A multi-objective optimization method and device for CO2 flooding and storage in a high-water-cut fault block oil reservoir

Abstract

The application discloses a kind of high water cut fault block reservoir CO2 drive oil storage multi-objective optimization methods, by establishing high water cut fault block reservoir CO2 drive oil storage fluid-solid coupling model, sample generation sample is carried out based on Latin hypercube sampling method, long short-term memory neural network and the above sample are used to build target proxy model, introduce fast non-dominated sorting genetic multi-objective optimization algorithm, based on the proxy model constructed, multiple objective functions are optimized, and the Pareto solution set of target function is obtained.Such, based on the comprehensive framework of three-dimensional reservoir percolation and geomechanics coupling model, LSTM algorithm and multi-objective optimization function NSGA-II, the injection-production well production system scheme can be obtained by optimization, the maximum of oil recovery and CO2 storage capacity and the minimization of fault slip displacement are realized, and the CO2 drive oil storage combined optimization efficiency and result accuracy of high water cut fault block reservoir are improved.

Description

Technical Field

[0001] This invention relates to the field of oil and gas field development, and in particular to a multi-objective optimization method and apparatus for CO2 flooding and storage in high water-cut fault-block reservoirs. Background Technology

[0002] Old oilfields in eastern China (such as Daqing and Shengli oilfields) have entered the high water-cut development stage. Studies have shown that CO2 flooding in high water-cut reservoirs can generally extend the reservoir's production life by 15–20 years and increase the recovery rate by 8%–15%. Furthermore, during CO2-enhanced oil recovery (CO2-EOR), a portion of CO2 is permanently sequestered underground in various forms. To achieve maximum recovery and CO2 sequestration, the production strategies of injection and production wells must be optimized. In addition, since many oilfields typically have complex faults, the geomechanical risks of fault activation also need to be assessed in these reservoirs, transforming CO2 flooding and sequestration into a multi-objective optimization problem.

[0003] Among related technologies, the technical problems of CO2 flooding and sealing in high water-cut fault-block reservoirs are low CO2 flooding capacity, small CO2 storage capacity, and high geological risk. Summary of the Invention

[0004] To address the aforementioned problems in existing technologies, this invention provides a multi-objective optimization method and apparatus for CO2 flooding and storage in high water-cut fault-block reservoirs, thereby resolving the drawbacks of existing technologies in the development of high water-cut reservoirs from water-drive to gas-drive, such as low CO2 flooding capacity, low CO2 storage capacity, and high geological risks.

[0005] To achieve the above objectives, the technical solution of this invention is as follows:

[0006] The first aspect of the present invention provides a multi-objective method for CO2 flooding and oil sequestration in high water-cut fault-block reservoirs, the method comprising:

[0007] Establish a fluid-structure interaction numerical simulation model for CO2 flooding and storage in high water-cut fault-block reservoirs;

[0008] Based on the Latin hypercube sampling method, the injection well regime is sampled as input parameters, and calculations are performed based on the CO2 oil displacement and storage fluid-structure interaction numerical simulation model to obtain production and fault slip displacement samples under different injection regimes.

[0009] The constructed samples are input into a long short-term memory neural network and trained using the Adam optimization algorithm until the root mean square error meets the preset conditions, thus obtaining the target surrogate model.

[0010] Based on the target proxy model, a multi-objective optimization algorithm is used to optimize the recovery rate objective function, the CO2 storage rate objective function, and the fault slip displacement objective function to obtain the corresponding optimization results for each objective function.

[0011] A second aspect of the present invention provides a multi-objective optimization device for CO2 enhanced oil recovery and storage in high water-cut fault-block reservoirs, the device comprising:

[0012] A module was established to create a fluid-structure interaction numerical simulation model for CO2 flooding and storage in high water-cut fault-block oil reservoirs.

[0013] The calculation module is used to sample the injection well regime based on the Latin hypercube sampling method as input parameters, and to perform calculations based on the CO2 oil displacement and storage fluid-structure interaction numerical simulation model to obtain production and fault slip displacement samples under different injection regimes.

[0014] The training module is used to input the constructed samples into the long short-term memory neural network and train it using the Adam optimization algorithm until the root mean square error meets the preset conditions to obtain the target surrogate model.

[0015] The optimization module is used to optimize the recovery rate objective function, the CO2 storage rate objective function, and the fault slip displacement objective function based on the target proxy model using a multi-objective optimization algorithm, so as to obtain the corresponding optimization result for each objective function.

[0016] In some embodiments, the establishing module is further configured to determine the linear relationship between the ultimate strength, shear stress, and normal stress of each stressed surface; the linear relationship is shown in the following formula;

[0017]

[0018] In the formula, τ n σ is the shear stress on the failure surface. n C represents the normal stress at the failure surface, and C represents the cohesion of the rock. The internal friction angle of the rock;

[0019] Based on the shear stress σ at the fault plane t and σ s Determine the maximum shear stress τ max The maximum shear stress is shown in the following formula:

[0020]

[0021] The sliding displacement of the fracture surface is determined based on the maximum shear stress; the sliding displacement is shown in the following formula:

[0022]

[0023] In the formula, K t The stiffness of the fault;

[0024] The coefficient of friction for sliding friction is determined based on the value of the sliding displacement; the coefficient of friction is shown in the following formula:

[0025]

[0026]

[0027]

[0028] In the formula, The minimum friction angle, This represents the maximum value of the fault friction force. This represents the minimum value of fault friction.

[0029] The slip tendency of the fault is determined based on the maximum shear stress and the friction coefficient; the slip tendency is shown in the following formula:

[0030]

[0031] In the formula, cohes represents the fault cohesion, and the Mohr circle shifts to the left as the pore pressure increases.

