect diagram of the recovery factors of the basic plan and the optimized plan in Example 1.

[0067] Figure 5 It is a comparison effect diagram of the net present values of the basic plan and the optimized plan in Example 1.

[0068] Figure 6 It is a comparison effect diagram of the well interference degrees of the basic plan and the optimized plan in Example 1. Detailed Implementation Modes

[0069] Example 1: Select a certain block of an oilfield as the research object of this example. This block is a typical shale oil reservoir with the characteristics of "low porosity, low permeability, and rich organic matter": the average porosity is about 6.9%, the matrix permeability is 0.433×10 -3 μm 2 , the original formation pressure is 58.2 MPa, the reservoir burial depth is 3500 - 3750 m, and natural fractures are relatively developed in the area. This block adopts a three-dimensional development mode to achieve efficient production through multi-layer system well placement; the actual well placement includes 5 horizontal production wells, the wellbore trajectory is deployed along the dominant stress direction, the horizontal section length is 2000 - 3000 m, and the production wells adopt a constant-pressure depletion production method.

[0070] The multi-objective optimization method for shale oil vertical well pattern well positions provided in this embodiment includes the following steps: S1. Construct a multi-field coupling dynamic drainage volume characterization model based on reservoir pressure field evolution, fluid seepage characteristics, and fracture network response to quantitatively and dynamically evaluate the well interference in the shale oil reservoir.

[0071] (1) Definition and physical meaning of dynamic drainage volume (DDV)

[0072] During the development of the shale oil reservoir, affected by geological conditions and development dynamics, the formation pressure and fluid distribution inside the reservoir show complex evolution. DDV aims to quantify the effective drainage area at different production stages of the reservoir. DDV is accurately characterized by a multi-field coupling dynamic drainage volume characterization model of the reservoir pressure field, seepage characteristics, and fracture network response, dynamically reflecting the spatio-temporal variation characteristics of the effective drainage area in the reservoir. The multi-field coupling dynamic drainage volume characterization model integrates the following core dynamic characteristics: ① Dynamic evolution characteristics of pressure propagation: revealing the pressure diffusion law of the reservoir under development dynamics; ② Seepage law of oil-water two-phase fluid: quantifying the spatial distribution of fluid saturation changes; ③ Dynamic response of the fracture network: dynamically capturing the fracture geometry and its interaction with the pressure field and seepage field.

[0073] By constructing the multi-field coupling dynamic drainage volume characterization model, the spatial distribution characteristics of reservoir development benefits can be intuitively and quantitatively presented, providing a scientific basis for optimizing well spacing parameters and development strategies.

[0074] (2) Achieving high-precision calculation of dynamic drainage volume through the multi-field coupling dynamic drainage volume characterization model: ① Three-dimensional geological model coupling: Construct a three-dimensional reservoir model based on actual geological data, including geological parameters such as porosity, permeability, and natural fracture distribution; ② Geological mechanics and seepage coupling simulation: Use numerical simulation software to simulate the stress field and fracture network after hydraulic fracturing transformation of the reservoir, and calculate the pressure field and seepage field distributions at different time steps.

[0075] Based on the multi-phase seepage theory, establish a dynamic drainage volume calculation equation:

[0076] V leak ′(t)=φ(p)·[S o,l,m,n (t + 1)-S o,l,m,n (t)]·V eff (p)·△p eff (t);

[0077] Where, V leak ′(t) represents the dynamic drainage volume at time t, φ(p) represents the pressure-sensitive function, p represents the formation pressure, S o,l,m,n is the oil-phase saturation of the corresponding grid block at the (l, m, n) spatial coordinates, Veff (p) represents the effective transformation volume varying with pressure, Δp eff (t) is the production pressure difference at time t.

[0078] The calculation equation of the actual dynamic drainage volume can be corrected according to specific numerical simulation scenarios by combining factors such as porosity, oil saturation, and fracture extension scale.

