Double-layer optimization method considering well control and profile control and flooding agent parameter setting

Abstract

The invention discloses a double-layer optimization method considering well control and profile control and flooding agent parameter setting, and belongs to the field of oil and gas field development engineering.The method comprises the following steps that an oil reservoir profile control and flooding synergistic optimization control model is built, and dynamic profile control and flooding optimization variables are defined; based on actual reservoir operation constraints, constructing a concentration and dosage constraint judgment method of the profile control and flooding agent; a data-driven optimization algorithm is adopted, and a prediction model is trained and constructed to evaluate the effects of different profile control and flooding schemes; a plurality of injection agent opportunity schemes are generated based on a Gibbs sampling method, and each injection agent scheme is optimized to obtain the optimal injection and production liquid quantity and the optimal injection agent concentration; constructing a proxy model mapping relation; iteratively optimizing the profile control and flooding scheme, performing profile control and flooding simulation real evaluation, and updating the profile control and flooding experience library; and iteratively optimizing and outputting an optimal scheme. By introducing the dynamic profile control and displacement strategy and the advanced optimization algorithm, the design efficiency and effect of the profile control and displacement scheme in the complex oil reservoir environment are improved, and powerful support is provided for efficient development of oil and gas fields.

Description

Technical Field

[0001] The invention relates to the field of oil and gas field development engineering, and in particular to a double-layer optimization method considering well control and drive agent parameter settings. Background Art

[0002] At present, most of the main oil fields in my country adopt water injection development. For example, Daqing and Shengli oil fields have entered the high water content stage in the late stage of water injection development, and gradually formed some large pores with inefficient circulation, which reduced the water drive swept volume, reduced the recovery rate of crude oil, and affected the economic benefits of the oil fields. It is necessary to conduct in-depth research on deep adjustment and flooding technology in the high water content period of the reservoir to improve the reservoir development effect and effectively plug large pores and cracks. Due to the years of water injection development of the reservoir, the optimization of the deep adjustment and flooding system has become more difficult, and there are problems such as failure to inject, failure to plug, and short effective period of measures. Therefore, in order to ensure the large-scale promotion of deep adjustment and flooding, it is of great significance to carry out parameter optimization and adjustment and flooding simulation research of the deep adjustment and flooding system.

[0003] Development decisions usually need to consider many factors at the same time. At present, the optimization of flooding is mostly the optimization of well water injection rate, liquid production rate and flooding agent concentration. There are deficiencies in the optimization of flooding schemes involving slugging measures, resulting in insufficient integrity of flooding optimization schemes. In addition, due to the complexity of the flooding water plugging process and the inherent multi-scale heterogeneity of the reservoir, numerical simulation calculations take a long time. Most of the oilfield production optimization research based on data-driven agent models uses water drive as a means, and there are few optimization studies that add agents to flooding. In addition, due to the multi-dimensionality of flooding variables, the performance differences between a single evolutionary algorithm and an agent strategy in the early and late stages of optimization will affect the overall optimization performance, which can easily lead to incomplete development and fall into a local optimal situation. Summary of the invention

[0004] Aiming at the problem of efficient development and optimization in complex reservoir environments, the present invention proposes a double-layer optimization method that takes into account the setting of well control and displacement agent parameters. This method not only takes into account the fine setting of well control and displacement agent parameters, but also significantly improves the optimization efficiency and accuracy through the introduction of double-layer optimization mechanism and proxy model, especially for determining the best injection timing of the reservoir.

[0005] To achieve the above object, the technical solution of the present invention is implemented as follows: a double-layer optimization method considering well control and displacement agent parameter settings, comprising the following steps:

[0006] S1, constructing a reservoir flooding efficiency optimization control model, setting up reservoir flooding optimization variables, and dividing the optimization variables into upper-layer flooding discrete variables and lower-layer well control continuous variables;

[0007] S2, based on the application conditions of the displacement engineering of the displacement reservoir, a method for determining the concentration and dosage constraints of the displacement agent is established;

[0008] S3, using the flooding effect data set to train and build a machine learning efficient prediction model;

[0009] S4, randomly generate several injection timing schemes based on the Gibbs sampling method, and optimize each injection scheme to obtain its optimal injection and production liquid volume and optimal injection concentration scheme;

[0010] S5, constructing a proxy model mapping relationship, using the upper layer control and displacement discrete variables just calculated and the corresponding optimal lower layer well control continuous variables to construct a proxy model;

[0011] S6, iterative optimization, updating the flooding experience library and outputting the best solution, and continuously repeating step S6 until the preset maximum number of iterations is reached, and finally outputting the best development solution and the flooding balance factor of the best development solution.