[0032] In some embodiments, the calculation module is further configured to assume a structure-function with K-dimensional basic variables, where N represents the selection of N groups of x = (x1, x2, ..., x...). n The sample is given by matrix P, which is an N*K matrix. The column elements of P are randomly generated from integers from 1 to N, and each element has a different value. Matrix R is also an N*K matrix, and the column elements of R are randomly generated from rational numbers from 0 to 1, and each element has a different value.

[0033]

[0034]

[0035] In the formula, s ij It is the element in the i-th row and j-th column of matrix S. It is the distribution function that the j-th variable conforms to. It is the j-th variable of the i-th sample group obtained by Latin hypercube sampling.

[0036] In some embodiments, the training module is further configured to train the network using a long short-term memory neural network, defining the weights and bias parameters of the gates; the weights and biases are shown in the following formula:

[0037]

[0038] Where W represents weight, b represents bias, i represents input gate, f represents forget gate, and o represents output gate;

[0039] In a single hidden unit of the Long Short-Term Memory (LSTM) neural network, a sigmoid activation function is used to determine the forget gate; the forget gate is shown in the following equation:

[0040] f t =σ(W f ·[h t-1 ,x t ]+b f ),

[0041] In the formula, f t For the forgetting gate, σ represents the activation function, expressed as σ(x) = (1 + e^(-1 / x))^(-1 / x ... -x ) -1 Its output range is 0 to 1, x t h is the input for the current time step. t-1 W is the input from the previous time step. f With b f Weights and biases for the forget gate;

[0042] The input gate update is determined based on the Sigmoid activation function, and at the current step size, it is determined whether the current input has been written to the memory cell based on the input gate; the input gate is shown in the following formula:

[0043] i t =σ(W i ·[h t-1 ,x t ]+b i ),

[0044] In the formula, i t For the input gate, b i This is the bias matrix of the input gate;

[0045]

[0046] In the formula, tanh represents a hyperbolic tangent function with an output range of -1 to 1;

[0047] Update the current cell state based on the historical cell states, as shown in the following formula:

[0048]

[0049] In the formula, C t-1 For historical cell states, C t The current cell state, A vector of new candidate values;

[0050] The output gate controls whether the information stored in the memory cell affects subsequent network layers.

[0051] o t =σ(W o [h t -1,x t ]+b o ),

[0052] In the formula, o t For output gate, W o With b o These represent the weights and biases of the output gate, h. t This is the output value for the current cell state and also the input value for the next cell state.

[0053] In some embodiments, the optimization module is further configured to use the production system parameters of the injection-production well as independent variables, and the recovery rate, CO2 sequestration rate, and fault slip displacement as dependent variables, to maximize the objective functions of the recovery rate and the CO2 sequestration rate and minimize the objective function of the fault slip displacement; the following formula is defined:

[0054] maxF(x)=[(f1(x),f2(x),-f3(x)) T ],

[0055]

[0056] In the formula, OOIP represents the initial recoverable reserves of the reservoir; Let be the oil production at time t; The total amount of CO2 injected into the injection well at time t; D represents the total CO2 produced by the production well at time t, where T is the total time for reservoir development and evaluation, and D is the total CO2 produced by the production well at time t. s This represents fault slip displacement;

[0057] The NSGA-II algorithm is used to optimize the recovery rate objective function, the CO2 sequestration rate objective function, and the fault slip displacement objective function to obtain the corresponding optimization results.

[0058] A third aspect of the present invention provides an electronic device, comprising: a memory for storing executable instructions; and a processor for executing the executable instructions stored in the memory to implement the above-described multi-objective optimization method for CO2 flooding and sequestration in high water-cut fault-block oil reservoirs.

[0059] A fourth aspect of the present invention provides a computer-readable storage medium storing executable instructions for inducing a processor to execute the executable instructions to implement the above-described multi-objective optimization method for CO2 flooding and sequestration in high water-cut fault-block oil reservoirs.

[0060] The present invention provides a multi-objective optimization method for CO2 flooding and storage in high water-cut fault-block reservoirs. First, a fluid-structure interaction numerical simulation model for CO2 flooding and storage in high water-cut fault-block reservoirs is established. Second, samples are generated through sampling and numerical simulator calculations based on the Latin hypercube sampling method. A surrogate model is constructed using a long short-term memory neural network. Finally, the NSGA-II multi-objective optimization algorithm is introduced for multi-objective optimization to obtain the Pareto solution set of the objective function. Thus, on the one hand, based on a comprehensive framework of a three-dimensional fluid reservoir seepage and geomechanical coupling model, the LSTM artificial intelligence algorithm, and the NSGA-II multi-objective optimization algorithm, the present invention can obtain a production regime scheme for injection-production wells through optimization, thereby maximizing oil recovery and CO2 storage while minimizing fault slip displacement. Attached Figure Description

[0061] Figure 1 This is a flowchart illustrating a multi-objective optimization method for CO2 flooding and oil storage in a high water-cut fault-block reservoir provided in an embodiment of the present invention.

[0062] Figure 2 This is a schematic diagram of the Mohr-Coulomb yield criterion provided in an embodiment of the present invention;

[0063] Figure 3 This is a schematic diagram of the structure of the long short-term memory neural network model provided in the embodiment of the present invention;

[0064] Figure 4 This is a flowchart illustrating the NSGA-II algorithm provided in an embodiment of the present invention;

[0065] Figure 5 is a reservoir model attribute field diagram provided in the embodiment of the present invention, (a) is the permeability field, and (b) is the porosity field;

[0066] Figure 6 This is a schematic diagram of the network structure provided in an embodiment of the present invention;

[0067] Figure 7 These are the prediction results of the long short-term memory neural network surrogate model provided in the embodiments of the present invention. (a) is the prediction result of the recovery rate, (b) is the prediction result of the CO2 sequestration rate, and (c) is the prediction result of the fault sliding displacement.