[0079] The numerical simulation process of the multi - field coupling dynamic drainage volume characterization model is as follows: First, collect the rock physical parameters of the block (porosity, permeability, elastic modulus, Poisson's ratio, fracture distribution, etc.), and combine with the stress field distribution to build a three - dimensional geomechanics model; Second, use micro - seismic monitoring data and fracturing injection data to correct the fracture network and generate a fracture geometry closer to the actual field; Then, use numerical simulation to obtain the dynamic distribution of pressure and saturation and the fracture transformation volume of different grids; Finally, substitute the dynamic distribution of porosity, permeability, formation pressure, saturation, and fracture transformation volume of different grids into the dynamic drainage volume calculation equation to calculate the spatio - temporal distribution characteristics and evolution laws of the dynamic drainage volume at each time point.

[0080] The verification of the multi - field coupling dynamic drainage volume characterization model includes the following steps: Use micro - seismic monitoring data to verify the initial fracture geometry and transformation volume; Achieve accurate characterization of the well - to - well interference characteristics based on the history matching of production dynamic data. The verification results show that the multi - field coupling dynamic drainage volume characterization model has a high description ability for the dynamic response characteristics of shale reservoir, and its fitting accuracy is above 85%, providing a reliable basis for well pattern optimization.

[0081] The specific steps are as follows: First, input the rock mechanics parameters of the block (elastic modulus, Poisson's ratio), porosity, permeability, fracture distribution, etc. into the geomechanics - seepage coupling model to obtain the pressure field distribution and fracture geometry at each time step. Second, use micro - seismic monitoring data and fracturing injection data to correct the fracture network and generate a fracture geometry closer to the actual field. Finally, after completing the coupled iterative calculation at each time step, according to the dynamic drainage volume calculation equation, output the effective production volume corresponding to each well in the block, as Figure 1 shown. By comparing the micro - seismic monitoring results with the production history data, it is confirmed that DDV has a high simulation accuracy (error controlled within 10% - 15%) in the early stage of fracturing and the initial stage after production.

[0082] S2. Based on the multi - field coupling dynamic drainage volume characterization model, establish a quantitative evaluation system for well - to - well interference degree, and use the quantitative evaluation system for well - to - well interference degree to measure the spatial overlap degree of the drainage volume of each well and the pressure field difference, providing a quantitative basis for well pattern layout and interference control.

[0083] In the process of three-dimensional development of shale reservoirs, a reasonable evaluation of the well interference degree is an important prerequisite for optimizing the well pattern layout. Therefore, based on the multi-field coupling dynamic drainage volume characterization model, this embodiment proposes a quantitative evaluation system for well interference degree to scientifically measure the interference degree among multiple wells.

[0084] (1) The spatial overlap degree of the drainage volume is an important index to measure the well interference intensity. In this embodiment, the drainage volume overlap degree is used to quantify the dynamic well interference.

[0085] The calculation formula for the drainage volume overlap degree is:

[0086] V overlap (t) = V A (t) ∩ V B (t);

[0087]

[0088] Among them, V overlap represents the overlapping area of the dynamic drainage volume, V A (t) and V B (t) respectively represent the dynamic drainage volumes of Well A and Well B at time t, and R overlap represents the proportion of the drainage volume overlap degree. The numerical range of the proportion of the drainage volume overlap degree is from 0 to 1, and the larger the value, the more significant the well interference.

[0089] (2) To further reflect the degree of dynamic well interference, this embodiment introduces the pressure field difference coefficient. The larger the pressure field difference coefficient, the more significant the difference in the well pressure field, and the greater the intensity of the well interference is reflected. The calculation formula for the pressure field difference coefficient is:

[0090]

[0091] Among them, D pressure represents the pressure field difference coefficient, N is the total number of pressure field grid cells, p A (k, t) and p B (k, t) are the pressure values of the k-th grid cell of Well A and Well B at time t respectively.

