Net present value optimization method for continuous and discrete control of downhole inflow control valve of smart well

Abstract

The application discloses a kind of intelligent well downhole inflow control valve continuous and discrete control net present value optimization method, and specifically relates to oil and gas field development production optimization technical field.The method is by obtaining reservoir geologic model, well pattern and completion segmented information, constructs the ICV control vector of each well each segmented in each control cycle in production life period;With NPV as objective function, according to time step cumulative oil price income and deducts injection medium cost, produced water treatment cost and other expenses, and according to discount rate discount;According to control cycle and segmented structure to generate candidate control vector, input reservoir numerical simulator simulation, obtain oil production rate, water production rate and injection response and calculate NPV;When meeting termination criterion, output optimal ICV control strategy.The method supports bounded variable unconstrained transformation, continuous control time smoothing constraint and discrete gear feasibility and smoothing hard constraint, improves strategy implementability and economic benefit.

Description

Technical Field

[0001] This invention relates to the field of oil and gas field development and production optimization technology, specifically to a net present value optimization method for continuous and discrete control of intelligent well downhole inflow control valves. Background Technology

[0002] To achieve refined injection and production management along different sections of the wellbore, smart well completion technology is widely used. Among them, segmented control methods, represented by intelligent control valves (ICV) or inflow control devices (ICD), can adjust the inflow and outflow distribution of multiple sections of the wellbore based on downhole monitoring information and production response, thereby providing an engineering basis for improving profile utilization, suppressing water channeling (or gas channeling), and enhancing recovery rate.

[0003] However, existing research usually treats continuous and discrete ICVs separately, lacking a systematic comparison of multiple algorithms and multiple types of ICV control methods within a unified framework. Summary of the Invention

[0004] The purpose of this invention is to address the above-mentioned shortcomings by proposing a net present value optimization method for the continuous and discrete control of intelligent well downhole inflow control valves, which integrates continuous opening control and discrete gear control into the same modeling and solution process.

[0005] The present invention specifically adopts the following technical solution:

[0006] A net present value optimization method for continuous and discrete control of downhole inflow control valves in smart wells includes the following steps:

[0007] S1: Obtain reservoir geological model, well network and completion segment information, determine the number of production wells and injection wells, and determine the number of smart control valves, the number of control cycles and the duration of each control cycle for each well. The target control variable is the control vector u composed of the smart control valve settings of each well during the production life.

[0008] S2: The net present value is used as the objective function to evaluate the economic benefits of reservoir development. The net present value is calculated by accumulating over time steps, taking into account oil price revenue, water injection costs, and produced water treatment costs, and a discount rate is introduced to discount the cash flow.

[0009] S3: Based on the preset control cycle division and the intelligent control valve segment structure encoded according to the input format of the reservoir numerical simulator, combined with the optimization algorithm, candidate control vectors containing control variables of each well and each segment in each control cycle are constructed; and time smoothing constraints are introduced in both the continuous algorithm and the discrete algorithm.

[0010] S4: Input the obtained control vector u into the reservoir numerical simulator to obtain the oil production rate / water production rate of the production well and the water injection rate response of the injection well at each time step, and calculate the net present value of the objective function accordingly; when the objective function improvement is less than the threshold and the termination criterion is met, the optimal intelligent control valve control strategy is output to guide the layered injection and production regulation of the intelligent well.

[0011] Preferably, in step S1, each component of the control vector u corresponds to an upper and lower limit for the opening degree of the intelligent control valve. and Therefore, the optimization model established by S1 is a constrained optimization problem. To rewrite it as an unconstrained optimization problem, after defining the control variables in S1, an unconstrained variable x is introduced and the following transformation is performed:

[0012] Let the transformed variable vector be x, and its i-th element be:

[0013] ;

[0014] After logarithmic transformation, the domain of variable x is expanded to The original problem is equivalently rewritten as an unconstrained optimization in the x space, and the solution is performed directly in this space. After obtaining the optimal x, the following inverse transformation is used to map it back to the original variable space, thereby obtaining the control vector u:

[0015] .