[0068] Figure 8 This is the Pareto solution set optimized by NSGA-II provided in the embodiments of the present invention;

[0069] Figure 9 is a statistical result diagram of NSGA-II optimization provided in the embodiment of the present invention. (a) is a statistical result diagram of recovery rate, (b) is a statistical result diagram of storage rate, and (c) is a statistical result diagram of fault slip displacement.

[0070] Figure 10This is a schematic diagram of the composition and structure of the multi-objective optimization device for CO2 flooding and oil storage in high water-cut fault-block oil reservoirs provided in an embodiment of the present invention;

[0071] Figure 11 This is a schematic diagram of the composition and structure of the CO2 flooding and storage multi-objective optimization device for high water-cut fault-block oil reservoirs provided in this embodiment of the invention. Detailed Implementation

[0072] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0073] In the following description, references to "some embodiments" refer to a subset of all possible embodiments; however, it is understood that "some embodiments" may be the same or different subsets of all possible embodiments and may be combined with each other without conflict. Unless otherwise defined, all technical and scientific terms used in the embodiments of the invention have the same meaning as commonly understood by one of ordinary skill in the art to which the embodiments of the invention pertain. The terminology used in the embodiments of the invention is for the purpose of describing the embodiments of the invention only and is not intended to limit the invention.

[0074] In high water-cut fault-block oil reservoirs, the water cut gradually increases due to long-term water injection development or other extraction activities, leading to a decline in oil recovery. To improve recovery, various methods have been employed, among which CO2-enhanced oil recovery and storage (EOR) technology is an effective approach. This technology injects CO2 into the reservoir, utilizing the dissolving, expanding, and viscosity-reducing effects of CO2 to improve the fluidity of the crude oil, thereby achieving effective oil extraction. Simultaneously, the injected CO2 can be stored underground, reducing greenhouse gas emissions.

[0075] The following describes an exemplary application of the multi-objective optimization device for CO2 enhanced oil recovery and storage in high water-cut fault-block reservoirs according to embodiments of the present invention. This device can be implemented as a terminal or a server. In one implementation, the device can be implemented as a laptop, tablet, desktop computer, mobile device, or other types of terminal. In another implementation, it can also be implemented as a server. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. The terminal and server can be directly or indirectly connected via wired or wireless communication, which is not limited in the embodiments of the present invention. The following will illustrate an exemplary application of the CO2 flooding and sequestration multi-objective optimization device in high water-cut fault-block reservoirs as a server.

[0076] This invention provides a multi-objective optimization method for CO2 flooding and oil storage in high water-cut fault-block reservoirs. (See also...) Figure 1 , Figure 1 This is a flowchart illustrating a multi-objective optimization method for CO2 flooding and oil storage in high water-cut fault-block reservoirs provided in this invention. Figure 1 The steps shown are explained.

[0077] Step S101: Establish a fluid-structure interaction numerical simulation model for CO2 flooding and storage in high water-cut fault-block reservoirs.

[0078] In some embodiments, a high water-cut reservoir refers to an oil reservoir in which the formation has a high water content.

[0079] In some embodiments, a fault block refers to a rock mass formed by internal crustal stress and significantly displaced along a certain fault plane. In hydrocarbon reservoir geology, a fault-block reservoir refers to a hydrocarbon accumulation unit formed by the segmentation and closure of faults.

[0080] In practical applications, when conducting fluid-structure interaction numerical simulations for CO2-driven oil recovery and storage in high water-cut fault-block reservoirs, it is necessary to fully consider the geological characteristics and fluid distribution characteristics of the fault-block reservoir, as well as the interaction between fluid flow and reservoir deformation caused by CO2 injection. This interaction is known as fluid-structure interaction.

[0081] In some embodiments, the fluid-structure interaction numerical simulation model for CO2 flooding and sequestration in high water-cut fault-block reservoirs aims to predict and assess oil recovery, CO2 sequestration rate, and the possibility and impact of fault activation by simulating changes in fluid pressure within the reservoir, the mechanical response of rocks, and the geometric and physical properties of faults during CO2 injection.

[0082] In some embodiments, the slip state of the fault can be evaluated based on the Mohr-Coulomb yield criterion, taking into account the reservoir seepage-stress coupling equation, and the permeability at the fault can then be calculated.

[0083] In this invention, the CO2-assisted oil recovery and storage fluid-structure interaction numerical simulation model can be a model that integrates fluid mechanics and geomechanics to simulate the fault slip state. Of course, the CO2-assisted oil recovery and storage fluid-structure interaction numerical simulation model can also be other types of models, and this invention does not specifically limit them.

[0084] Step S102: Based on the Latin hypercube sampling method, the injection well regime is sampled as input parameters, and calculations are performed based on the CO2 oil displacement and storage fluid-structure interaction numerical simulation model to obtain production and fault slip displacement samples under different injection regimes.

[0085] In some embodiments, Latin Hypercube Sampling (LHS) is a method of approximately random sampling from a multivariate parameter distribution. It belongs to the stratified sampling technique and reduces the correlation between input variables by uniformly sampling in each dimension, thereby improving the accuracy and efficiency of model prediction.

[0086] In this invention, the injection well regime is sampled based on Latin hypercube sampling as input parameters. The CO2 oil displacement and storage fluid-structure interaction numerical simulation model established in step S101 above is used for calculation, and the output results are read, i.e., the samples.

[0087] Step S103: Input the constructed sample into the long short-term memory neural network and train it using the Adam optimization algorithm until the root mean square error meets the preset condition to obtain the target surrogate model.