[0092] (3) Time-weighted overlap degree calculation: To comprehensively reflect the well interference characteristics in different production stages, this embodiment designs a time-weighted function w(t) to dynamically adjust the weight of the interference evaluation. Its calculation formula is:

[0093] w(t) = e -αt ;

[0094] Among them, w represents the time-weighted function, and α is the attenuation coefficient.

[0095] Combined with the drainage volume overlap degree and the pressure field difference coefficient, a comprehensive evaluation index I of the interference degree between wells is defined. total , and its calculation formula is:

[0096]

[0097] Among them, T is the total time step; β and γ are weight coefficients used to balance the contributions of the drainage volume overlap degree and the pressure field difference coefficient to the interference degree.

[0098] (4) Evaluation index system for the interference degree between wells: Based on the above calculations, in this embodiment, an evaluation system for the interference degree between wells including the drainage volume overlap degree, the pressure field difference coefficient, and the time-weighted function is established. By comprehensively considering these three indicators, the interference degree between wells can be divided into three levels, as shown in Table 1.

[0099] Table 1 Classification criteria for the interference degree between wells

[0100]

[0101] In this embodiment, 5 horizontal wells in this block are selected as the target well group, and the DDV distributions of each well at the same moment are superimposed and the pressure field difference coefficient is calculated respectively to obtain the comprehensive evaluation index I of the interference degree between wells. total . In the first 6 months of the simulation period, due to the influence of fracture interference and fracturing transformation, the I value of some well groups total reached 0.55 - 0.60, showing a medium to strong interference level; with the adjustment of the production system and the pressure decay, after 1 year of production, the interference degree gradually decreased to about 0.35, indicating that the interference will be alleviated in time sequence, which is in line with the on-site observation phenomenon.

[0102] S3. Combine the CNN-LSTM deep learning prediction model to achieve multi-source data fusion and high-precision prediction of the dynamic drainage volume, recovery rate, net present value (NPV), and the interference degree between wells.

[0103] To solve the problem of low calculation efficiency in optimizing the well pattern layout under complex geological conditions, this embodiment proposes a CNN-LSTM deep learning prediction model based on the convolutional neural network (CNN) and the long short-term memory network (LSTM). This model integrates multi-dimensional geological parameters and production dynamic data, and can quickly improve the collaborative prediction ability of the dynamic drainage volume, recovery rate, net present value, and the interference degree between wells through efficient multi-dimensional feature extraction and sequence learning.

[0104] (1) Architecture design of the CNN-LSTM deep learning prediction model.

[0105] I. CNN module: Spatial feature extraction.

[0106] Use a convolutional neural network to process geological parameters (such as porosity, permeability, fracture distribution, etc.). Through multi-layer convolution and pooling operations, extract the spatial distribution characteristics of the parameter field.

[0107] By converting geological parameter fields (such as permeability fields, oil saturation fields, etc.) into RGB three-channel images, capture the dynamic changes of reservoir heterogeneity and fracture networks, and improve the accuracy of reservoir modeling.

[0108] II. LSTM Module: Temporal Feature Learning.

[0109] For time series data such as historical production and pressure changes, use LSTM to capture the time-dependent relationships of production dynamics. Through a two-layer LSTM structure, predict the evolution trends of dynamic drainage volume and well interference degree over time.

[0110] III. Multi-source Data Fusion: Deeply fuse the spatial features extracted by the CNN module, the temporal features captured by the LSTM module, and tabular data (such as well locations, horizontal section lengths, etc.). Make full use of multi-source information to achieve collaborative prediction, and finally predict the dynamic drainage volume, recovery factor, net present value, and well interference degree. The design of multi-source data fusion enables the model to adapt to different geological conditions and engineering parameters and generate targeted prediction results.