[0016] Preferably, the expression for the net present value of the objective function in S2 is:

[0017] ;

[0018] Where J is the optimization objective function, u is the setting of all smart control valves in the well during its production life, n represents the nth time step of the reservoir simulator, and N... t It is the total number of time steps, and the time at the end of the nth time step is determined by... express, It is the size of the nth time step, and b is the annual discount rate. and These refer to the number of production wells and injection wells, respectively. This is the price of oil, in $ / m³ 3 ; , These represent the produced water treatment cost and the injection cost, respectively, in $ / m³. 3 ; and These represent the average oil production rate and average water production rate of the j-th producing well at the n-th time step, respectively, in units of STB / day. This represents the average water injection rate at the k-th injector at the n-th time step. The formula ignores the revenue from hydrocarbon production and the disposal cost of the injected gas.

[0019] Preferably, in S3, when encoding the obtained control vector u according to the segmented control input format of the reservoir numerical simulator CMG, the method for embedding the intelligent control valve into the segmented control of the reservoir numerical simulator CMG is as follows:

[0020] By changing the WI multiplier, the flow rate into the wellbore in the simulator is altered, which is equivalent to changing the opening of the bottom-hole flow control valve. This control method achieves the effect of regulating the flow rate of the fluid entering the wellbore.

[0021] ;

[0022] in, The corresponding opening degree of the intelligent control valve, where 0 corresponds to the intelligent control valve being completely closed, 1 represents the intelligent control valve being completely open, and a value between 0 and 1 corresponds to the intelligent control valve being partially open.

[0023] Preferably, in S3, when using a continuous algorithm for continuous gear control, the method for introducing time smoothing constraints on the control variables during the generation of the time series disturbance of the obtained control vector is as follows:

[0024] The total control vector for the kth iteration is expressed as: ;

[0025] The control subvector of the m-th well is formed by splicing the control sequences of all the smart control valves in that well: ;

[0026] And the mth well The control sequence of an intelligent control valve (ICV) throughout the entire production period is represented as follows: ;

[0027] in To control the number of steps, The number of intelligent control valves (ICVs) in the m-th well, and the gradient vector of the search direction. The order of the components must be consistent with Complete consistency is required to ensure that covariance smoothing works on the correct channels;

[0028] To achieve time smoothing of the ICV gradient in an intelligent control valve, a predefined covariance matrix is ​​introduced during the random disturbance and gradient estimation process. In the case of optimizing only the intelligent control valve ICV, it is usually assumed that different wells and different intelligent control valve ICV channels are independent of each other, that is, no cross-well or cross-valve correlations are introduced, and time correlations are only applied within the time series of the same ICV. Therefore, Construct a block diagonal matrix with the following structure: ;

[0029] Each block Indicates the first The covariance of the j-th ICV in the wellbore across different control steps;

[0030] To smooth the ICV control sequence of the intelligent control valve over time, a spherical covariance model is used to assign correlation to the control steps of the same intelligent control valve, for the control step index. Given time-related length parameters and variance parameter The spherical covariance function is defined as follows:

[0031] ;

[0032] This allows us to obtain the covariance submatrix of each ICV channel. ,when From time to time ,when The covariance is 0, thus controlling the correlation between steps to remain only within a finite time window, achieving time smoothing;

[0033] parameter Determines the smoothing strength: The larger the value, the wider the time correlation range, and the smoother the generated perturbations and gradients; however, An excessively large gradient may result in a small amplitude of the gradient after normalization, thereby reducing the gradient quality and update efficiency. Therefore, it is necessary to select a reasonable value based on the control step length and production dynamics.

[0034] Build Then, the lower triangular matrix is ​​obtained through Cholesky decomposition: ;

[0035] And generate independent standard normal random vectors This forms a smart control valve disturbance control vector with an event-dependent structure: ;

[0036] in The number of perturbation samples, due to Spherical time correlation was introduced into the channel of each intelligent control valve to mitigate disturbances. The temporal variation is continuous / smooth, which makes subsequent correlation gradient estimation tend to produce a smooth search direction. The estimated search direction can then be further multiplied by... Implement additional smoothing filtering.