[0088] In some embodiments, the surrogate model is constructed based on a long short-term memory neural network.

[0089] Step S104: Based on the target proxy model, a multi-objective optimization algorithm is used to optimize the recovery rate objective function, the CO2 sequestration rate objective function, and the fault slip displacement objective function to obtain the corresponding optimization results for each objective function.

[0090] In some embodiments, the multi-objective optimization algorithm may be the NSGA-II algorithm or other multi-objective optimization algorithms. This application does not specifically limit it.

[0091] The method provided by this invention first establishes a numerical simulation model of CO2 flooding and sequestration fluid-structure interaction in high water-cut fault-block reservoirs. Second, it uses the Latin hypercube sampling method to sample and generate samples based on the CO2 flooding and sequestration fluid-structure interaction model. A surrogate model is constructed using a long short-term memory neural network, and the NSGA-II multi-objective optimization algorithm is introduced for multi-objective optimization to obtain the Pareto solution set of the objective function. Thus, this invention, on the one hand, based on a comprehensive framework of a three-dimensional reservoir seepage and geomechanical coupling model, the LSTM artificial intelligence algorithm, and the NSGA-II multi-objective optimization algorithm, can obtain a production regime scheme for injection-production wells through optimization, maximizing oil recovery and CO2 storage while minimizing fault slip displacement.

[0092] In some embodiments, step S101 above may include the following:

[0093] For any stress surface under consideration, the linear relationship between its ultimate strength, shear stress and normal stress is shown in Equation (1);

[0094]

[0095] In the formula, τ n σ is the shear stress on the failure surface. n C represents the normal stress at the failure surface, and C represents the cohesion of the rock. The internal friction angle of the rock.

[0096] In some embodiments, when a point in the material slips (shears) on a plane, the shear stress τ acting on that plane is... n In addition to overcoming the inherent cohesive force C of the material, it is also necessary to overcome the normal stress σ acting on the surface. n The resulting frictional force.

[0097] The stress component acting on the fault plane is the normal stress σ' at the fault surface. n Based on the shear stress σ of the fault plane t and σ s The maximum shear stress τ at the location can be determined. max for

[0098] The sliding displacement of the fracture surface is determined based on the maximum shear stress; the sliding displacement is shown in the following formula:

[0099]

[0100] In the formula, Kt To determine the stiffness of the fault, when the maximum shear stress reaches a certain level, the grids on both sides of the fault will displace, and this movement will affect the sliding friction of the grid blocks.

[0101] According to the slip displacement D s The value of is used to determine the coefficient of friction for sliding friction, as shown in equations (3) to (5):

[0102]

[0103] In the formula, The minimum friction angle is reached when the fault surface begins to slip, at which point the slip friction force decreases; the displacement of the fault from the initial movement to the slip displacement is less than D. smax period, This represents the maximum value of the fault friction force. This represents the minimum value of fault friction.

[0104] The slip tendency of the fault is determined based on the maximum shear stress and the friction coefficient; the slip tendency is shown in the following formula:

[0105]

[0106] In the formula, cohes represents the fault cohesion.

[0107] like Figure 2 As shown, the Mohr circle shifts to the left as pore pressure increases. Material failure occurs when the Mohr circle contacts the failure envelope. The slip tendency of the fault is 1 when the Mohr circle contacts the failure envelope. For engineering safety considerations, a safety factor F less than 1 is introduced in actual reservoirs. s The slip tendency of the fault is greater than F. s During ×ST, the fault is activated.

[0108] In some embodiments, step S102 above may include the following:

[0109] Applying Latin hypercube sampling to the sampling of training sets for long short-term memory neural networks can improve the reliability of numerical simulations and the representativeness of the samples. Assume a structure-function with K basic variables, where N represents the number of sets of x = (x1, x2, ..., x...). n The sample is given by matrix P, which is an N*K matrix. The column elements of P are randomly generated from integers from 1 to N, and each element has a different value. Matrix R is also an N*K matrix, and the column elements of R are randomly generated from rational numbers from 0 to 1, and each element has a different value.

[0110]

[0111]

[0112] In the formula, s ij It is the element in the i-th row and j-th column of matrix S. It is the distribution function that the j-th variable conforms to. It is the j-th variable of the i-th sample obtained through Latin hypercube sampling.

[0113] In some embodiments, step S103 above may include the following:

[0114] The system is trained using a Long Short-Term Memory (LSTM) neural network, with gate weights and bias parameters defined as follows:

[0115]

[0116] Where W represents weight, b represents bias, i represents input gate, f represents forget gate, and o represents output gate.

[0117] like Figure 3 As shown, in a single hidden unit, the first part is the forget gate, which selectively forgets information from the previous time step, allowing the LSTM to retain more important information. In a single hidden unit of a Long Short-Term Memory (LSTM) neural network, the sigmoid activation function is used to determine the forget gate; the forget gate is shown in the following equation:

[0118] f t =σ(W f ·[h t-1 ,x t ]+b f (10),

[0119] In the formula, f t For the forgetting gate, σ represents the activation function, expressed as σ(x) = (1 + e^(-1 / x))^(-1 / x ... -x ) -1 Its output range is 0 to 1, x t h is the input for the current time step. t-1 W is the input from the previous time step. f With b f Weights and biases for the forget gate;

[0120] The input gate update is determined based on the Sigmoid activation function, and at the current step size, it is determined whether the current input has been written to the memory cell based on the input gate; the input gate is shown in the following formula:

[0121]

[0122] In the formula, i t For the input gate, b i This is the bias matrix of the input gate;

[0123]

[0124] In the formula, tanh represents a hyperbolic tangent function with an output range of -1 to 1;

[0125] Update the current cell state based on the historical cell states, as shown in the following formula:

[0126]

[0127] In the formula, C t-1 For historical cell states, C t The current cell state, A vector of new candidate values;

[0128] The output gate controls whether the information stored in the memory cell affects subsequent network layers.

