Injection-production regulation method, device and equipment

Abstract

The embodiment of the specification provides an injection-production regulation method, device and equipment, relates to the technical field of oil and gas development, and comprises the following steps: inputting first geological development data of multiple wells in an oil reservoir into an injection-production regulation model to obtain a regulation strategy of each well; dividing the multiple wells into multiple well groups; selecting a target well group with the highest regulation priority from multiple unoptimized well groups; iteratively performing optimization and merging operations on the target well group until an optimization termination condition is met; wherein the optimization and merging operations comprise the following steps: optimizing the regulation strategy of each well in the target well group by using the injection-production regulation model; merging the optimized target well group and a well group with the highest regulation priority in an adjacent well group to form a new target well group, wherein the adjacent well group is a well group that has not been optimized and shares at least one production well with the target well group; iteratively performing the step of selecting the target well group until a selection termination condition is met; and regulating injection-production parameters of corresponding wells according to the optimized regulation strategy of each well in the oil reservoir.

Description

Technical Field

[0001] The embodiments in this specification relate to the field of oil and gas development technology, specifically to an injection-production control method, apparatus, and equipment. Background Technology

[0002] As oil and gas field development enters its mid-to-late stages, increased formation heterogeneity, complex well network relationships, and prominent injection-production conflicts make traditional static injection-production schemes unable to continuously meet production demands. In recent years, with the development of digital and smart oilfield technologies, optimization methods based on numerical simulation and data-driven approaches have been increasingly applied to the control of injection-production systems. These methods can improve injection-production matching efficiency and capacity utilization to a certain extent.

[0003] However, the above methods are based on single production data and rely on a global optimization framework, attempting to obtain the optimal solution at the overall well network scale in one go. However, the reservoir system exhibits high nonlinearity and uncertainty. The global optimization lacks a real-time adaptive adjustment mechanism when injection and production conditions and reservoir status change, which can easily lead to instability, difficulty in convergence, or slow convergence speed. It cannot meet the control requirements in a dynamic production environment, resulting in low injection and production control efficiency. Summary of the Invention

[0004] The purpose of the embodiments in this specification is to provide an injection-progression control method, apparatus, and equipment to overcome the problems of existing methods being unable to meet the control requirements in dynamic production environments and having low injection-progression control efficiency.

[0005] To solve the above-mentioned technical problems, the specific technical solutions of the embodiments in this specification are as follows:

[0006] On the one hand, the embodiments of this specification provide an injection-propagation control method, including:

[0007] The first geological development data of multiple wells in the reservoir are input into the injection-production control model to obtain the control strategy for each well;

[0008] The multiple wells are divided into multiple well groups, each well group including an injection well and multiple adjacent production wells;

[0009] Select the target well group with the highest control priority from multiple well groups that have never been optimized;

[0010] The optimization and merging operations of the target well group are performed iteratively until the optimization termination condition is met. This includes: using the injection-production control model to optimize the control strategy of each well in the target well group; merging the optimized target well group with the well group with the highest control priority in its adjacent well groups to form a new target well group. The adjacent well groups are those that have not been optimized and share at least one production well with the target well group.

[0011] The above steps for selecting the target well group are executed iteratively until the selection termination condition is met;

[0012] Based on the optimized control strategy for each well in the reservoir, the injection and production parameters of the corresponding wells are adjusted.

[0013] Furthermore, the method also includes:

[0014] Acquire geological attribute data, production dynamic data, inter-well connectivity data, and engineering attribute data from multiple wells in the reservoir;

[0015] The geological attribute data, production dynamic data, well connectivity data, and engineering attribute data are spliced ​​together;

[0016] The vectorized and stitched data yields the first geological development data for multiple wells.

[0017] Furthermore, determining the control strategy for each well based on the first injection-production control data of each well includes:

[0018] Based on the first injection-production control data, multiple third injection-production control data are generated using multiple expert models;

[0019] Based on the injection-production control model and the confidence levels of each expert model, the first injection-production control data and multiple third injection-production control data are weighted and fused to obtain the fourth injection-production control data for each well.

[0020] Based on the fourth injection-production control data of each well, the control strategy for each well is determined.

[0021] Furthermore, dividing the multiple wells into multiple well groups includes:

[0022] Based on a preset spatial window, the multiple wells are divided into multiple well groups with the injection well as the spatial center.

[0023] Furthermore, the method also includes:

[0024] Calculate the total oil production data of multiple production wells in each well group;

[0025] Based on the total oil production data of each well group, the control priority of each well group is determined, and the control priority of each well group is positively correlated with the total oil production data of that well group.

[0026] Furthermore, the optimization of the control strategy for each well within the target well group using the injection-production control model includes:

[0027] Using the control strategies of each well in multiple unoptimized well groups as constraints, the injection-production control model is used to optimize the control strategies of each well in the target well group.

[0028] Furthermore, the optimization of the control strategies of each well in the target well group using an injection-production control model, with the control strategies of each well in the unoptimized well groups as constraints, includes:

[0029] Obtain the second geological development data of multiple wells optimized using the aforementioned control strategy;

[0030] Using the control strategies of each well in multiple unoptimized well groups as constraints, the second geological development data is input into the injection-production control model to maximize the total oil production of multiple production wells in the target well group, thereby obtaining the optimized control strategies of multiple wells in the target well group.