[0037] Preferably, in S3, when using a discrete algorithm for discrete gear control, a mechanism for handling discrete feasibility and time smoothing hard constraints is adopted during the generation of the obtained control vector u:

[0038] Set a time smoothing hard constraint threshold L, where L = 1 or L is a positive integer not greater than 2, so that any well section satisfies ;

[0039] In generating candidate files Time based on the previous cycle gear Constructing a allowed set and The intersection;

[0040] Candidate gear positions are generated periodically in chronological order: For t=1, an initial gear position distribution or initial value is used for generation. For t≥2, only from Selected from This ensures that the candidate sequence naturally satisfies the hard constraints;

[0041] When the time window When performing local reconstruction, apply a connection constraint to the left boundary of the window. and To satisfy hard constraints, apply a connection constraint to the right boundary of the window. and If the hard constraints are satisfied, backtrack and re-evaluate within the allowed set if the constraints are not satisfied.

[0042] The candidate gear sequence that meets the hard constraints is processed After being mapped to an opening sequence, the simulation evaluation and update process begins.

[0043] The present invention has the following beneficial effects:

[0044] This method reduces the workload of modeling and implementing different valve types separately by incorporating continuous opening control and discrete valve position control into the same modeling and solution process. For bounded control variables, unconstrained transformation can be used to improve the stability of high-dimensional control variable solutions. For continuous valve position control, range constraints and time smoothing constraints are imposed on the control variables to ensure the feasibility and temporal smoothness of the control strategy. For discrete valve position control, discrete feasibility and time smoothing constraints are directly added during the candidate solution generation stage, ensuring that the resulting control sequence meets engineering regulation requirements. This scheme can output an ICV control strategy that can be directly used for field stratified injection and production regulation, and improves the economic indicators of the development. Attached Figure Description

[0045] Figure 1 A flowchart illustrating the net present value optimization method for continuous and discrete control of downhole flow control valves in smart wells;

[0046] Figure 2 Net present value obtained by a continuous optimization algorithm;

[0047] Figure 3 This is the net present value obtained by the discrete optimization algorithm. Detailed Implementation

[0048] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings and specific examples:

[0049] Combination Figure 1 A net present value optimization method for continuous and discrete control of downhole inflow control valves in smart wells includes the following steps:

[0050] S1: Obtain reservoir geological model, well network and completion segment information, determine the number of production wells and injection wells, and determine the number of intelligent control valves (ICVs) for each well, the number of control cycles and the duration of each control cycle. The target control variable is the control vector u composed of the intelligent control valve settings of each well during the production life.

[0051] For ICV (Integrated Capacity Variability), upper and lower bound constraints are introduced to define the feasible region of the optimization problem. Boundary constraints are most common in reservoir production optimization. low with u up Let represent the lower and upper bounds of the control vector u, respectively. Then, considering only the boundary constraints, the optimal control problem can be expressed as: ;

[0052] Methods for rewriting constrained optimization problems into unconstrained optimization problems:

[0053] For the boundary-constrained optimization problem, a logarithmic transformation is performed on each element of the vector u, thus rewriting the constrained optimization problem as an unconstrained optimization problem. Let the transformed variable vector be x, and its i-th element be:

[0054] ;

[0055] After logarithmic transformation, the domain of variable x is expanded to The original problem can be equivalently rewritten as an unconstrained optimization in the x space, and can be solved directly in this space. After obtaining the optimal x, it can be mapped back to the original variable space using the following inverse transformation to obtain the control vector u:

[0056] ;

[0057] S2: The economic benefits of reservoir development are evaluated using net present value (NPV) as the objective function. The NPV is calculated cumulatively over time steps, taking into account oil price revenue, water injection costs, and produced water treatment costs. A discount rate is also introduced to discount the cash flow.

[0058] The expression for the net present value of the objective function is:

[0059] ;

[0060] Where J is the optimization objective function, u is the setting of all smart control valves in the well during its production life, n represents the nth time step of the reservoir simulator, and N... t It is the total number of time steps, and the time at the end of the nth time step is determined by... express, It is the size of the nth time step, and b is the annual discount rate. and These refer to the number of production wells and injection wells, respectively. This is the price of oil, in $ / m³ 3 ; , These represent the produced water treatment cost and the injection cost, respectively, in $ / m³. 3 ; and These represent the average oil production rate and average water production rate of the j-th producing well at the n-th time step, respectively, in units of STB / day. This represents the average water injection rate at the k-th injector at the n-th time step. The formula ignores the revenue from hydrocarbon production and the disposal cost of the injected gas.