[0129] o t =σ(W o [h t -1,x t ]+b o (14),

[0130] In the formula, o t For output gate, W o With b o These represent the weights and biases of the output gate, h. t This is the output value for the current cell state and also the input value for the next cell state.

[0131] In some embodiments, step S104 above may include the following:

[0132] In CO2 injection development of high water-cut fault-block reservoirs, the well control parameters of injection and production wells ultimately affect the reservoir's recovery rate, storage rate, and geological risk of fault activation. Therefore, the well control operation parameters of injection and production wells are taken as inputs, and the recovery rate f1(x), storage rate f2(x), and fault slip displacement f3(x) are taken as three objective functions as outputs. The problem is transformed into maximizing f1(x) and f2(x) and minimizing f3(x), as defined below:

[0133] maxF(x)=[(f1(x),f2(x),-f3(x)) T (15),

[0134]

[0135]

[0136] f3 = Ds (18),

[0137] In the formula, OOIP represents the initial recoverable reserves of the reservoir; Let be the oil production at time t; The total amount of CO2 injected into the injection well at any given time; D represents the total CO2 produced by the production well at time t, where T is the total time for reservoir development and evaluation, and D is the total CO2 produced by the production well at time t. s This represents fault slip displacement;

[0138] In solving multi-objective optimization problems, the NSGA-II algorithm is adopted, which is evolved from the genetic algorithm. Genetic algorithms generally include three basic operations: selection, crossover, and mutation. NSGA-II improves upon the genetic algorithm by introducing non-dominated sorting algorithms, crowding operators for fitness strategies, and elitist strategies. The algorithm flow is roughly as follows:

[0139] Set the crowding distance of each individual to 0, and let the two individuals at the boundary be d0 and d2. l The crowding degree is positive infinity. All other individuals i = 1, 2, ..., l-1 are allocated using the following formula:

[0140]

[0141] In the formula, f j i+1 Let f be the j-th objective function value at point i+1. j i-1 Let represent the j-th objective function value at point i-1.

[0142] In some embodiments, the complexity of the NSGA-II algorithm depends on the sorting method. The computational complexity is highest, mNlgN, when all individuals are in a non-dominated set.

[0143] In some embodiments, to ensure a more uniform distribution of non-dominated solution sets, the crowding operator is defined based on i d with i rank For two individuals i and j, if i rank <j rank If i > 0, then individual i is better. rank =j rank In this case, it is better to select individuals with lower crowding. Individuals with lower crowding mean that there are relatively fewer individuals around them, and they need to be retained to maintain diversity.

[0144] The NSGA-II algorithm flow is as follows: Figure 4As shown: First, initialize the parent population P0 to generate a child population Q0. Second, merge all individuals in the population and sort them according to their non-dominant relationship. Calculate the crowding degree of individuals in each non-dominant layer. Then, select individuals with better non-dominant order and crowding degree according to the crowding degree comparison operator to form a new parent population P1. Finally, generate the next generation population Q1 through a series of operations including selection, crossover, and mutation. Merge P1 and Q1 into a new population. Repeat the above operations until the program's requirements are met.

[0145] The NSGA-II algorithm is used to optimize the recovery rate objective function, the CO2 sequestration rate objective function, and the fault slip displacement objective function to obtain the corresponding optimization results.

[0146] The following will describe an exemplary application of the embodiments of the present invention in a practical application scenario.

[0147] The purpose of this invention is to provide a joint optimization method for high water-cut reservoirs based on long short-term memory neural networks. Addressing the shortcomings in research on CO2 flooding and storage in high water-cut reservoirs, this invention establishes a fluid-structure interaction numerical simulation model for CO2 flooding and storage in high water-cut fault-block reservoirs using CMG. Samples are generated using Latin hypercube sampling and a simulator. A surrogate model is constructed using a long short-term memory neural network. The NSGA-II multi-objective optimization algorithm is then introduced to perform multi-objective optimization based on the surrogate model, resulting in a Pareto solution set containing three objectives.

[0148] This embodiment relates to the application of a multi-objective optimization method for oil displacement and storage in high water-cut fault-block reservoirs based on long short-term memory neural networks in a specific reservoir as follows:

[0149] Based on the method described in the preceding embodiments, a numerical simulation model coupling flow and geomechanics was established using the CMG-GEM module. The reservoir was set as a dual-porosity, dual-permeability model, as shown in Figure 5. The meshes for geomechanics and flow simulations were identical. Initial conditions for overlying stress were provided to simulate the influence of the overlying rock. The model has 52 × 49 × 1 = 2548 meshes, including 4 injection wells and 9 production wells, with a five-point well configuration. The porosity and permeability distributions of the model are shown in Figure 5(a) and Figure 5(b). Reservoir and fluid parameters are shown in Table 1. During the initial water drive process, to maintain an injection-production ratio of 1:1, the production rate of the production well located at the center of the model was 160 m³ / s. 3 / day, the production well at the boundary center is 80m 3 / day, corner point is 40m 3 / day, injection-production ratio is 1:1. To simulate the development effect of switching from water drive to gas drive at different stages during the development process, this invention uses the model after water drive development as the restart model input into the gas drive. Since the initial contact miscibility pressure is 25MPa and the multi-stage contact miscibility pressure is 18MPa, CO2 injection in this reservoir belongs to CO2 miscibility drive.