[0031] Furthermore, the optimization termination conditions include: the target well group has no adjacent well groups, or the number of injection wells contained in the target well group reaches a preset upper limit;

[0032] The iterative optimization and merging operation on the target well group until the optimization termination condition is met includes:

[0033] Determine whether the optimized target well group meets the optimization termination conditions;

[0034] If the conditions are met, the optimization of the target well group is terminated.

[0035] If not, based on the optimized control strategy of each well in the target well group and the control strategy of each well in the adjacent well group with the highest control priority of the target well group, a new control strategy for each well in the new target well group is determined; the optimization and merging steps are executed iteratively until the optimization termination condition is met.

[0036] Furthermore, the termination condition includes: there are no unoptimized well groups.

[0037] Furthermore, the injection-production control model is constructed in the following manner:

[0038] Obtain third-party geological development data and corresponding injection-production tag data from multiple wells in the reservoir;

[0039] The injection-production control model is trained based on the third geological development data and the corresponding injection-production tag data.

[0040] Furthermore, the method also includes:

[0041] Obtain the fourth geological development data and corresponding injection-production tag data of the multiple wells after being regulated using the target regulation strategy;

[0042] Based on the fourth geological development data and the corresponding injection-production tag data, the injection-production control model is fine-tuned.

[0043] On another front, embodiments of this specification provide an injection-production control device, comprising:

[0044] The input module is used to input the first geological development data of multiple wells in the reservoir into the injection-production control model to obtain the control strategy for each well;

[0045] A partitioning module is used to divide the multiple wells into multiple well groups, each well group including an injection well and multiple adjacent production wells;

[0046] The selection module is used to select the target well group with the highest control priority from multiple well groups that have not been optimized;

[0047] The optimization module is used to iteratively perform optimization and merging operations on the target well group until the optimization termination condition is met. This includes: using the injection-production control model to optimize the control strategy of each well in the target well group; merging the optimized target well group with the well group with the highest control priority in its adjacent well groups to form a new target well group, where the adjacent well groups are those that have not been optimized and share at least one production well with the target well group.

[0048] The iteration module is used to iteratively execute the above steps for selecting the target well group until the selection termination condition is met;

[0049] The control module is used to adjust the injection and production parameters of each well in the reservoir according to the optimized control strategy.

[0050] In another aspect, a computer device is provided, including a memory for storing computer programs and a processor for executing the computer programs to implement the above-mentioned injection and extraction control method.

[0051] Furthermore, embodiments of this specification also provide a readable storage medium storing a computer program that, when executed by a processor, implements the method described in any of the foregoing descriptions.

[0052] Furthermore, embodiments of this specification also provide a computer program product, which, when run by the processor of a computer device, executes instructions for any of the methods described above.

[0053] As can be seen from the technical solutions provided in the embodiments of this specification above, the embodiments of this specification can input the first geological development data of multiple wells in an oil reservoir into an injection-production control model to obtain the control strategy for each well; divide the multiple wells into multiple well groups, each well group including an injection well and multiple adjacent production wells; select the target well group with the highest control priority from the multiple unoptimized well groups; iteratively perform optimization and merging operations on the target well group until the optimization termination condition is met; including: using the injection-production control model to optimize the control strategy of each well in the target well group; merging the optimized target well group with the well group with the highest control priority in its adjacent well groups to form a new target well group, wherein the adjacent well groups are unoptimized well groups that share at least one production well with the target well group; iteratively execute the above steps of selecting the target well group until the selection termination condition is met; and adjust the injection-production parameters of the corresponding wells according to the optimized control strategy of each well in the oil reservoir. Through the divide-and-conquer strategy, the complex global optimization task is decomposed into a series of sequentially executed, controllable-scale local optimization problems. This incremental approach of well group optimization -> merging -> re-optimization effectively avoids the technical difficulties, computational oscillations, or convergence problems caused by excessive dependent variables and constraints in one-time global optimization. Using the control strategies of each well in the unoptimized well group as constraints ensures that the rest of the system remains stable while optimizing the current local area, preventing strategy oscillations. Each local optimization is performed under defined boundary conditions, making the evolution of the entire system smooth and controllable, and the optimization results can be effectively solidified and transferred to the next step. Furthermore, adjacent well groups sharing at least one identical producing well with the target well group ensures that the expansion of the optimization scope follows the real dynamic interference relationships between wells. The optimization path constructed by this method is not a mathematical abstraction but closely follows the natural connectivity of underground fluids, giving the final target control strategy a solid reservoir physics foundation and strong engineering feasibility. Overall, through multiple optimization and merging iterations, the quality of the strategy can be continuously corrected and improved. The final control strategy is the result of multiple rounds of collaborative consideration from local to global perspectives, possessing both a global vision and engineering feasibility. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the accompanying drawings used in the description of the embodiments or prior art will be briefly introduced below.