[0061] To make the economic evaluation more closely reflect on-site development decisions, the cash flow item of the Net Present Value (NPV) can further include, in addition to crude oil revenue and produced water treatment costs, the multi-media injection costs and related operating costs associated with the injection development process. These multi-media injections include, but are not limited to, the injection of injected water, injected gas, and chemical agents required for chemical flooding / profile control and water shut-off. Corresponding cost items include, but are not limited to, the cost of preparing and injecting injected water, the purchase cost or opportunity cost of injected gas, and the energy consumption costs for compression, pressurization, transportation, and metering; the material costs and preparation and dosage costs of chemical agents such as polymers / surfactants / alkalis; and the surface treatment and operation and maintenance costs related to the above injection methods, wellhead and platform fixed operating costs, well opening and closing costs, and treatment operation costs. The above economic parameters can be set as constants or as functions that vary with time steps to reflect the impact of fluctuations in oil prices and injected media prices, changes in injection ratios, changes in energy consumption, and changes in treatment unit consumption on economic indicators, thereby improving the adaptability and robustness of the objective function to the entire life-cycle development system.

[0062] S3: Based on the pre-defined control cycle division and the intelligent control valve segment structure encoded according to the input format of the reservoir numerical simulator CMG, and combined with the optimization algorithm, candidate control vectors containing the control variables of each well and each segment in each control cycle are constructed. Specifically, to avoid excessively drastic changes in ICV opening between adjacent time steps that would violate valve operating constraints, and to reduce mechanical wear caused by frequent adjustments and thus extend valve life, time smoothing constraints are introduced in both the continuous and discrete algorithms.

[0063] When parameterizing / encoding candidate control vectors according to the segmented control input format of the reservoir numerical simulator (CMG), the method for embedding intelligent control valves into the segmented control of the CMG is as follows:

[0064] Under the assumption of radial seepage, single-phase steady-state flow satisfies Darcy's law. ;

[0065] In simulations, multiphase flow is typically used, at which point the formula becomes: ;

[0066] In the formula, The flow rate of fluid phase j at time step n; The well index is located at layer l. The total fluidity of fluid phase j in the well grid block; The oil pressure of the well; This refers to the bottom hole pressure.

[0067] By changing the WI multiplier, the flow rate into the wellbore in the simulator is altered, which is equivalent to changing the opening of the bottom-hole flow control valve. This control method achieves the effect of regulating the fluid flow rate into the wellbore.

[0068] ;

[0069] in, The corresponding opening degree of the intelligent control valve, where 0 corresponds to the intelligent control valve being completely closed, 1 represents the intelligent control valve being completely open, and a value between 0 and 1 corresponds to the intelligent control valve being partially open.

[0070] Continuous optimization algorithms include at least one or more of the following: ensemble perturbation gradient algorithms, stochastic approximation gradient algorithms, synchronous perturbation stochastic approximation algorithms, or particle swarm optimization algorithms. Discrete optimization algorithms include at least one or more of the following: simulated annealing, genetic algorithms, cross-entropy methods, or large neighborhood search.

[0071] It should be noted that in step S3, when continuous opening control is used, the feasibility and timing smoothness of the control strategy are ensured by imposing range constraints on the control variables and introducing time smoothing constraints.