[0150] Table 1

[0151]

[0152] In this invention, a production dynamic prediction model based on LSTM is established. First, the model framework needs to be designed. Its optimal framework consists of four hidden layers, one input layer, and one output layer, as shown in the figure. Sequential input is selected, first entering the first LSTM layer. After obtaining the output, it is input to ReLU to extract relevant features. This process is repeated once more with LSTM and ReLU layers before being passed to a fully connected layer for final output. The specific structure of the model is as follows: Figure 6 As shown.

[0153] Figure 7 The results show that the LSTM surrogate model has a good fit coefficient R0 when the water content is 0.7%. 2 The value performed well in the training set, and the retention rate was related to the fault slip displacement R. 2 The value is high, while in the test set, R... 2 All values ​​are greater than 0.7, indicating a good correlation between the actual and predicted values.

[0154] Depend on Figure 8 As shown in Figure 9, with a population size of 100 and 100 iterations, the optimization results of the NSGA-II algorithm are as follows: all solutions are non-dominated, and the distribution of Pareto solutions in three-dimensional space is approximately "L"-shaped. The statistical results of the objective function are shown in Figure 9. It is easy to see from the figure that the distribution pattern of the Pareto optimal concentrated solution is similar for different water cuts, regardless of whether it is for recovery rate, storage rate, or fault slip displacement.

[0155] This invention addresses the shortcomings in CO2 flooding and storage research in high water-cut reservoirs by establishing a fluid-solid numerical simulation model for CO2 flooding and storage in high water-cut fault-block reservoirs using CMG. Samples are generated through Latin hypercube sampling and simulator calculation. A surrogate model is constructed based on the samples using a long short-term memory neural network. The NSGA-II multi-objective optimization algorithm is introduced to perform three-objective optimization on the surrogate model, resulting in a Pareto solution set containing three objectives.

[0156] Figure 10 This is a schematic diagram of the composition structure of the high water-cut reservoir co-optimization device based on long short-term memory neural network provided in an embodiment of the present invention, as shown below. Figure 10As shown, the multi-objective optimization device 1000 for CO2 flooding and storage in high water-cut fault-block oil reservoirs based on long short-term memory neural networks includes: a modeling module 1001 for establishing a fluid-structure interaction numerical simulation model for CO2 flooding and storage in high water-cut fault-block oil reservoirs; a calculation module 1002 for sampling input of the injection well regime of the CO2 flooding and storage model based on the Latin hypercube sampling method, and performing calculations based on the CO2 flooding and storage fluid-structure interaction numerical simulation to obtain samples; a training module 1003 for inputting the established samples into the long short-term memory neural network, training it using the Adam optimization algorithm until the root mean square error meets the preset conditions to obtain a surrogate model; and an optimization module 1004 for optimizing the recovery rate objective function, the CO2 storage rate objective function, and the fault slip displacement objective function based on the target surrogate model using a multi-objective optimization algorithm to obtain the corresponding optimization results.

[0157] In some embodiments, the establishing module 1001 is further configured to determine the linear relationship between the ultimate strength, shear stress, and normal stress of each stressed surface; the linear relationship is shown in the following formula:

[0158]

[0159] In the formula, τ n σ is the shear stress on the failure surface. n C represents the normal stress at the failure surface, and C represents the cohesion of the rock. The internal friction angle of the rock;

[0160] Based on the shear stress σ at the fault plane t and σ s Determine the maximum shear stress τ max The maximum shear stress is shown in the following formula:

[0161]

[0162] The sliding displacement of the fracture surface is determined based on the maximum shear stress; the sliding displacement is shown in the following formula:

[0163]

[0164] In the formula, K t The stiffness of the fault;

[0165] The coefficient of friction for sliding friction is determined based on the value of the sliding displacement; the coefficient of friction is shown in the following formula:

[0166]

[0167]

[0168]

[0169] In the formula, The minimum friction angle, This represents the maximum value of the fault friction force. This represents the minimum value of fault friction.

[0170] The slip tendency of the fault is determined based on the maximum shear stress and the friction coefficient; the slip tendency is shown in the following formula:

[0171]

[0172] In the formula, cohes represents the fault cohesion, and the Mohr circle shifts to the left as the pore pressure increases.

[0173] In some embodiments, the calculation module 1002 is further configured to assume a structure-function with K-dimensional basic variables, where N represents the selection of N groups of x = (x1, x2, ..., x...). n The sample is given by matrix P, which is an N*K matrix. The column elements of P are randomly generated from integers from 1 to N, and each element has a different value. Matrix R is also an N*K matrix, and the column elements of R are randomly generated from rational numbers from 0 to 1, and each element has a different value.

[0174]

[0175]

[0176] In the formula, s ij It is the element in the i-th row and j-th column of matrix S. It is the distribution function that the j-th variable conforms to. It is the j-th variable of the i-th sample obtained through Latin hypercube sampling.

[0177] In some embodiments, the training module 1003 is further configured to train the long short-term memory neural network, defining the weights and bias parameters of the gates; the weights and biases are shown in the following formula:

[0178]

[0179] Where W represents weight, b represents bias, i represents input gate, f represents forget gate, and o represents output gate;

[0180] In a single hidden unit of the Long Short-Term Memory (LSTM) neural network, a sigmoid activation function is used to determine the forget gate; the forget gate is shown in the following equation:

[0181] f t =σ(W f ·[h t-1 ,x t]+b f ),

[0182] In the formula, f t For the forgetting gate, σ represents the activation function, expressed as σ(x) = (1 + e^(-1 / x))^(-1 / x ... -x ) -1 Its output range is 0 to 1, x t h is the input for the current time step. t-1 W is the input from the previous time step. f With b f Weights and biases for the forget gate;

[0183] The input gate update is determined based on the Sigmoid activation function, and at the current step size, it is determined whether the current input has been written to the memory cell based on the input gate; the input gate is shown in the following formula:

[0184] i t =σ(W i ·[h t-1 ,x t ]+b i ),

[0185] In the formula, i t For the input gate, b i This is the bias matrix of the input gate;

[0186]

[0187] In the formula, tanh represents a hyperbolic tangent function with an output range of -1 to 1;

[0188] Update the current cell state based on the historical cell states, as shown in the following formula:

[0189]

[0190] In the formula, C t-1 For historical cell states, C t The current cell state, A vector of new candidate values;

[0191] The output gate controls whether the information stored in the memory cell affects subsequent network layers.