[0012] Further, step S1 includes:

[0013] S11, according to the reservoir development requirements, with the goal of maximizing the reservoir adjustment and displacement balance factor, a reservoir adjustment and displacement efficiency optimization control model is constructed, as shown in the following formula:

[0014]

[0015] In the formula, z j is the water content of each displacement well, %; DF(z) represents the reservoir displacement balance factor, decimal; k represents the number of displacement wells, It represents the average water content of all flooding wells, %;

[0016] S12, setting up dynamic drive optimization variables, including upper layer drive discrete variables and lower layer well control continuous variables.

[0017] Further, step S2 includes:

[0018] S21, set the upper and lower limits of the injection-production system and maintain the injection-production balance;

[0019] Step S2 includes:

[0020] S21, set the upper and lower limits of the injection-production system and maintain the injection-production balance;

[0021] In order to ensure the safety and effectiveness of the flooding process, it is necessary to set the maximum and minimum injection rate and the limit of the production rate. At the same time, it is necessary to maintain the balance between injection and production and keep the well control pressure stable.

[0022] S22, setting the upper and lower limits of the concentration of the control and displacement agent, and limiting the total amount of the control and displacement agent;

[0023] According to the physical and chemical properties of the displacement agent and reservoir conditions, set reasonable upper and lower limits of the displacement agent concentration;

[0024] S23, setting specific injection concentration constraint conditions for each injection opportunity.

[0025] Furthermore, step S3 includes: training and constructing an efficient prediction model according to the accumulated adjustment and drive scheme results corresponding to different parameters.

[0026] Further, step S4 includes:

[0027] S41, using the Gibbs sampling method to generate NZ injection timing plans, denoted as in represents the upper layer displacement discrete variable, NZ represents the number of samples;

[0028] S42, for each injection timing scheme, the Gibbs sampling method is used to generate NZ j The combination of injection volume and injection concentration is denoted as in represents the lower well control continuous variable of group j, NZ j It represents the number of samples, and then the concentration and dosage of the displacement agent are judged. If there are individuals that do not meet the constraints, Gibbs sampling is performed again until all individuals meet the constraints and are grouped together;

[0029] S43, set up a parallel calculation of an NZ group, find the corresponding optimal injection and production fluid volume and injection concentration scheme for NZ injection timing schemes, j is the number of groups currently being parallel calculated, and use the reservoir control and displacement simulator to calculate the NZ in the jth group. j The jth group of displacement balance factor data sets are obtained by performing displacement simulation calculations for each scheme to obtain the block displacement balance factor DF in the reservoir displacement efficiency optimization control model. Construct the jth group's control and driving experience database DF j ;

[0030] S44, using NZ in group j j The radial basis function proxy model is constructed by using the schemes and their corresponding block control and displacement balance factors DF.

[0031] Further, step S5 includes: using the upper layer of NZ samples to adjust the discrete variables The corresponding optimal lower well control continuous variable A new radial basis function surrogate model is trained to approximate the mapping relationship between the upper layer solution and the corresponding lower layer optimal solution.

[0032] Further, step S6 includes: using the proxy model constructed in step S5 to adjust the discrete variable individuals of the new upper layer Perform the optimal solution for the corresponding lower layer The prediction of each new upper layer discrete variable individual The best lower layer optimal solution Use the displacement simulation to truly evaluate the objective function and store it in the displacement experience library. Repeat step S6 until the number of iterative optimizations reaches the maximum number of evaluations preset in the reservoir numerical simulator of the displacement mechanism. Record the samples and the objective functions of the samples output by the reservoir numerical simulator of the displacement mechanism during each iterative optimization. After the iterative optimization is completed, output the samples obtained by each iterative optimization and the objective functions corresponding to the samples to obtain the optimal displacement injection-production scheme and the objective functions of the optimal development scheme.