[0072] In S3, when using a continuous algorithm for continuous gear control, the method for introducing time smoothing constraints on the control variables during the generation of the time series disturbance of the obtained control vector is as follows:

[0073] The total control vector for the kth iteration can be expressed as: ;

[0074] The control subvector of the m-th well is formed by splicing the control sequences of all the smart control valves in that well: ;

[0075] And the mth well The control sequence of an intelligent control valve (ICV) throughout the entire production period is represented as follows: ;

[0076] in To control the number of steps, The number of intelligent control valves (ICVs) in the m-th well, and the gradient vector of the search direction. The order of the components must be consistent with Complete consistency is required to ensure that covariance smoothing works on the correct channels;

[0077] To achieve time smoothing of the ICV gradient in an intelligent control valve, a predefined covariance matrix is ​​introduced during the random disturbance and gradient estimation process. In the case of optimizing only the intelligent control valve ICV, it is usually assumed that different wells and different intelligent control valve ICV channels are independent of each other, that is, no cross-well or cross-valve correlations are introduced, and time correlations are only applied within the time series of the same ICV. Therefore, Construct a block diagonal matrix with the following structure: ;

[0078] Each block Indicates the first The covariance of the j-th ICV in the wellbore across different control steps.

[0079] To smooth the ICV control sequence of the intelligent control valve over time, a spherical covariance model is used to assign correlation to the control steps of the same intelligent control valve, for the control step index. Given time-related length parameters and variance parameter The spherical covariance function is defined as follows:

[0080] ;

[0081] This allows us to obtain the covariance submatrix of each ICV channel. ,when From time to time ,when The covariance is 0, thus controlling the correlation between steps to remain only within a finite time window, achieving time smoothing;

[0082] parameter Determines the smoothing strength: The larger the value, the wider the time correlation range, and the smoother the generated perturbations and gradients; however, An excessively large gradient may result in a small amplitude of the gradient after normalization, thereby reducing the gradient quality and update efficiency. Therefore, it is necessary to select a reasonable value based on the control step length and production dynamics.

[0083] Build Then, the lower triangular matrix is ​​obtained through Cholesky decomposition: ;

[0084] And generate independent standard normal random vectors This forms a smart control valve disturbance control vector with an event-dependent structure: ;

[0085] in The number of perturbation samples, due to Spherical time correlation was introduced into the channel of each intelligent control valve to mitigate disturbances. The temporal variation is continuous / smooth, which makes subsequent correlation gradient estimation tend to produce a smooth search direction. The estimated search direction can then be further multiplied by... Implement additional smoothing filtering.

[0086] Then, in S3, when using a discrete algorithm for discrete gear control, a mechanism for handling discrete feasibility and time smoothing hard constraints is adopted during the generation of the obtained control vector u:

[0087] Set a time smoothing hard constraint threshold L, where L = 1 or L is a positive integer not greater than 2, so that any well section satisfies ;

[0088] In generating candidate files Time based on the previous cycle gear Constructing a allowed set and The intersection;

[0089] Candidate gear positions are generated periodically in chronological order: For t=1, an initial gear position distribution or initial value is used for generation. For t≥2, only from Selected from This ensures that the candidate sequence naturally satisfies the hard constraints;

[0090] When the time window When performing local reconstruction, apply a connection constraint to the left boundary of the window. and To satisfy hard constraints, apply a connection constraint to the right boundary of the window. and If the hard constraints are satisfied, backtrack and re-evaluate within the allowed set if the constraints are not satisfied.

[0091] The candidate gear sequence that meets the hard constraints is processed After being mapped to an opening sequence, the simulation evaluation and update process begins.

[0092] S4: Input the obtained control vector u into the reservoir numerical simulator to obtain the oil production rate / water production rate of the production well and the water injection rate response of the injection well at each time step, and calculate the net present value of the objective function accordingly; when the objective function improvement is less than the threshold and the termination criterion is met, the optimal intelligent control valve control strategy is output to guide the layered injection and production regulation of the intelligent well.

[0093] The S4 termination criteria include, but are not limited to, one or more of the following combinations: the improvement of the objective function NPV in a number of consecutive iterations is less than a preset threshold; the number of iterations reaches a preset maximum value; the computation time reaches a preset upper limit; or the degree of constraint violation of the candidate solution is lower than a preset threshold and the objective function no longer improves significantly.

[0094] Based on the optimal control vector u in step S4 and the response results obtained from the numerical simulator evaluation, a set of predictive indicators for decision comparison can also be output simultaneously, including but not limited to: expected NPV, cumulative oil production, cumulative water production, cumulative injection volume, water cut change trend, and contribution indicators of key wells / key intervals, so as to provide interpretable quantitative basis for intelligent well stratified injection and production control.