[0192] o t =σ(W o [h t -1,x t ]+b o ),

[0193] In the formula, o t For output gate, W o With bo These represent the weights and biases of the output gate, h. t This is the output value for the current cell state and also the input value for the next cell state.

[0194] In some embodiments, the optimization module 1004 is further configured to use the production system parameters of the injection-production well as independent variables, and the recovery rate, CO2 sequestration rate, and fault slip displacement as dependent variables, to maximize the objective functions of the recovery rate and CO2 sequestration rate and minimize the objective function of the fault slip displacement; the following definitions apply:

[0195] maxF(x)=[(f1(x),f2(x),-f3(x)) T ],

[0196]

[0197] In the formula, OOIP represents the initial recoverable reserves of the reservoir; Let be the oil production at time t; The total amount of CO2 injected into the injection well at time t; D represents the total CO2 produced by the production well at time t, where T is the total time for reservoir development and evaluation, and D is the total CO2 produced by the production well at time t. s This represents fault slip displacement;

[0198] The NSGA-II algorithm is used to optimize the recovery rate objective function, the CO2 sequestration rate objective function, and the fault slip displacement objective function to obtain the corresponding optimization results.

[0199] It should be noted that the description of the apparatus in this embodiment is similar to the description of the method embodiment described above, and has similar beneficial effects to the same method embodiment; therefore, it will not be repeated. For technical details not disclosed in this apparatus embodiment, please refer to the description of the method embodiment of this invention for understanding.

[0200] It should be noted that, in the embodiments of the present invention, if the above-mentioned multi-objective optimization method for CO2 enhanced oil recovery and storage in high water-cut fault-block reservoirs is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of the present invention, or the part that contributes to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a terminal to execute all or part of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), magnetic disks, or optical disks. Thus, the embodiments of the present invention are not limited to any specific hardware and software combination.

[0201] Correspondingly, embodiments of the present invention provide a multi-objective optimization device for CO2 enhanced oil recovery and storage in high water-cut fault-block reservoirs. Figure 11 This is a schematic diagram of the composition structure of the high water-cut reservoir joint optimization equipment provided in an embodiment of the present invention, as shown below. Figure 11 As shown, the high water-cut reservoir co-optimization device 1100 includes at least: a processor 1101 and a computer-readable storage medium 1102 configured to store executable instructions, wherein the processor 1101 typically controls the overall operation of the high water-cut reservoir co-optimization device. The computer-readable storage medium 1102 is configured to store instructions and applications executable by the processor 1101, and can also cache data to be processed or processed by each module in the high water-cut fault-block reservoir CO2 flooding and sequestration multi-objective optimization device 1100, which can be implemented by flash memory or random access memory (RAM).

[0202] This invention provides a storage medium storing executable instructions. When these executable instructions are executed by a processor, they cause the processor to perform the method provided in this invention, for example... Figure 1 The method shown.

[0203] In some embodiments, the storage medium may be a computer-readable storage medium, such as a ferromagnetic random access memory (FRAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), flash memory, magnetic surface memory, optical disc, or a compact disk-read-only memory (CD-ROM); or it may be a device that includes one or any combination of the above-mentioned memories.

[0204] In some embodiments, executable instructions may take the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0205] As an example, executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file containing other programs or data, for example, in one or more scripts within a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple co-located files (e.g., files storing one or more modules, subroutines, or code sections). As an example, executable instructions may be deployed to execute on a single electronic device, or on multiple electronic devices located in one location, or on multiple electronic devices distributed across multiple locations and interconnected via a communication network.

[0206] The above description is merely an embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of the present invention are included within the scope of protection of the present invention.

[0207] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of the invention. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of the invention, the sequence numbers of the above-described processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the invention. The sequence numbers of the above-described embodiments of the invention are merely descriptive and do not represent the superiority or inferiority of the embodiments.

[0208] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. In the several embodiments provided by this invention, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components may be combined, or integrated into another system, or some features may be ignored or not performed.

[0209] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A multi-objective optimization method for CO2 flooding and oil sequestration in high water-cut fault-block reservoirs, characterized in that, The method includes: A fluid-structure interaction numerical simulation model for CO2 flooding and storage in a high water-cut fault-block reservoir was established. The model was based on the Mohr-Coulomb yield criterion to evaluate the fault slip state and was coupled with the reservoir seepage-stress equation to simulate fluid pressure changes, rock mechanical response and fault activation risk during CO2 injection. Based on the Latin hypercube sampling method, the injection well regime is sampled as input parameters, and calculations are performed based on the CO2 oil displacement and storage fluid-structure interaction numerical simulation model to obtain production and fault slip displacement samples under different injection regimes. The sample is input into a long short-term memory neural network and trained using the Adam optimization algorithm until the root mean square error meets the preset condition to obtain the target surrogate model. Based on the target proxy model, a multi-objective optimization algorithm is used to optimize the recovery rate objective function, the CO2 storage rate objective function, and the fault slip displacement objective function to obtain the corresponding optimization results for each objective function.