[0111] (2) Model Training: In this embodiment, a total of 400 sets of reservoir numerical simulation schemes are designed and generated. Among them, 320 sets are used to train the basic framework of the CNN-LSTM deep learning prediction model, 40 sets are used to determine the hyperparameters of the hybrid neural network, and the remaining 40 sets are used to test the generalization ability of the CNN-LSTM deep learning prediction model. The production time period of each simulation scheme is 3 years, and the dynamic adjustment period of the single-well working system is set to 2 months, with a total of 18 time steps to fully reflect the dynamic response characteristics of the reservoir.

[0112] The training data includes the following main inputs: ① Geological and fracture information: permeability, porosity, fracture network distribution; ② Production dynamic data: production, water cut, bottom-hole flowing pressure; ③ DDV time series: the dynamic oil infiltration volume output at each time step.

[0113] Among them, the CNN module is responsible for extracting spatial features (such as formations, fracture grids, etc.), the LSTM module is responsible for capturing time series dependencies, and finally outputs the comprehensive predicted values of the target parameters (dynamic drainage volume, net present value, recovery factor, well interference degree).

[0114] The training set was established using the Python programming language by invoking the INTERSECT numerical simulator of Schlumberger. In the specific implementation process, the os.system() function was first used to execute the preset cmd instruction sequence to achieve batch processing of the automatically generated.afi files. On this basis, sequential Gaussian simulation (SGSIM) was used to generate random geological samples to ensure data coverage and rationality.

[0115] (3) Model verification: To evaluate the accuracy of different deep learning models in predicting the dynamic oil seepage volume in the reservoir, in this embodiment, the coefficient of determination (R 2 ), mean absolute error (MAE), and root mean square error (RMSE) were selected as three indicators to conduct performance comparison tests on the CNN model, LSTM model, and CNN-LSTM deep learning prediction model. The results are shown in Table 2.

[0116] Table 2 Performance evaluation of different reservoir prediction models based on deep learning

[0117] <![CDATA[R 2 > MAE RMSE CNN model 0.881 0.0391 0.0455 LSTM model 0.853 0.0458 0.0491 CNN-LSTM deep learning prediction model 0.913 0.0248 0.0297

[0118] The test results show that the CNN-LSTM deep learning prediction model is significantly superior to the individual CNN model or LSTM model in terms of R 2 , MAE, and RMSE, indicating that the CNN-LSTM deep learning prediction model in this embodiment can capture the spatio-temporal evolution law of the dynamic oil seepage volume in the reservoir more accurately and achieve high-precision prediction.

[0119] (4) Test set results: The last 40 sets of reservoir numerical simulation schemes were used for testing. The results show that the average relative error of production prediction is less than 10%; the R 2 of well interference degree prediction can reach 0.92; the prediction of recovery factor and net present value also has high accuracy, as Figure 2 shown.

[0120] S4. Implement multi-objective optimization based on the NSGA-III multi-objective optimization algorithm. Under the premise of meeting engineering and economic constraints, comprehensively balance the recovery factor, net present value, and well interference degree. Through the organic combination of the CNN-LSTM deep learning prediction model and multi-objective optimization, finally obtain the optimal well spacing and layer spacing layout that takes into account the recovery factor, net present value, and well interference degree.

[0121] With the support of the CNN-LSTM deep learning prediction model, this embodiment proposes a three-dimensional well pattern optimization model based on the NSGA-III multi-objective optimization algorithm. This three-dimensional well pattern optimization model takes well spacing and layer spacing as key optimization parameters, comprehensively considers the recovery factor, net present value, and well interference degree, and realizes the three-dimensional optimization of the well pattern layout in the horizontal and vertical dimensions.

[0122] (1) Construct a 3D well pattern optimization model based on the NSGA-III multi-objective optimization algorithm. The construction method is as follows:

[0123] I. Determine the objective functions.