[0055] Figure 1 This is a schematic diagram of an injection and extraction system provided in the embodiments of this specification;

[0056] Figure 2 This is a flowchart of an injection-propulsion control method provided in the embodiments of this specification;

[0057] Figure 3This is a flowchart illustrating the overall logic of an injection-propulsion control method provided in the embodiments of this specification;

[0058] Figure 4 This is a schematic diagram of well combination in an injection-production control method provided in the embodiments of this specification;

[0059] Figure 5 This is a schematic diagram illustrating a method for controlling injection and production provided in the embodiments of this specification, in which the current well and the target well group share a production well;

[0060] Figure 6 This is a schematic diagram of the structural composition of an injection-production control device provided in the embodiments of this specification;

[0061] Figure 7 This is a schematic diagram of the structural composition of the computer device provided in the embodiments of this specification. Detailed Implementation

[0062] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0063] It should be noted that the terms "first," "second," etc., used in this specification, claims, and the foregoing drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0064] Figure 1 This specification provides an example of an injection and collection system, which may include:

[0065] Automatic Data Integration Module 1: This module is responsible for the automatic collection, cleaning, and standardization of heterogeneous data from multiple sources, including reservoirs, geology, production, and equipment. Through a unified interface and data format conversion mechanism, it achieves seamless integration of production data and model data. Ultimately, this forms a complete, accurate, and usable unified data foundation, providing support for subsequent intelligent analysis and decision-making.

[0066] Injection-Production Control Strategy Intelligent Generation Module 2: Based on integrated data and artificial intelligence models, it provides intelligent control strategy analysis functions, automatically analyzes the dynamic relationship and response rules between oil wells and water injection wells, and provides input data for setting injection-production control methods and control parameters.

[0067] Injection-Production Control Parameter Adaptive Adjustment Module 3: This module takes real-time production data and system status to form characteristic parameters, inputs them into the aforementioned intelligent injection-production control strategy generation module, predicts the injection-production control strategy, and dynamically corrects key parameters of the injection-production system (such as injection volume, injection pressure, etc.) based on the obtained injection-production control strategy. Furthermore, it uses a multi-level adaptive algorithm to ensure that the system can maintain the optimal estimate of injection-production system control parameters even under complex well group interference conditions.

[0068] Module 4, the automatic injection-production adjustment plan generation module, transforms intelligently generated strategies and parameters into executable production operation plans, automatically generating corresponding adjustment instructions. The plan includes water injection well allocation, oil well production control, and operational procedure descriptions. This achieves seamless integration from decision-making to execution, improving the efficiency and accuracy of injection-production adjustment work.

[0069] Well Workover Plan Development Module 5: Based on wellbore conditions, completion information, and production dynamics, this module intelligently analyzes potential wellbore problems and optimization needs. It automatically outputs workover objectives, operational procedures, and construction parameters to form a complete workover plan, providing a basis for on-site construction.

[0070] Remote Control Module 6: Enables remote real-time monitoring and control of the entire injection and production system, supporting unified management across regions and multiple well groups. This module can read and manipulate field parameters via the network, significantly improving the flexibility of control. It allows for rapid response and adjustment in case of emergencies, reducing the cost of manual intervention.

[0071] Remote Valve Control Module 7: Supports remote on / off control and flow regulation of valves in wellhead and pipeline systems, ensuring the accuracy and timeliness of system operation. It can automatically issue valve adjustment commands according to the control strategy, achieving real-time closed-loop control. This improves operational safety and reduces the risks associated with on-site manual operation.

[0072] Remote Variable Frequency Drive Control Module 8: Enables remote start / stop and frequency adjustment of variable frequency drives for equipment such as electric submersible pumps and water injection pumps. This module can automatically match the drive output power according to production parameters, optimizing energy consumption and production capacity.

[0073] Sensor Module 9: Responsible for real-time acquisition of key parameters such as pressure, flow rate, temperature, and liquid level in the wellbore, formation, and surface facilities. Sensor data is transmitted wirelessly or via wired connection to the aforementioned automatic data integration module, providing real-time information for intelligent analysis and control.

[0074] Corresponding to the above-described injection and production system, this specification provides an injection and production control method in its embodiments. Figure 2 This is a flowchart of an injection-progression control method provided in the embodiments of this specification. Figure 3 This is a flowchart illustrating the overall logic of an injection-propagation control method provided in the embodiments of this specification. In specific implementation, it includes the following steps:

[0075] S10: Input the first geological development data of multiple wells in the reservoir into the injection-production control model to obtain the control strategy for each well.

[0076] In some embodiments, the reservoir contains multiple wells, including multiple injection wells and multiple production wells. The injection wells inject displacement media into the formation, and the production wells produce crude oil. These wells are spatially distributed within the reservoir and collectively constitute the reservoir's development well network. Specifically, the injection wells inject displacement media such as water, gas, or chemicals into the formation to maintain formation energy and drive crude oil flow. The production wells produce crude oil displaced by the displacement media to the area surrounding the wellbore. These wells are deployed according to the reservoir's geological structure and fluid distribution characteristics, forming a regular or irregular well network pattern in space to ensure effective control of the reservoir's affected area.

[0077] In some embodiments, the aforementioned first geological development data may be obtained by: acquiring geological attribute data, production dynamic data, inter-well connectivity data, and engineering attribute data of multiple wells in the reservoir; splicing the geological attribute data, production dynamic data, inter-well connectivity data, and engineering attribute data; and vectorizing the spliced ​​data to obtain the first geological development data of the multiple wells.