[0095] A comparative analysis is conducted on the net present value (NPV) optimization results before and after implementing the NPV optimization method for continuous and discrete control of the intelligent well downhole inflow control valve (ICV) provided by this invention. (See reference...) Figure 2 and Figure 3 As shown, the net present value (NPV) before implementing the optimization method (without optimization) is 0.52 × 10⁻⁶. 8 After implementing the optimization method, the minimum net present value (NPV) is 1.12 × 10⁻⁶. 8 $, with a maximum of 2.10 × 10 8 It can be seen that implementing the net present value optimization method for continuous and discrete control of downhole inflow control valves (ICV) in smart wells provided by this invention improves the economic indicators of oilfield development.

[0096] In summary, the above examples, through numerical simulation, demonstrate the net present value optimization method for continuous and discrete control of the downhole inflow control valve (ICV) in smart wells provided by this invention. This method can output an ICV control strategy that can be directly used for on-site stratified injection and production regulation, and improve the economic indicators of the development.

[0097] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A net present value optimization method for continuous and discrete control of downhole inflow control valves in intelligent wells, characterized in that, Includes the following steps: S1: Obtain reservoir geological model, well network and completion segment information, determine the number of production wells and injection wells, and determine the number of smart control valves, the number of control cycles and the duration of each control cycle for each well. The target control variable is the control vector u composed of the smart control valve settings of each well during the production life. S2: The net present value is used as the objective function to evaluate the economic benefits of reservoir development. The net present value is calculated by accumulating over time steps, taking into account oil price revenue, water injection costs, and produced water treatment costs, and a discount rate is introduced to discount the cash flow. S3: Based on the preset control cycle division and the intelligent control valve segment structure encoded according to the input format of the reservoir numerical simulator, combined with the optimization algorithm, candidate control vectors containing control variables of each well and each segment in each control cycle are constructed; and time smoothing constraints are introduced in both the continuous algorithm and the discrete algorithm. S4: Input the obtained control vector u into the reservoir numerical simulator to obtain the oil production rate / water production rate of the production well and the water injection rate response of the injection well at each time step, and calculate the net present value of the objective function accordingly; when the objective function improvement is less than the threshold and the termination criterion is met, the optimal intelligent control valve control strategy is output to guide the layered injection and production regulation of the intelligent well.

2. The net present value optimization method for continuous and discrete control of intelligent well downhole inflow control valves as described in claim 1, characterized in that, Each component of the control vector u in S1 corresponds to an upper and lower limit for the opening of the intelligent control valve. and Therefore, the optimization model established by S1 is a constrained optimization problem. To rewrite it as an unconstrained optimization problem, after defining the control variables in S1, an unconstrained variable x is introduced and the following transformation is performed: Let the transformed variable vector be x, and its i-th element be: ； After logarithmic transformation, the domain of variable x is expanded to The original problem is equivalently rewritten as an unconstrained optimization in the x space, and the solution is performed directly in this space. After obtaining the optimal x, the following inverse transformation is used to map it back to the original variable space, thereby obtaining the control vector u: 。 3. The net present value optimization method for continuous and discrete control of intelligent well downhole inflow control valves as described in claim 1, characterized in that, The expression for the net present value of the objective function in S2 is: ； Where J is the optimization objective function, u is the setting of all smart control valves in the well during its production life, n represents the nth time step of the reservoir simulator, and N t It is the total number of time steps, and the time at the end of the nth time step is determined by... express, It is the size of the nth time step, and b is the annual discount rate. and These refer to the number of production wells and injection wells, respectively. This is the price of oil, in $ / m³ 3 ; , These represent the produced water treatment cost and the injection cost, respectively, in $ / m³. 3 ; and These represent the average oil production rate and average water production rate of the j-th producing well at the n-th time step, respectively, in units of STB / day. This represents the average water injection rate at the k-th injector at the n-th time step. The formula ignores the revenue from hydrocarbon production and the disposal cost of the injected gas.