[0124] Objective 1: Maximize the recovery factor max f1(x): Improve the utilization degree of the reservoir and optimize the development effect; Objective 2: Maximize the net present value max f2(x): Considering production, oil price, investment cost, and discount rate comprehensively to optimize the economic benefits; Objective 3: Minimize the well interference degree min f3(x): Reduce the interference between horizontal wells and vertical interval wells and improve the rationality of well pattern deployment.

[0125] II. Determine the decision variables.

[0126] Take the horizontal well spacing (affecting the well interference degree and reservoir utilization range) and vertical well spacing (reflecting the development degree and economic rationality of wells in different intervals) as decision variables to optimize the variable set x:

[0127] x = {(X1, Y1), (X2, Y2), …, (X n , Y n ), d h}, Δh};

[0128] Among them, (X i , Y i ) is the well location coordinate of the i-th well, i = 1, 2, …, n; d h is the horizontal spacing between adjacent wells; Δh is the vertical interval between adjacent wells.

[0129] S43. Determine the constraint conditions:

[0130] Reservoir and geological constraints, the well location is not deployed in areas with porosity lower than 4% or inaccessible areas:

[0131]

[0132] Among them, is the minimum allowable horizontal well spacing, is the maximum allowable horizontal well spacing, is the minimum allowable horizontal well interval, is the maximum allowable horizontal well interval.

[0133] The well location and interval should avoid interbeds and fault fracture zones:

[0134] d h,min ≤ d h ≤ d h,max , Δh min ≤ Δh ≤ Δh max;

[0135] Among them, d h,min is the minimum allowable horizontal well spacing, and d h,max is the maximum allowable horizontal well spacing, and Δh min is the minimum allowable horizontal well layer spacing, and Δh max is the maximum allowable horizontal well layer spacing.

[0136] The total investment cost of the economic constraint does not exceed the budget cap:

[0137]

[0138] Among them, Cost(X i , Y i ) is the total cost of deploying wells at the coordinate position (X i , Y i ), and Budget max is the maximum investment budget cap allowed for the project.

[0139] In this embodiment, during multi-well network three-dimensional development, three objectives of recovery factor, net present value, and well interference degree are concerned; the decision variables include well position coordinates, horizontal section length, and injection-production mode.

[0140] (2) The design and solution method of the NSGA-III multi-objective optimization algorithm includes:

[0141] I. Population initialization: Adopt the Latin hypercube sampling method to randomly generate an initial scheme of well spacing and layer spacing that meets the constraint conditions, covering the design space.

[0142] II. Fitness evaluation and reference point generation: Call the CNN-LSTM deep learning prediction model to predict the recovery factor, net present value, and well interference degree of each well network layout scheme, calculate the fitness of each individual according to the objective function value, and sort the individuals in combination with the effects of well spacing and layer spacing adjustment.

[0143] III. Solution set evolution and reference point distribution: Generate reference points in the "recovery factor - net present value - well interference degree" objective space to guide the distribution of Pareto front solutions and ensure a reasonable trade-off between well spacing and layer spacing.

[0144] IV. Crossover mutation and iterative update: Use simulated binary crossover and multi-point mutation operations to iteratively adjust the well spacing and layer spacing to optimize the objective values.

[0145] V. Convergence determination: Set the maximum number of iterations or the convergence condition of the Pareto front solutions, and output a set of optimal well network layouts that are balanced and compromised in terms of recovery factor, net present value, and well interference degree.

[0146] In this embodiment, the final selection is a crossover mutation coefficient of 0.45, a mutation coefficient of 0.05, a population size of 200, and 300 iterations to output the Pareto front solutions, as Figure 3 shown.

[0147] (3) After completing the solution of the NSGA-III multi-objective optimization algorithm, a set of Pareto front solutions is obtained, and each solution corresponds to a set of well pattern deployments and engineering parameter combinations.