[0078] Multi-source heterogeneous data covering geological, production, connectivity, and engineering dimensions can be obtained from oilfield databases and real-time monitoring systems. Geological attribute data may include, but is not limited to, static parameters such as formation depth, permeability distribution, porosity, effective thickness, and original formation pressure within the control area of ​​each well, used to characterize the physical properties and fluid storage capacity of the reservoir. Production dynamic data can cover time-varying sequence data such as oil production, water production, water injection, wellhead pressure, and bottomhole flowing pressure, reflecting the real-time production status and historical trends of the well. Inter-well connectivity data can include inter-well influence coefficients and flow distribution relationships obtained through well test interpretation, interference testing, or dynamic analysis, quantifying the dynamic coupling effects within the well group. Engineering attribute data can include engineering information such as completion structure, perforation location, artificial lift method, and downhole tool status, describing the well completion and operational context.

[0079] In some embodiments, the four types of multimodal data mentioned above are preprocessed using reservoirs as sample units and wells within the reservoirs as the order. After standardization and normalization, they are then multidimensionally spliced ​​in the order of geology-dynamics-connectivity-engineering to form a wide feature vector that comprehensively reflects the state of the wellbore-formation system. Subsequently, this wide vector is dimensionality-reduced and vectorized through an embedded layer or a fully connected network to generate a unified feature representation with high information density, i.e., the first geological development data. This process maps heterogeneous data to a unified vector space, which can provide standardized input for subsequent injection-production control models.

[0080] By constructing a unified feature vector, the barriers between heterogeneous data such as geological, dynamic, and engineering data are resolved, enabling subsequent injection-production control models to make decisions based on comprehensive and consistent information, significantly improving the completeness of feature expression and the reliability of decision-making basis.

[0081] In some embodiments, step S10 may specifically include: inputting the first geological development data into the injection-production control model to obtain the first injection-production control characteristic data for each well. The first injection-production control characteristic data includes multiple items among the control type, pressure response, production change rate, water cut increase rate, inter-well interference coefficient, and energy consumption coefficient of the well. Based on the first injection-production control data of each well, a control strategy for each well is determined. The control strategy includes at least the control method of the well and multiple corresponding control parameters.

[0082] The obtained initial geological development data can be input into a pre-trained injection-production control model. This model can be a deep neural network, which calculates through forward propagation and outputs the initial injection-production control feature data for each well. This initial injection-production control feature data can be a multi-dimensional vector including multiple indicators.

[0083] The control type can be a classification label indicating the recommended macro-level operational direction. Pressure response can include a quantitative prediction of the magnitude and direction of formation pressure changes after injection-production adjustments. Production change rate can include the predicted percentage change in oil production after adjustments. Water cut increase rate can include the estimated trend and rate of change in water cut in the produced fluid after adjustments. Inter-well interference coefficient can quantify the dynamic impact of adjustments in this well on the production status of adjacent wells. Energy consumption coefficient can assess the unit energy cost of increasing oil production involved in implementing this control strategy. By using interpretable control indicators with physical meaning, such as pressure response and interference coefficients, which have clear reservoir engineering implications, rather than black-box results, the transparency of the decision-making process and the interpretability of the results are greatly enhanced.

[0084] Based on the characteristic data output by the injection-production control model, a control strategy can be automatically generated for each well. This strategy can be a structured decision tuple, including the control method for each well and a set of multiple control parameters. The control method can be a concrete operational instruction determined based on the control type. The control parameter set can be a list of parameters, including but not limited to specific control targets such as target oil production, water injection volume, and flowing pressure control values. The initial values ​​of these parameters can be directly mapped from numerical indicators such as pressure response and production change rate predicted by the model after inverse normalization. This initial control strategy, including the control method and control parameters, provides a precise and quantitative starting point for subsequent well group collaborative optimization, ensuring that the entire control process is built on a solid foundation of data-driven and intelligent prediction from the initial stage.

[0085] In some embodiments, based on the feature data output by the injection-propulsion control model Obtain from it For each well Regulation type Pressure response Production change rate Water content rise rate Inter-well interference coefficient Energy consumption coefficient Key parameters, etc. Set as the current control characteristic parameter Based on Set the control method for each well in the reservoir. and the corresponding control parameters ,in Adjustment amount for oil production, This is the adjustment amount for the water injection volume. This is the bottom hole flowing pressure adjustment amount, which is consistent for both production wells and injection wells. This is for information on the perforation layer, where the absolute value represents the number of layers to be added. Positive numbers indicate perforation filling, and negative numbers indicate perforation plugging, arranged sequentially. The number of actual control parameters... Selection can be based on on-site needs and priorities. For each well... At the current time step The corresponding adjustment methods and parameter sets for the next step are as follows: Based on the type of regulation right Wellheads are given control methods .

[0086] In some embodiments, for production wells, control methods may include no control, increasing oil production, controlling water cut (equivalent to reducing oil production), controlling bottomhole flowing pressure, increasing perforation levels, plugging perforation levels, and converting to water injection wells. Increasing oil production and controlling water cut are for renewal... The method for controlling bottom hole flowing pressure is to update Adding perforation layers and sealing perforation layers are for updating. The injection well is for renewal. and clear The settings are as follows: For both production and injection wells, when the control mode is uncontrolled, the current control parameters do not need to be changed. After the control mode changes, only the corresponding parameters are updated. Except for control modes for wells switching from water injection to production, other control parameters are not cleared. That is, the constraint parameters formed during the change of control mode are retained, thereby achieving adaptive adjustment of the control mode and constraint parameters of a single well.