[0033] Compared with the prior art, the double-layer optimization method considering well control and displacement agent parameter setting described in the present invention has the following advantages:

[0034] Aiming at the problem of insufficient injection timing measures in the current reservoir production optimization process, especially in the optimization of the injection scheme, this paper proposes an innovative two-layer adjustment and displacement optimization method. This method not only covers the key decision point of injection timing, but also goes deep into the parameters of specific execution levels such as injection and production fluid volume and injection concentration by finely dividing the optimization variables, thus achieving comprehensive optimization from top to bottom, multi-level and multi-dimensional.

[0035] Specifically, the present invention decomposes the complex optimization problem into two levels: the upper level focuses on the selection of injection timing, and the optimization of this discrete variable directly determines the basic framework of the control and displacement strategy; the lower level makes fine adjustments to the continuous variables for the injection and production fluid volume and injection concentration, ensuring the efficiency and accuracy of the control and displacement measures during the specific implementation process. This hierarchical optimization strategy effectively improves the scientificity and rationality of the formulation of the reservoir development plan, making the final injection timing closer to the actual reservoir, and ensuring the integrity and comprehensiveness of the optimization plan.

[0036] In addition, the present invention cleverly combines the idea of ​​global optimal search with local fine optimization. By introducing the finite difference evolution algorithm, a diverse optimization subpopulation is generated globally, providing a rich candidate solution for finding the potential global optimal solution. At the same time, the use of the prediction model to conduct a preliminary evaluation of these subpopulations greatly accelerates the optimization process and reduces the computational cost. On this basis, combined with the real evaluation of the numerical simulation of the adjustment and drive, the prediction results of the proxy model are verified and updated, further improving the accuracy and reliability of the optimization results.

[0037] The present invention effectively avoids the defect of traditional optimization methods that they are prone to fall into local optimum by continuously iterating the optimization and updating of the adjustment and driving experience library. This dynamic update mechanism enables the optimization process to continuously approach the global optimal solution, and the optimization solution finally obtained is not only technologically advanced, but also shows significant economic and social benefits in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0039] Figure 1 A flow chart of a double-layer optimization method considering well control and displacement agent parameter settings provided by an embodiment of the present invention;

[0040] Figure 2 The optimization curve diagram provided by this embodiment of the present invention;

[0041] Figure 3 The cumulative oil production provided by the best development solution in this embodiment of the present invention;

[0042] Figure 4 The cumulative water production provided by the best development solution in this embodiment of the present invention;

[0043] Figure 5 The cumulative water injection volume provided by the best development solution in this embodiment of the present invention;

[0044] Figure 6 The water content of each production well provided by the best development plan in this embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention, that is, the examples described are only part of the present invention, not all of the examples. The present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification:

[0046] The present invention proposes a two-layer optimization method that considers well control and displacement agent parameter settings, such as Figure 1As shown. In this embodiment, the research block adopts the polymer drive development mode, the experimental model is the reservoir model set by Eclipse, the grid distribution is 25×25×1, the size of each grid is 100×100×20 meters, and 8 water injection wells and 4 drive wells are set in the research block. The eight water injection wells are located at the edge of the block, and the coordinates are (1,1), (13,1), (25,1), (1,13), (25,13), (1,25), (13,25), (25,25), and the four drive wells are located inside the block, and the coordinates are (7,7), (19,7), (7,19), (19,19). The optimal development plan for the research block is determined by the double-layer optimization method considering the well control and drive agent parameter settings proposed in the present invention, which specifically includes the following steps:

[0047] S1, construct the reservoir control model for optimizing oil reservoir displacement and efficiency, set up reservoir control optimization variables, and divide the optimization variables into upper layer control discrete variables and lower layer well control continuous variables; the specific method is as follows:

[0048] S11, construct an optimization control model for reservoir adjustment and flooding efficiency enhancement; construct a multi-dimensional reservoir comprehensive simulation system, and then construct a comprehensive reservoir model, integrating multi-source data such as geological structure, fluid properties and production dynamics, to provide a solid foundation for subsequent optimization.