[0148] The methods for in-depth analysis and comparative evaluation of the optimization results include: I. The optimization results are presented in the three-dimensional objective space composed of recovery factor, net present value, and well interference degree in the form of the Pareto front. Each solution in the Pareto front represents a well pattern optimization scheme, and the spatial distribution of these schemes reflects the trade-off relationship between different objectives in multi-objective optimization.

[0149] II. Select several representative solutions from the Pareto optimal solution set for analysis to understand the characteristics of different well pattern layout schemes. Then, the decision maker can select the most suitable scheme from the complete Pareto optimal solution set based on actual needs and preferences.

[0150] III. Sensitivity and uncertainty analysis: For the finally selected most suitable scheme, by changing the well spacing, layer spacing, number of wells, and horizontal section length, observe the influence degree on the recovery factor, net present value, and well interference degree.

[0151] IV. Field test and model iteration: According to the comparison between the monitoring results and the predicted values, continuously correct the CNN-LSTM deep learning prediction model and the well interference degree evaluation parameters to form a closed loop of "numerical simulation - deep learning - multi-objective optimization - field feedback".

[0152] In the Pareto optimal solution set, three types of typical optimization schemes are selected for analysis: one is the high-production scheme, where the recovery factor is increased from 22% of the base scheme to 28%, an increase of 6%; the second is the high-benefit scheme, where the net present value is increased from 15×10 7 yuan to 20×10 7 yuan, an increase of 5×10 7 yuan; the third is the low-well-interference scheme, where the well interference degree is reduced from 0.5 to 0.2, a decrease of 0.3. As Figure 4 , Figure 5 and Figure 6 shown, by optimizing the injection-production relationship and well pattern parameters, the high-production scheme achieves a better reservoir utilization effect, the high-benefit scheme achieves a significant improvement in economy, and the low-well-interference scheme ensures development stability by regulating the injection-production ratio.

[0153] The application of this embodiment shows that this embodiment can effectively solve the problem of optimizing the three-dimensional well pattern under the condition of well interference. Its optimization scheme performs excellently in aspects such as improving the recovery rate, increasing the net present value, and reducing the well interference degree, fully demonstrating the high credibility and robustness of the three-dimensional well pattern optimization model based on the NSGA-III multi-objective optimization algorithm, and providing reliable technical support for the decision-making and deployment of reservoir engineers under complex development conditions.

[0154] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions, or substitutions made by those skilled in the art within the substantial scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A multi-objective optimization method for shale oil three-dimensional well pattern well positions, characterized in that, It includes the following steps: S1. Construct a multi-field coupled dynamic oil drainage volume characterization model based on the evolution of reservoir pressure field, fluid seepage characteristics and fracture network response to realize the quantitative and dynamic evaluation of well interference in shale reservoirs; S2. Based on the multi-field coupled dynamic oil drainage volume characterization model, establish a quantitative evaluation system for well interference degree, and use the quantitative evaluation system for well interference degree to measure the spatial overlap degree of oil drainage volume and pressure field difference of each well, providing a quantitative basis for well pattern layout and interference control; S3. Combine the CNN-LSTM deep learning prediction model to realize multi-source data fusion and high-precision prediction of dynamic oil drainage volume, recovery rate, net present value and well interference degree; S4. Implement multi-objective optimization based on the NSGA-III multi-objective optimization algorithm, comprehensively balance the recovery rate, net present value and well interference degree on the premise of meeting engineering and economic constraints, and finally obtain the optimal well spacing and layer spacing layout that takes into account the recovery rate, net present value and well interference degree through the organic combination of the CNN-LSTM deep learning prediction model and multi-objective optimization.