[0087] In some embodiments, for injection wells, control methods may include no control, increasing injection volume, decreasing injection volume, increasing perforation levels, sealing perforation levels, or converting to a production well. Increasing and decreasing injection volume are updates. Adding perforation layers and sealing perforation layers are for updating Converting production wells to renewal wells and clear The settings.

[0088] S20: Divide the multiple wells into multiple well groups, each well group including an injection well and multiple adjacent production wells.

[0089] In some embodiments, step S20 may specifically include: dividing the multiple wells into multiple well groups based on a preset spatial window, each well group including an injection well and multiple adjacent production wells.

[0090] It can identify the well type of all wells in the reservoir, classifying them into injection wells and production wells. Subsequently, it can calculate the Euclidean distance between all well pairs in the reservoir and use the global minimum well spacing, that is, the shortest distance between any two wells, as the spatial reference.

[0091] Based on the calculated global minimum well spacing, a preset spatial window can be constructed. This spatial window can be a circular area centered on each injection well, and its radius, i.e., the radius of influence, can be set to A times the global minimum well spacing, where A is a positive number greater than 1, for example, A can be 1.5. Specifically, the value of A can be set based on reservoir engineering experience, aiming to reasonably define the direct influence range of a single injection well and ensure that the production wells and the central injection well within the well group have a significant hydrodynamic connection.

[0092] In some embodiments, each injection well can be traversed, and the following operations can be performed to complete the well group division: For the current injection well, search the global well network of the reservoir for all production wells whose distance from the injection well is less than or equal to its radius of influence. These selected production wells, along with the injection well itself, are grouped into a newly created well group unit. This unit is defined as a well group centered on the injection well.

[0093] In actual oil reservoirs, a single production well may be located within the influence radius of multiple injection wells. It is permissible to allow such production wells to belong to multiple well groups simultaneously, thus accurately reflecting the multi-source disturbances they experience in subsequent optimization.

[0094] By dynamically constructing well groups centered on injection wells and based on actual well spacing, a genuine hydrodynamic relationship between wells within each optimization unit is ensured. This allows subsequent collaborative optimization to be based on a physical foundation consistent with reservoir mechanisms, avoiding optimization distortions caused by arbitrary grouping. Furthermore, the massive global well network problem is decomposed into multiple controllable and closely interconnected local well group problems, creating conditions for implementing a divide-and-conquer incremental optimization strategy and fundamentally avoiding the computational explosion and convergence difficulties commonly encountered when directly performing global optimization. In addition, the above partitioning method does not rely on fixed grids or rules and can adapt to various irregular and heterogeneous actual well network deployments, exhibiting good versatility and robustness.

[0095] S30: Select the target well group with the highest control priority from multiple well groups that have never been optimized.

[0096] In some embodiments, the control priority of each well group can be a quantitative ranking index, where a higher value indicates that the well group should be given greater priority in the subsequent optimization process.

[0097] In some embodiments, step S30 may specifically include: selecting a target well group sequentially from the remaining unoptimized well groups based on the control priority of each well group.

[0098] In some embodiments, all well groups can be sorted from highest to lowest based on their total well group production, and an ordered set of well groups can be established with the well group as the smallest unit. The total number of well groups is .from Take out the first well group As the current well group .

[0099] The control priority of each well group can be calculated, which can be determined by the production data of each well group. That is, all well groups can be sorted in order of total oil production from high to low to form an orderly well group processing queue, thereby ensuring that the well group that contributes the most to the oilfield production can get the optimized resources first.

[0100] S40: Iteratively perform optimization and merging operations on the target well group until the optimization termination condition is met; including: using the injection-production control model to optimize the control strategy of each well in the target well group; merging the optimized target well group with the well group with the highest control priority in its adjacent well groups to form a new target well group, wherein the adjacent well groups are those that have not been optimized and share at least one production well with the target well group.

[0101] In some embodiments, step S40 may specifically include: iteratively performing optimization and merging operations on the target well group until the optimization termination condition is met; including: using an injection-production control model to optimize the control strategy of each well in the target well group; merging the optimized target well group with the well group with the highest control priority in its adjacent well groups to form a new target well group, wherein the adjacent well groups are well groups that have not been optimized and share at least one production well with the target well group.

[0102] In some embodiments, the above-mentioned optimization of the control strategy of each well in the target well group using the injection-production control model may specifically include: using the control strategies of each well in multiple unoptimized well groups as constraints, and using the injection-production control model to optimize the control strategy of each well in the target well group.

[0103] Specifically, the control strategies of all other unoptimized wells not selected for the target well group—that is, wells that are not optimized and are not in the target well group—are fixed as boundary conditions and constraint parameters for the optimization process. The injection and production parameters corresponding to these strategies remain unchanged in this optimization. This method of fixing the strategies of non-target well groups as constraints essentially solidifies the reliable results of previous optimizations into the basis for subsequent optimizations. It ensures the continuity of the optimization process, prevents strategy oscillations during optimization, and greatly enhances the stability of the entire system.