[0049] According to the needs of reservoir development, with the goal of maximizing the reservoir adjustment and displacement balance factor, an optimization control model for reservoir adjustment and displacement efficiency is constructed, as shown in the following formula:

[0050]

[0051] In the formula, z j is the water content of each displacement well, %; DF(z) represents the reservoir displacement balance factor, decimal; k represents the number of displacement wells, Represents the average water content of all displacement wells, %.

[0052] In this embodiment, the objective function is the reservoir displacement balance factor. Since the pressure of the displacement well is constant at 395 bar in this embodiment, the optimization variables in this embodiment are the injection timing, injection volume (here is the water injection volume) and injection concentration of the water injection well; in this embodiment, the total number of time steps t is set to 10, the time interval of the time step is set to 180 days, and the production cycle is 1800 days. Since there are 8 water injection wells to be optimized in this embodiment, the dimension of the upper layer displacement discrete variable optimized in this embodiment is 10, and the upper layer variable is optimized once in each time step, and the dimension is 10. The two variables of liquid volume and concentration of the 8 water injection wells are optimized in each time step. After optimization, the dimension of the lower layer well control continuous variable is 160. The lower boundary of the upper layer displacement discrete variable injection timing is 0, and the upper boundary is 1; the upper boundary of the water injection volume in the lower layer well control continuous variable is 800sm 3 / day, the lower limit is 600sm 3 / day; the upper limit of injection concentration is 0.1kg / sm 3 , the lower boundary is 0.001kg / sm 3 .

[0053] S12, define the optimization parameter set of dynamic displacement strategy, establish reservoir displacement optimization variables, and divide the optimization variables into upper layer displacement discrete variables and lower layer well control continuous variables to fully cover the key parameters of the displacement strategy.

[0054] S2, according to the actual limiting conditions of reservoir field operation (such as pressure threshold, fluid compatibility, etc.), design the precise control logic of the concentration and dosage of the displacement agent to ensure the safety, economy and efficiency of the displacement process; step S2 describes in detail the application conditions of the displacement engineering and the method for judging the constraint of the displacement agent dosage, and the specific method is as follows:

[0055] Establish reservoir control and displacement optimization variables, which consist of injection timing, injection and production fluid volume, and injection concentration. These variables together determine the effect and efficiency of the control and displacement strategy;

[0056] Categories of optimization variables include:

[0057] Upper-level control and displacement discrete variables: including injection timing, which is a discrete variable that represents the decision to perform control and displacement operations at different time points or stages;

[0058] Lower well control continuous variables: including injection and production fluid volume and injection agent concentration; injection and production fluid volume refers to the amount of liquid injected or produced during the adjustment and displacement process, which is a continuous variable and can be adjusted according to actual needs; injection agent concentration refers to the concentration ratio of the adjustment and displacement agent in the injected liquid, which is also a continuous variable, and its optimization is crucial to improving the adjustment and displacement effect.

[0059] The idea of ​​this displacement optimization method is to find the optimal injection-production system under the optimal injection timing, so the optimization variables are divided into upper displacement discrete variables and lower well control continuous variables. The upper displacement discrete variables are injection timing (discrete variables), and the lower well control continuous variables are injection fluid volume and injection concentration (continuous variables). The injection timing is set as a binary variable, representing whether displacement is performed at each time step, 1 represents yes, and 0 represents no. In this example, because the number of time steps for simulated mining is set to 10, the injection timing variable size is set to 10. The injection fluid volume represents the liquid production and injection volume of each displacement well and water injection well at each time step, which can be continuously valued within a given range. In this example, because only 8 water injection wells are optimized, the variable dimension of the injection fluid volume is set to 8 at each time step. The injection concentration represents the concentration of the displacement agent injected into each water injection well at each time step, which can be continuously valued within a given range. Similarly, the variable dimension of the injection concentration is set to 8 at each time step;

[0060] Based on the application conditions of the displacement engineering of the displacement reservoir, a method for determining the concentration and dosage constraints of the displacement agent is constructed. The specific method is as follows:

[0061] S21, set the upper and lower limits of the injection-production system and maintain the injection-production balance;

[0062] In order to ensure the safety and effectiveness of the flooding process, it is necessary to set the upper and lower limits of the injection-production system, including the maximum and minimum injection rate, as well as the limit of the production rate. At the same time, the injection-production balance must be maintained, that is, the amount of injected liquid and the amount of produced liquid should be balanced overall to avoid excessive pressure fluctuations in the reservoir.