2. The multi-objective optimization method for shale oil vertical well pattern well positions according to claim 1, wherein, In step S1, based on the multi-phase seepage theory, a dynamic oil drainage volume calculation equation is established: V leak ′(t) = φ(p)·[S o,l,m,n (t + 1) - S o,l,m,n (t)]·V eff (p)·Δp eff (t); Among them, V leak ′(t) represents the dynamic oil drainage volume at time t, φ(p) represents the pressure-sensitive function, p represents the formation pressure, and S o,l,m,n is the oil-phase saturation of the corresponding grid block under the (l, m, n) spatial coordinates, V eff (p) represents the effective stimulation volume that changes with pressure, and △p eff (t) is the production pressure difference at time t; The numerical simulation process of the multi-field coupled dynamic oil drainage volume characterization model is as follows: First, collect rock physical parameters, including but not limited to porosity, permeability and elastic modulus, and build a three-dimensional geomechanics model in combination with the stress field distribution; Second, use microseismic monitoring data and fracturing pumping data to correct the fracture network and generate a fracture geometry close to the actual field; Then, use numerical simulation to obtain the dynamic distribution of pressure and saturation and the fracture transformation volume of different grids; Finally, substitute the porosity, permeability, formation pressure, saturation dynamic distribution and fracture transformation volume of different grids into the dynamic oil drainage volume calculation equation to calculate the spatio-temporal distribution characteristics and evolution laws of the dynamic oil drainage volume at each time point.

3. The multi-objective optimization method for shale oil vertical well pattern well positions according to claim 2, wherein, In step S2, the quantitative evaluation system for well interference degree includes the oil drainage volume overlap degree, pressure field difference coefficient and time weighting function; The calculation formula for the oil drainage volume overlap degree is: V overlap V(t) = V A V(t) ∩ V B V(t); Among them, V overlap represents the overlapping area of the dynamic drainage volume, V A (t) and V B (t) respectively represent the dynamic drainage volumes of Well A and Well B at time t, and R overlap represents the ratio of the overlapping degree of the drainage volume; The calculation formula for the pressure field difference coefficient is: Among them, D pressure represents the pressure field difference coefficient, N is the total number of pressure field grid cells, and p A (k, t) and p B (k, t) are the pressure values of the k-th grid cell of well A and well B at time t, respectively; The calculation formula for the time weighting function is: w(t) = e -αt ; Among them, w represents the time weighting function, and α is the attenuation coefficient; Comprehensive evaluation index I of interference degree between wells total The calculation formula is as follows: Among them, T is the total time step, and both β and γ are weight coefficients; When 0 ≤ I total < 0.3, the interference level between wells is weak interference; when 0.3 ≤ I total < 0.6, the interference level between wells is medium interference; when 0.6 ≤ I total ≤ 1, the interference level between wells is strong interference.

4. The multi-objective optimization method for shale oil vertical well pattern well positions according to claim 3, characterized in that In step S3, the architecture of the CNN-LSTM deep learning prediction model includes a CNN module and an LSTM module; Among them, the CNN module uses a convolutional neural network to process geological parameters, and through multi-layer convolution and pooling operations, extracts the spatial distribution characteristics of the parameter field; The geological parameters include porosity, permeability, oil saturation, formation pressure and fracture transformation volume; The LSTM module is aimed at time series data including historical production and pressure changes, uses a long short-term memory network to capture the time-dependent relationship of production dynamics, and predicts the evolution trend of dynamic oil drainage volume and well interference degree over time through a double-layer LSTM structure. The CNN-LSTM deep learning prediction model deeply integrates the spatial features extracted by the CNN module, the temporal features captured by the LSTM module, and tabular data, makes full use of multi-source information to achieve collaborative prediction, and finally predicts the dynamic oil drainage volume, recovery rate, net present value, and inter-well interference degree.