[0104] In some embodiments, under these constraints, an injection-production control model can be used to recalculate and collaboratively optimize the control strategies for each well within the target well group. The optimization objective can be to maximize the total oil production or overall development benefits of the target well group while satisfying global constraints.

[0105] In some embodiments, after local optimization is completed, a merging operation can be performed to expand the optimization scope. Specifically, adjacent well groups that are spatially related to the target well group can be identified, defined as well groups that share at least one identical producing well with the target well group. From all adjacent well groups, the adjacent well group with the highest control priority is selected. The optimized target well group can be merged with this adjacent well group with the highest control priority to form a new, larger target well group. After merging, the original two well groups can be removed from the queue of well groups to be processed.

[0106] The above optimization and merging process can be repeated to form an iterative loop. Each iteration is based on the results of the previous round, gradually expanding the scope of optimization. The loop stops when one of the following optimization termination conditions is met: 1) The target well group no longer has any adjacent well groups; 2) The number of injection wells contained in the target well group reaches a preset upper limit (e.g., 5 wells).

[0107] In some embodiments, for the current well group Based on the current regulatory characteristic parameters Modify the control methods and parameters for all wells, due to the current control characteristic parameters. Parameters in All parameters have been normalized, so the specific values ​​that need to be set for the control parameters can be calculated using these parameters. For example, increasing production can be achieved using pressure response. The physical meaning of its coefficient is the proportion of production increase under a unit pressure increment; the larger the value, the greater the production increase. After updating the control mode and parameters of this well group, while keeping the control mode and parameters of other wells outside this well group unchanged, the injection-production control model is called to obtain the output control characteristic parameters. .from Take out and The well group with the highest total production among adjacent well groups. The two wells are combined into one well group. Hejing Group from Remove from the list and set the well group as the current well group. ,like Figure 4 As shown, well group B1 and well group B2 are merged, and optimization and merging are performed iteratively until any one of the following conditions is met: 1) There is no longer any connection with the current well group. Up to adjacent well groups; or 2) the current well group. Includes the number of injection wells The limit of more than 5 groups prevents the dynamic expansion into groups from becoming too large and causing difficulties in global optimization.

[0108] This divide-and-conquer strategy decomposes the complex global optimization problem into a series of controllable local optimization problems, significantly reducing computational complexity and avoiding the convergence difficulties and instability issues common in traditional global optimization methods. Prioritization based on production contribution ensures optimal allocation of optimization resources, while the well-merging mechanism naturally diffuses local optimization results to the global optimum through shared production wells, forming a reliable path from local optima to global optima.

[0109] S50: Iteratively execute the above steps for selecting the target well group until the selection termination condition is met.

[0110] In some embodiments, the selection of termination conditions includes: there are no unoptimized well groups.

[0111] In some embodiments, step S50 may specifically include: iteratively executing the steps of selecting target well groups until there are no unoptimized well groups.

[0112] In some embodiments, a set of well groups consisting of multiple well groups can be checked. If there are unvisited well groups, the first well group is selected as the target well group, that is, the well group with the highest control priority among the remaining well groups. The following steps are repeated: using the control strategies of each well in the unoptimized well group that is not the target well group as constraints, the control strategies of each well in the injection and production control model are optimized; the optimized target well group is merged with the well group with the highest control priority in its adjacent well groups to form a new target well group; the optimization and merging operations are iteratively executed until the optimization termination condition is met.

[0113] Can be checked If there are still unvisited well groups, then the first well group (the one with the highest production among the remaining well groups) is selected as the current well group. Repeat step S30 until... Empty. Since two adjacent well groups will share at least one production well, such as... Figure 5 As shown, well group B1 and well group B2 share production well B4. Adaptive adjustment of the production well control method is achieved through multiple iterations per well group. During well group reconfiguration, the control method and injection / production priority of key wells are dynamically adjusted. At this time... The collection already contains the well's control methods and control parameters.

[0114] S60: Adjust the injection and production parameters of the corresponding wells according to the optimized control strategy for each well in the reservoir.

[0115] In some embodiments, step S60 may specifically include: adjusting the injection and production parameters of the corresponding wells according to the optimized control strategy for each well in the reservoir.

[0116] The target control strategy for each well can be a structured data object containing the control method (operation type instruction) and a set of multiple control parameters (specific numerical targets). The target control strategy can be parsed to generate corresponding injection and production parameter adjustment instructions. For production wells, the control parameters may include the target oil production value, bottom hole flowing pressure control value, and nozzle size, etc.; for injection wells, the control parameters may include the target water injection volume value, upper limit of injection pressure, and pump frequency setting value, etc.

[0117] Parameter commands can be sent to field equipment via an integrated remote control module. Specifically, the remote valve control module can automatically adjust the wellhead nozzle opening to change the production pressure differential and achieve production or flowing pressure control targets. The remote frequency converter driver control module can precisely adjust the pump speed to achieve continuous control of the fluid production. Furthermore, the remote frequency converter driver control module can adjust the injection pump frequency to precisely control the injection volume and injection pressure.