[0063]

[0064] In the formula, I wi,k Indicates the water injection volume of each injection well at each time step; Indicates the lower limit of water injection volume of the injection well; Indicates the upper limit of water injection volume of the injection well; P lj,k Indicates the liquid production of each well at each time step; Indicates the lower limit of fluid production of the drive well; represents the upper limit of the fluid production of the drive well; k represents the time step;

[0065] S22, setting the upper and lower limits of the concentration of the control and displacement agent, and limiting the total amount of the control and displacement agent;

[0066] According to the physical and chemical properties of the displacement agent and the reservoir conditions, reasonable upper and lower limits of the displacement agent concentration are set, which helps to ensure the effective distribution and effect of the displacement agent in the reservoir, while avoiding the negative impact of excessive concentration. In addition, in order to control costs and avoid waste, it is necessary to stipulate the upper limit of the total displacement agent dosage;

[0067]

[0068] In the formula, c wi,k It represents the concentration of injection agent at each time step in each injection well; Indicates the lower limit of injection agent concentration in water injection wells; represents the upper limit of injection well injection agent concentration; C represents the total amount of displacement agent used; k represents the time step;

[0069] S23, for each injection timing, set a specific injection concentration constraint. Set the injection concentration to be non-zero in the time step when the injection timing variable is 1, and set the injection concentration to be zero in the time step when the variable is 0; specifically, when the injection timing variable is 1 (indicating that the adjustment and displacement operation is performed at this time point), the injection concentration cannot be 0 to ensure the effective injection of the adjustment and displacement agent. On the contrary, in the time step when the injection timing variable is 0 (indicating that the adjustment and displacement operation is not performed at this time point), the injection concentration should be set to 0 to avoid unnecessary waste and potential risks. This constraint condition helps to achieve precise control of the adjustment and displacement process and improve the adjustment and displacement effect and economy.

[0070] st c wi,k =T k c wi,k i=1,2,…,z (4)

[0071] In the formula, c wi,k represents the concentration of the injection agent at each time step in each injection well; T k Indicates the timing of injection at each time step;

[0072] S3, build an efficient prediction model, use the previously accumulated data set of adjustment and flooding effects to train and build an efficient prediction model; this model implements a two-layer optimization strategy:

[0073] Upper-level optimization: global search for discrete choices of injection timing;

[0074] Lower-level optimization: For each selected injection timing, a lightweight proxy model is used to replace high-cost numerical simulations to accurately adjust the injection rate and the combination of chemical flooding agents to quickly determine the optimal continuous variable configuration;

[0075] S31, using a pre-prepared data set containing each sample, wherein each sample contains an upper layer displacement discrete variable (injection timing) and a corresponding optimal lower layer well control continuous variable (injection rate and chemical displacement agent ratio), to train a proxy model for quickly and accurately predicting the optimal continuous variable configuration under different injection timings (also called displacement timings);

[0076] S32, for the newly given upper layer control and displacement discrete variable, use the proxy model constructed in step S31 to predict the corresponding optimal lower layer continuous variable at the control and displacement opportunity.

[0077] S4, based on the Gibbs sampling method, generates an initial plan, randomly generates several injection timing plans, and uses the prediction model to optimize each injection plan to obtain the corresponding optimal injection rate and chemical displacement agent ratio plan. The specific method is as follows:

[0078] S41, using the Gibbs sampling method to generate NZ injection timing plans, denoted as in represents the upper layer control and displacement discrete variable, NZ represents the number of samples, and in this example, NZ is set to 20;

[0079] S42, for each injection timing scheme, the Gibbs sampling method is used to generate NZ j The combination of injection volume and injection concentration is denoted as in represents the lower well control continuous variable of group j, NZ j Represents the number of samples. In this example, NZ is set j is 200;

[0080] Then, each combination scheme is classified as a complete group of flooding schemes if it meets all the conditions after the concentration and dosage constraints of the flooding agent are judged. That is, the concentration and dosage constraints of the flooding agent are judged. If there are individuals that do not meet the constraints, Gibbs sampling is performed again until all individuals meet the constraints and are grouped together. For example, NZ j The solution of injection volume and injection concentration: Finally, we get the NZ group solution, which includes NZ×NZ j Initial samples;