2. The method according to claim 1, characterized in that, The establishment of a fluid-structure interaction numerical simulation model for CO2 flooding and storage in high water-cut fault-block reservoirs includes: Determine the linear relationship between the ultimate strength, shear stress, and normal stress of each stressed surface; the linear relationship is shown in the following formula; ， In the formula, The shear stress on the failure surface, The normal stress on the failure surface, C The cohesive force of the rock; The internal friction angle of the rock; Based on the shear stress at the fault plane and Determine the maximum shear stress The maximum shear stress is shown in the following formula: ， The sliding displacement of the fracture surface is determined based on the maximum shear stress; the sliding displacement is shown in the following formula: ， In the formula, The stiffness of the fault; The coefficient of friction for sliding friction is determined based on the value of the sliding displacement; the coefficient of friction is shown in the following formula: ， ， ， In the formula, The minimum friction angle, This represents the maximum value of the fault friction force. This represents the minimum value of fault friction. The slip tendency of the fault is determined based on the maximum shear stress and the friction coefficient; the slip tendency is shown in the following formula: ， In the formula, As the pore pressure increases due to the cohesive force within the fault, the Mohr circle shifts to the left.

3. The method according to claim 1, characterized in that, The Latin hypercube sampling method is used to sample the injection well regime as input parameters, and calculations are performed based on the CO2-driven oil recovery and storage fluid-structure interaction numerical simulation model to obtain production and fault slip displacement samples under different injection regimes, including: Suppose a has K Structure-function functions of basic variables, N Indicates selection N Group The sample, matrix P yes The matrix, P The column elements are composed of 1~ N The integers are randomly generated and each element has a different value; the matrix... R Too The matrix, R The column elements are randomly generated from rational numbers between 0 and 1, and each element has a different value; ， ， In the formula, It is a matrix S The Line number j Column elements, It is the first j The distribution function that each variable conforms to. The first was obtained through Latin hypercube sampling. The first group of samples j One variable.

4. The method according to claim 1, characterized in that, The process of inputting the samples into a long short-term memory neural network and training it using the Adam optimization algorithm until the root mean square error meets a preset condition, thereby obtaining the target surrogate model, includes: The system is trained using a Long Short-Term Memory (LSTM) neural network, with gate weights and bias parameters defined as follows: ， in, W Indicates weight, b Indicates bias. Indicates the input gate. f Represents the Gate of Oblivion o Indicates the output gate; In a single hidden unit of the Long Short-Term Memory (LSTM) neural network, a sigmoid activation function is used to determine the forget gate; the forget gate is shown in the following equation: ， In the formula, For the forget gate, σ represents the activation function, and its expression is: Its output range is 0 to 1. Input for the current time step. For the input of the previous time step, and Weights and biases for the forget gate; The input gate update is determined based on the Sigmoid activation function, and at the current step size, it is determined whether the current input has been written to the memory cell based on the input gate; the input gate is shown in the following formula: ， In the formula, For input gate, This is the bias matrix of the input gate; ， In the formula, tanh represents a hyperbolic tangent function with an output range of -1 to 1; Update the current cell state based on the historical cell states, as shown in the following formula: ， In the formula, For historical cell states, The current cell state, A vector of new candidate values; The output gate controls whether the information stored in the memory cell affects subsequent network layers. ， In the formula, For output gate, and These represent the weights and biases of the output gate, respectively. This is the output value for the current cell state and also the input value for the next cell state.

5. The method according to claim 1, characterized in that, Based on the aforementioned proxy model, a multi-objective optimization algorithm is used to optimize the recovery rate objective function, the CO2 sequestration rate objective function, and the fault slip displacement objective function, obtaining the corresponding optimization results for each objective function, including: Using the production system parameters of injection-production wells as independent variables, and recovery rate, CO2 sequestration rate, and fault slip displacement as dependent variables, the following formula is defined to maximize the objective functions of recovery rate and CO2 sequestration rate, and minimize the objective function of fault slip displacement: ， ， ， ， In the formula, OOIP represents the initial recoverable reserves of the reservoir; for t Oil production over a period of time; for t The total amount of CO2 injected into the injection well within a given time period; for t The total amount of CO2 produced by the production well within a given time period. T The total time for reservoir development evaluation. D s This represents fault slip displacement; The NSGA-II algorithm is used to optimize the recovery rate objective function, the CO2 sequestration rate objective function, and the fault slip displacement objective function to obtain the corresponding optimization results.

6. A multi-objective optimization device for CO2 enhanced oil recovery and storage in high water-cut fault-block oil reservoirs based on long short-term memory neural networks, characterized in that, The device includes: A module was established to create a fluid-structure interaction numerical simulation model for CO2 flooding and storage in high water-cut fault-block oil reservoirs. The calculation module is used to sample the injection well regime based on the Latin hypercube sampling method as input parameters, and to perform calculations based on the CO2 oil displacement and storage fluid-structure interaction numerical simulation model to obtain production and fault slip displacement samples under different injection regimes. The training module is used to input the samples into a long short-term memory neural network and train it using the Adam optimization algorithm until the root mean square error meets the preset conditions to obtain the target surrogate model. The optimization module is used to optimize the recovery rate objective function, the CO2 storage rate objective function, and the fault slip displacement objective function based on the target proxy model using a multi-objective optimization algorithm, so as to obtain the corresponding optimization result for each objective function.

7. An electronic device, characterized in that, include: Memory, used to store executable instructions; The processor, when executing executable instructions stored in the memory, implements the multi-objective optimization method for CO2 flooding and sequestration in high water-cut fault-block oil reservoirs as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing executable instructions for causing a processor to execute the executable instructions to implement the multi-objective optimization method for CO2 flooding and sequestration in high water-cut fault-block oil reservoirs as described in any one of claims 1 to 5.

Images