[0118] For wells that are marked as requiring workover measures in the strategy (such as perforation repair, sealing, etc.), a detailed workover plan preparation module can be used to generate a detailed work instruction, including the procedure arrangement, construction parameters and quality control points, and assign it to the on-site work team for execution.

[0119] After all parameter adjustment commands are executed, the actual changes of key parameters can be monitored in real time through the sensor module, including: the flow sensor monitors the actual value of the production / injection volume; the pressure sensor monitors the actual value of the wellhead pressure and bottom hole flowing pressure; and the power sensor monitors the changes in equipment energy consumption.

[0120] By seamlessly integrating intelligent strategy analysis with a remote control system, the manual intervention required in traditional methods is eliminated, transforming optimization decisions into on-site actions within minutes, significantly improving control response speed and execution efficiency. Furthermore, a complete record is maintained of strategy content, execution instructions, actual parameter changes, and equipment status, forming a full-chain data archive from decision-making to execution, providing comprehensive data support for effect evaluation, accountability, and system optimization.

[0121] In some embodiments, determining the control strategy for each well based on the first injection-production control data of each well in step S10 may specifically include: generating multiple third injection-production control data using multiple expert models based on the first injection-production control data; weighting and fusing the first injection-production control data and the multiple third injection-production control data according to the injection-production control model and the confidence level of each expert model to obtain the fourth injection-production control data for each well; and determining the control strategy for each well based on the fourth injection-production control data for each well.

[0122] Expert models can be auxiliary decision-making models built on different algorithm architectures or training strategies, including but not limited to: engineering experience simulators based on random forests; time series prediction models using attention mechanisms; and neural network models based on physical constraints.

[0123] The expert models can receive the same first injection and injection control data. Each expert model independently outputs its prediction results, forming a multi-dimensional feature vector with the same structure as the first injection and injection control data, which is the third injection and injection control data.

[0124] A first confidence level can be calculated for the injection-production control model. Specifically, the first confidence level can be based on the model's historical accuracy on the validation set and the degree of certainty in the prediction results. Similarly, a second confidence level can be calculated for each expert model. Specifically, the second confidence level can be based on its prediction accuracy and stability performance under specific operating conditions. Different expert models have differentiated advantages for specific reservoir conditions or development stages. The confidence level assessment mechanism can dynamically adjust the contribution of each model, resulting in more stable performance when facing challenging scenarios such as heterogeneous reservoirs and complex well networks.

[0125] The confidence level can be used as a weight to fuse the first and third injection-import control data to obtain more reliable fourth injection-import control data. Based on the fused fourth injection-import control data, the following preset strategy mapping rules can be applied: discretize the continuous control feature data into specific control methods; convert standardized parameters into actual engineering parameters through inverse normalization; and generate a control strategy containing specific operation instructions and parameter values.

[0126] By using a multi-model collaborative decision-making mechanism, the biases or errors that may arise from a single model are effectively avoided. The weighted fusion method makes full use of the advantages of each model under different working conditions to form a complementary effect. Even if a certain expert model has an abnormal output, the overall decision can still maintain a high level of accuracy.

[0127] In some embodiments, the above-mentioned division of the multiple wells into multiple well groups with the injection well as the spatial center based on a preset spatial window may specifically include: acquiring spatial distribution data, formation connectivity data, and production dynamic data of each injection well and each production well in the reservoir; calculating the corresponding spatial window parameters for each injection well based on the formation connectivity data and production dynamic data, wherein the parameters include at least the influence radius and direction factor; determining the spatial influence domain with a specific shape and range for each injection well according to the calculated spatial window parameters; assigning the production wells located within the spatial influence domain of the injection well to its respective well group, forming a well group centered on the injection well; the method for determining the spatial window parameters includes: calculating based on the pressure response relationship between the injection well and the surrounding production wells, the fluid connectivity coefficient, and historical interference data, wherein the spatial window expands in the direction of strong connectivity and contracts in the direction of weak connectivity.

[0128] It can acquire multi-dimensional data in reservoir development, including: spatial distribution data, precise coordinates of each injection well and production well, and distances between wells; formation connectivity data obtained through geological modeling and well test interpretation; and production dynamic data, including real-time monitoring data such as inter-well pressure response characteristics and flow rate changes.

[0129] Based on the aforementioned multi-source data, a unique spatial window parameter can be calculated for each injection well. By analyzing the pressure drop transmission velocity and amplitude between the injection well and surrounding production wells, and combining pressure monitoring data and interference test results, an influence radius calculation model considering reservoir heterogeneity is established to obtain the influence radius. By analyzing the fluid connectivity coefficients, historical interference patterns, and injection-production response characteristics in each direction, vector analysis is used to quantify the connectivity strength in each direction, generating a direction-specific connectivity index, i.e., the direction factor. By introducing formation connectivity and production dynamic data, well group division is upgraded from simple geometric distance judgment to intelligent identification based on reservoir seepage mechanisms, greatly improving the scientificity and accuracy of well group division. The adaptive spatial window can effectively capture the dominant seepage channels and flow boundaries in heterogeneous reservoirs, accurately identify the true influence range of injection wells, and lay a reliable physical foundation for subsequent collaborative optimization.