[0081] S43, set up a parallel calculation of an NZ group, find the corresponding optimal injection and production fluid volume and injection concentration scheme for NZ injection timing schemes, j is the number of groups currently being parallel calculated, and use the reservoir control and displacement simulator to calculate the NZ in the jth group. j The jth group of displacement balance factor data sets are obtained by performing displacement simulation calculations for each scheme to obtain the block displacement balance factor DF in the reservoir displacement efficiency optimization control model. Construct the jth group's control and driving experience database DF j ;

[0082] S44, using NZ in group j j The radial basis function proxy model is constructed by combining the schemes and their corresponding block control and displacement balance factors DF. The weighted basis function and the objective function of approximating injection and production optimization are used to approximate the mapping relationship between a certain scheme and the corresponding block control and displacement balance factor. The radial basis function formula is as follows:

[0083]

[0084] In the formula, is the predicted value of the radial basis function network model, w i is the weight coefficient of the i-th sample, ||·|| is the Euclidean norm, is the basis function, x i is the injection-production system of the ith sample, x is the decision variable; N is the total number of samples, pieces;

[0085] S45, using finite difference evolution algorithm to DF j Middle and lower well control continuous variables Optimize and get NZ j A new lower well control continuous variable individual is created and constraint processing is performed; the finite difference evolution algorithm strategy is:

[0086] v i =x i +F(x best -x i )+F(x r1 -x r2 ) (6)

[0087] Where: v i represents the decision variable generated by the i-th mutation strategy; x r1 、x r2 are different individuals randomly selected from the population; x best is the individual with the best value of the driving objective function in the current population; F is the mutation operator, which is a constant and F∈(0,2]; x i is the i-th individual in the population. After the mutation is completed, a crossover operation is required. For each dimension of the decision variable j=1,2,...d, a random number rand(0,1) in the interval [0,1] needs to be generated. If the random number is greater than the crossover probability CR, no crossover operation is performed and the value of this dimension remains unchanged. If the random number is less than the crossover probability CR, a crossover operation is required to determine the value u of this dimension. The rules are as follows:

[0088]

[0089] Where CR represents the crossover probability, a decimal; u j Represents the value of the random number of dimension j;

[0090] S46, using the proxy model to analyze NZ j The new lower layer well control continuous variable individuals are evaluated, the corresponding block adjustment and displacement balance factors are quickly predicted, the optimal scheme evaluated is selected, the optimal scheme is simulated and evaluated for real adjustment and displacement and stored in the adjustment and displacement experience database DF j Repeat this step until the preset number of optimizations is reached. In this example, the number of optimizations is set to 200. After reaching the preset number of optimizations, the upper layer of the sample in the jth group is output as the discrete variable. The optimal lower well control continuous variable under

[0091] S47, enter the optimization of the next group, repeat steps S44-S46 until all NZ group samples are optimized. When all NZ group samples are optimized, output the upper layer drive discrete variables of NZ samples. The corresponding optimal lower well control continuous variable

[0092] S5, constructing the proxy model mapping relationship, using the upper layer displacement discrete variables just calculated and the corresponding optimal lower layer well control continuous variables to construct the proxy model, so as to approximate the mapping relationship between the upper layer injection timing and the lower layer optimal solution, the specific method is as follows:

[0093] Upper layer drive discrete variables using NZ samples The corresponding optimal lower well control continuous variable A new radial basis function proxy model is trained to approximate the mapping relationship between the upper layer drive timing discrete variables and the corresponding lower layer optimal solution well control continuous variable optimal solution. The training principle is the same as S44 in S4.