4. The net present value optimization method for continuous and discrete control of intelligent well downhole inflow control valves as described in claim 1, characterized in that, In S3, when encoding the obtained control vector u according to the segmented control input format of the reservoir numerical simulator CMG, the method for embedding the intelligent control valve into the segmented control of the reservoir numerical simulator CMG is as follows: By changing the WI multiplier, the flow rate into the wellbore in the simulator is altered, which is equivalent to changing the opening of the bottom-hole flow control valve. This control method achieves the effect of regulating the flow rate of the fluid entering the wellbore. ； in, The corresponding opening degree of the intelligent control valve, where 0 corresponds to the intelligent control valve being completely closed, 1 represents the intelligent control valve being completely open, and a value between 0 and 1 corresponds to the intelligent control valve being partially open.

5. The net present value optimization method for continuous and discrete control of intelligent well downhole inflow control valves as described in claim 1, characterized in that, In S3, when using a continuous algorithm for continuous gear control, the method for introducing time smoothing constraints on the control variables during the generation of the time series disturbance of the obtained control vector is as follows: The total control vector for the kth iteration is expressed as: ; The control subvector of the m-th well is formed by splicing the control sequences of all the smart control valves in that well: ; in Let I represent the number of intelligent control valves (ICVs) in the m-th well, and let I represent the number of intelligent control valves (ICVs) in the m-th well. The control sequence of an intelligent control valve (ICV) throughout the entire production period is represented as follows: ; in To control the number of steps, the gradient vector of the search direction is... The order of the components must be consistent with Complete consistency is required to ensure that covariance smoothing works on the correct channels; To achieve time smoothing of the ICV gradient in an intelligent control valve, a predefined covariance matrix is ​​introduced during the random disturbance and gradient estimation process. , Construct a block diagonal matrix with the following structure: ; Each block Indicates the first The covariance of the j-th ICV in the wellbore across different control steps; To smooth the ICV control sequence of the intelligent control valve over time, a spherical covariance model is used to assign correlation to the control steps of the same intelligent control valve, for the control step index. Given time-related length parameters and variance parameter The spherical covariance function is defined as follows: ； This yields the covariance submatrix for each ICV channel. ,when From time to time ,when The covariance is 0, thus controlling the correlation between steps to remain only within a finite time window, achieving time smoothing; parameter Determines the smoothing strength: The larger the value, the wider the time correlation range, and the smoother the generated perturbations and gradients; however, An excessively large gradient will result in a small amplitude of the gradient after normalization, thereby reducing the gradient quality and update efficiency. Therefore, it is necessary to select a reasonable gradient based on the control step length and production dynamics. Build Then, the lower triangular matrix is ​​obtained through Cholesky decomposition: ; And generate independent standard normal random vectors This forms a smart control valve disturbance control vector with an event-dependent structure: ; in The number of perturbation samples, due to Spherical time correlation was introduced into the channel of each intelligent control valve to mitigate disturbances. The temporal variation is continuous / smooth, causing subsequent correlation gradient estimation to tend to produce a smooth search direction. Further multiplying the estimated search direction by... Implement additional smoothing filtering.

6. The net present value optimization method for continuous and discrete control of intelligent well downhole inflow control valves as described in claim 1, characterized in that, In S3, when using a discrete algorithm for discrete gear control, a mechanism for handling discrete feasibility and time smoothing hard constraints is adopted during the generation of the obtained control vector u: Set a time smoothing hard constraint threshold L, where L = 1 or L is a positive integer not greater than 2, so that any well section satisfies ; In generating candidate files Time based on the previous cycle gear Constructing a allowed set and The intersection; Candidate gear positions are generated periodically in chronological order: For t=1, an initial gear position distribution or initial value is used for generation. For t≥2, only from Selected from This ensures that the candidate sequence naturally satisfies the hard constraints; When the time window When performing local reconstruction, apply a connection constraint to the left boundary of the window. and To satisfy hard constraints, apply a connection constraint to the right boundary of the window. and If the hard constraints are satisfied, backtrack and re-evaluate within the allowed set if the constraints are not satisfied. The candidate gear sequence that meets the hard constraints is processed After being mapped to an opening sequence, the simulation evaluation and update process begins.

Images