5. The multi-objective optimization method for shale oil vertical well pattern well positions according to claim 4, wherein, In step S4, a 3D well pattern optimization model based on the NSGA-III multi-objective optimization algorithm is constructed. The construction method is as follows: S41. Determine the objective functions: The objectives are to maximize the recovery rate max f1(x), maximize the net present value max f2(x), and minimize the inter-well interference degree min f3(x). S42. Determine the decision variables: The horizontal well spacing and vertical well spacing are used as decision variables, and the optimization variable set is x. x = {(X1, Y1), (X2, Y2), …, (X n , Y n ), d h , Δh}; Among them, (X i , Y i ) is the well location coordinate of the i-th well, where i = 1, 2, …, n; d h is the horizontal spacing between adjacent wells; Δh is the vertical layer spacing between adjacent wells; S43. Determine the constraint conditions: Reservoir and geological constraints: Well positions are not deployed in areas with porosity less than 4% or inaccessible areas. Among them, is the minimum allowable horizontal well spacing, is the maximum allowable horizontal well spacing, is the minimum allowable horizontal well layer spacing, is the maximum allowable horizontal well layer spacing; Well positions and intervals should avoid interbeds and fault fracture zones. d h,min ≤ d h ≤ d h,max , Δh min ≤ Δh ≤ Δh max ; where d h,min is the minimum allowable horizontal well spacing, d h,max is the maximum allowable horizontal well spacing, Δh min is the minimum allowable horizontal well layer spacing, Δh max is the maximum allowable horizontal well layer spacing; Economic constraint: The total investment cost does not exceed the budget ceiling. Among them, Cost(X i ,Y i ) is the total cost of deploying a well at the coordinate position (X i ,Y i ), and Budget max is the upper limit of the maximum investment budget allowed for the project.

6. The multi-objective optimization method for shale oil vertical well pattern well positions according to claim 5, characterized in that In step S4, the design and solution method of the NSGA-III multi-objective optimization algorithm includes: S51. Population initialization: Using the Latin hypercube sampling method, randomly generate initial well spacing and interval schemes that meet the constraint conditions to cover the design space. S52. Fitness evaluation and reference point generation: Call the CNN-LSTM deep learning prediction model to predict the recovery rate, net present value, and inter-well interference degree of each well pattern layout scheme, calculate the fitness of each individual according to the objective function values, and sort the individuals in combination with the effects of well spacing and interval adjustment. S53. Solution set evolution and reference point distribution: Generate reference points in the "recovery rate - net present value - inter-well interference degree" objective space to guide the distribution of Pareto front solutions and ensure a reasonable trade-off between well spacing and interval. S53. Crossover mutation and iterative update: Use simulated binary crossover and multi-point mutation operations to iteratively adjust the well spacing and interval to optimize the objective values. S54. Convergence determination: Set the maximum number of iterations or the convergence condition of the Pareto front solutions, and output a set of optimal well pattern layouts that are balanced and compromised in terms of recovery rate, net present value, and inter-well interference degree.

7. The multi-objective optimization method for shale oil three-dimensional well pattern well positions according to claim 6, wherein After solving the NSGA-III multi-objective optimization algorithm, a set of Pareto front solutions is obtained. Each solution corresponds to a set of well pattern deployment and engineering parameter combinations. The methods for analyzing and comparing and evaluating the optimization results include: S61. The optimization results are presented in the 3D objective space in the form of a Pareto front. Each solution in the Pareto front represents a well pattern optimization scheme, and the spatial distribution of the scheme reflects the trade-off relationship between different objectives in multi-objective optimization. S62. Select several representative solutions from the Pareto optimal solution set for analysis. Each solution corresponds to a different well pattern layout scheme. The decision maker can select the most suitable scheme from the complete Pareto optimal solution set based on actual needs and preferences. S63. Sensitivity and uncertainty analysis: For the finally selected most suitable scheme, by changing the well spacing, interval, number of wells, and horizontal section length, observe the influence degree on the recovery rate, net present value, and inter-well interference degree. S64. Field test and model iteration: According to the comparison between the monitoring results and the predicted values, continuously correct the CNN-LSTM deep learning prediction model and the evaluation parameters of the interference degree between wells, and form a closed loop of "numerical simulation - deep learning - multi-objective optimization - field feedback".

Images