[0094] S6, optimization iteration and real evaluation, finite difference evolution algorithm is used to process the upper layer of flooding discrete variables, to produce evolutionary sub-populations, and the proxy model is used to predict the lower layer optimal variables corresponding to each injection timing in the sub-population, and the real evaluation of flooding simulation is carried out to verify the prediction effect, and the evaluation results are stored in the flooding experience library to continuously optimize the model performance. The specific method is as follows:

[0095] S61, using finite difference evolution algorithm to adjust the upper discrete variables Optimize and obtain NZ new upper layer control and displacement discrete variable individuals The principle of finite difference evolution optimization is the same as S45 in S4;

[0096] S62, using the proxy model constructed in step S5, for each new upper layer discrete variable individual Perform the optimal solution for the corresponding lower layer The prediction of each new upper layer discrete variable individual The best lower layer optimal solution Use flooding simulation to truly evaluate the objective function and store it in the flooding experience library.

[0097] S63, iterative optimization and output of the best solution, continuously repeating step S6 until the preset maximum number of iterations is reached, and finally outputting the best flooding solution, including the best flooding timing, injection rate and chemical flooding agent ratio, and the flooding balance factor of the best development solution verified by the reservoir numerical simulator. Specifically, continue to perform iterative optimization, repeating step S6 until the number of iterative optimizations reaches the preset maximum number of evaluations in the reservoir numerical simulator of the flooding mechanism. In this example, the maximum number of evaluations is set to 1000. During each iterative optimization, the samples output by the reservoir numerical simulator of the flooding mechanism and the objective function of the samples are recorded. After the iterative optimization is completed, the samples obtained by each iterative optimization and the objective function corresponding to the samples are output to obtain the best flooding injection-production solution and the objective function of the best development solution. It should be noted that before the proxy model is obtained, the flooding library can be updated according to the design plan and the reservoir flooding simulator. After the proxy model is obtained, the flooding experience library is updated using the relationship between the proxy model data.

[0098] In order to verify the optimization effect of the method of the present invention, the conventional model optimization method and the method of the present invention are used to determine the optimal injection and production scheme of the study block, that is, the best development scheme of the study block, and the curve of the block's displacement balance factor changing with the number of displacement simulations is output, as shown in FIG. Figure 2 As shown by Figure 2 It can be seen that the method of the present invention can quickly obtain an injection-production scheme with a higher block displacement balance factor, thereby achieving balanced displacement of the remaining oil in the block.

[0099] Figure 3 is the cumulative oil production of the optimal injection-production scheme determined by the method of the present invention and the conventional optimization method, Figure 4 is the cumulative water production of the optimal injection-production scheme determined by the method of the present invention and the conventional optimization method, Figure 5 It is the cumulative water injection volume of the optimal injection-production scheme determined by the method of the present invention and the conventional optimization method. Figure 6 The water content of each production well when the optimal injection-production scheme is determined by the present invention and the conventional optimization method. The cumulative oil production is the revenue item, and the cumulative water production and cumulative water injection are both cost items. Figures 3 to 5 It can be seen that the injection-production scheme determined by the method of the present invention can reduce the investment costs in terms of cumulative water production and cumulative water injection while increasing the cumulative oil production, thus having higher economic benefits.

[0100] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention. The present invention has wide applicability and flexibility. Any person with professional knowledge in this field can make various reasonable adjustments, improvements and derivatives on the basis of this specification after reading and understanding it in depth. These changes may include but are not limited to equivalent substitutions of technology, subtle changes in design, and even comprehensive innovations while keeping the core technical ideas unchanged. All these innovations and changes, as long as they can reflect the core concepts and values ​​pursued by the present invention, should be regarded as effective extensions of the present invention and be protected accordingly.

Claims

1. A two-layer optimization method considering well control and displacement agent parameter settings, characterized in that: The following steps are involved: S1, constructing a reservoir flooding efficiency optimization control model, setting up reservoir flooding optimization variables, and dividing the optimization variables into upper-layer flooding discrete variables and lower-layer well control continuous variables; S2, based on the application conditions of the displacement engineering of the displacement reservoir, a method for determining the concentration and dosage constraints of the displacement agent is established; S3, using the flooding effect data set to train and build a machine learning efficient prediction model; S4, randomly generate several injection timing schemes based on the Gibbs sampling method, and optimize each injection scheme to obtain its optimal injection and production liquid volume and optimal injection concentration scheme; S5, constructing a proxy model mapping relationship, using the upper layer control and displacement discrete variables just calculated and the corresponding optimal lower layer well control continuous variables to construct a proxy model; S6, iterative optimization, updating the flooding experience library and outputting the best solution, and continuously repeating step S6 until the preset maximum number of iterations is reached, and finally outputting the best development solution and the flooding balance factor of the best development solution.

2. A double-layer optimization method considering well control and displacement agent parameter settings according to claim 1, characterized in that: Step S1 includes: S11, according to the reservoir development requirements, with the goal of maximizing the reservoir adjustment and displacement balance factor, a reservoir adjustment and displacement efficiency optimization control model is constructed, as shown in the following formula: In the formula, z j is the water content of each displacement well, %; DF(z) represents the reservoir displacement balance factor, decimal; k represents the number of displacement wells, It represents the average water content of all flooding wells, %; S12, setting up dynamic drive optimization variables, including upper layer drive discrete variables and lower layer well control continuous variables.

3. A double-layer optimization method considering well control and displacement agent parameter settings according to claim 1, characterized in that: Step S2 includes: S21, set the upper and lower limits of the injection-production system and maintain the injection-production balance; In order to ensure the safety and effectiveness of the flooding process, it is necessary to set the maximum and minimum injection rate and the limit of the production rate. At the same time, it is necessary to maintain the balance between injection and production and keep the well control pressure stable. S22, setting the upper and lower limits of the concentration of the control and displacement agent, and limiting the total amount of the control and displacement agent; According to the physical and chemical properties of the displacement agent and reservoir conditions, set reasonable upper and lower limits of the displacement agent concentration; S23, setting specific injection concentration constraint conditions for each injection opportunity.

4. A double-layer optimization method considering well control and displacement agent parameter settings according to claim 3, characterized in that: Step S3 includes: training and constructing an efficient prediction model according to the accumulated results of the control and driving schemes corresponding to different parameters.

5. A double-layer optimization method considering well control and displacement agent parameter settings according to claim 3, characterized in that: Step S4 includes: S41, using the Gibbs sampling method to generate NZ injection timing plans, denoted as in represents the upper layer displacement discrete variable, NZ represents the number of samples; S42, for each injection timing scheme, the Gibbs sampling method is used to generate NZ j The combination of injection volume and injection concentration is denoted as in represents the lower well control continuous variable of group j, NZ j It represents the number of samples, and then the concentration and dosage of the displacement agent are judged. If there are individuals that do not meet the constraints, Gibbs sampling is performed again until all individuals meet the constraints and are grouped together; S43, set up a parallel calculation of an NZ group, find the corresponding optimal injection and production fluid volume and injection concentration scheme for NZ injection timing schemes, j is the number of groups currently being parallel calculated, and use the reservoir control and displacement simulator to calculate the NZ in the jth group. j The jth group of displacement balance factor data sets are obtained by performing displacement simulation calculations for each scheme to obtain the block displacement balance factor DF in the reservoir displacement efficiency optimization control model. Construct the jth group's control and driving experience database DF j ; S44, using NZ in group j j The radial basis function proxy model is constructed by using the schemes and their corresponding block control and displacement balance factors DF.

6. A double-layer optimization method considering well control and displacement agent parameter settings according to claim 5, characterized in that: Step S5 includes: using the upper layer of NZ samples to adjust the discrete variables The corresponding optimal lower well control continuous variable A new radial basis function surrogate model is trained to approximate the mapping relationship between the upper layer solution and the corresponding lower layer optimal solution.

7. A double-layer optimization method considering well control and displacement agent parameter settings according to claim 6, characterized in that: Step S6 includes: using the proxy model constructed in step S5 to adjust the discrete variable individuals of the new upper layer Perform the optimal solution for the corresponding lower layer The prediction of each new upper layer discrete variable individual The best lower layer optimal solution Use the displacement simulation to truly evaluate the objective function and store it in the displacement experience library. Repeat step S6 until the number of iterative optimizations reaches the maximum number of evaluations preset in the reservoir numerical simulator of the displacement mechanism. Record the samples output by the reservoir numerical simulator of the displacement mechanism and the objective function of the samples in each iterative optimization. After the iterative optimization is completed, output the samples obtained in each iterative optimization and the objective function corresponding to the samples to obtain the optimal displacement injection-production scheme and the objective function of the optimal development scheme.

Images