[0130] Based on the calculated spatial window parameters, a spatial influence domain with a specific shape and range is constructed for each injection well. A contour tracing algorithm can be used to expand the boundary in the strongly connected directions and correspondingly shrink the boundary in the weakly connected directions, centered on the injection well, forming an irregularly shaped influence domain that highly matches the actual underground seepage field. Based on real-time production dynamic data, a dynamic update mechanism for the spatial influence domain can be established to ensure that the influence domain range can adaptively adjust with changes in reservoir conditions.

[0131] All production wells located within the spatial influence domain of each injection well are automatically assigned to the well group to which that injection well belongs, ultimately forming multiple adaptive well groups centered around the injection well.

[0132] Well groups based on real seepage relationships ensure strong interactions between wells within the group, enabling targeted collaborative optimization and significantly improving the effectiveness and pertinence of control measures, thus providing technical support for refined oilfield management.

[0133] In some embodiments, the control priority of each well group in step S30 above is determined by the following method: calculating the total oil production data of multiple producing wells in each well group; determining the control priority of each well group based on the total oil production data of each well group, wherein the control priority of each well group is positively correlated with the total oil production data of that well group.

[0134] Total oil production data can represent the cumulative oil production of all producing wells in a specific well group within a set statistical period, directly reflecting the contribution weight of the well group in the current oilfield production structure.

[0135] Daily oil production data for each well within a well group can be obtained from a real-time database. Data quality verification and outlier filtering are then performed to ensure the reliability of the basic data. The verified production data are summed by well group dimension. After calculating the total oil production for all well groups, priorities are determined according to the following rules: well groups are sorted from highest to lowest total oil production, and either sequential numbering or exponential weighting is used to convert the sorting position into a priority value. This ensures that the control priority is positively correlated with the total oil production data; that is, well groups with higher production receive higher priority values.

[0136] By positively correlated regulation priority with well group oil production, the system ensures that well groups contributing the most to oilfield production receive priority access to optimized resources. This allows limited regulation resources to be invested in the most efficient processes, maximizing overall development benefits. Furthermore, this priority calculation based on accurate production data completely avoids subjective bias, providing a transparent and traceable decision-making basis for the optimization sequence, significantly improving the scientific rigor and repeatability of the regulation process. This optimization sequence, starting with high-yield well groups, establishes an efficient path for subsequent well group merging and global optimization, enabling the system to quickly identify the most critical production challenges, accelerate the overall convergence process, and improve the execution efficiency of the optimization algorithm.

[0137] In some embodiments, during the continuous operation of the injection-production system, priority assessment can be automatically updated with each data acquisition cycle. That is, when the oilfield production plan is adjusted or the well group composition changes, the control priority calculation process can be retried. This dynamic updating characteristic of the priority mechanism enables the system to reflect changes in the oilfield production status in a timely manner, ensuring that the optimization focus is always consistent with the current actual production structure, and improving the adaptability of the injection-production system in dynamic development environments.

[0138] In some embodiments, determining the control priority of each well group based on the total oil production data of each well group may further include: determining a first control priority for each well group based on the total oil production data of each well group, wherein the first control priority is positively correlated with the total oil production data of each well group; dividing multiple well groups into head well groups and long-tail well groups based on the production data of each well group using a preset cumulative production percentage threshold; calculating a second control priority for the head well group using a first evaluation model, wherein the second control priority of the head well group is positively correlated with the energy consumption efficiency of the well group; calculating a second control priority for the long-tail well group using a second evaluation model, wherein the second control priority of the long-tail well group is positively correlated with the control potential and the urgency of the well group's operating conditions; coupling the first control priority and the second control priority to obtain a third control priority for each well group; wherein the head well group and the long-tail well group are divided in the following manner: well groups whose cumulative oil production reaches a preset percentage of the total production are identified as head well groups, and the remaining well groups are identified as long-tail well groups.

[0139] Based on the actual total oil production data of each well group, the first control priority can be calculated. This priority has a simple positive correlation with the well group's production, and is achieved using linear weighting or sequential ranking methods to ensure that high-yielding well groups receive basic priority attention, providing a basic ranking framework for subsequent optimization.

[0140] Furthermore, the long-tail theory can be introduced to classify well groups. In practice, the cumulative oil production percentage distribution curve of all well groups is calculated. Well groups whose cumulative oil production reaches a preset percentage of total production (e.g., a typical value of 70%-80%) are identified as head well groups, while the remaining well groups are identified as long-tail well groups. This classification method overcomes the limitations of evaluating a single production index, making priority determination more consistent with the actual pattern in reservoir development where a few well groups contribute the majority of production, thus providing a theoretical basis for optimizing the precise allocation of resources.

[0141] For different types of well groups, specialized evaluation models can be used for refined assessment: For head well groups, a first evaluation model is used to calculate their second control priority, which is positively correlated with the well group's energy efficiency. By quantitatively analyzing the energy cost per unit output, resources are guided towards high-yield and high-efficiency well groups. For tail well groups, a second evaluation model is used to calculate their second control priority, which is positively correlated with both the well group's control potential and the urgency of its operating conditions. By comprehensively assessing the distribution of remaining oil, the expected increase in oil production from measures, and the level of production risk, the optimization potential of low-yield well groups is fully explored. This differentiated evaluation strategy focuses on energy efficiency assessment for head well groups, promoting quality and efficiency improvement in major producing areas; for tail well groups, it considers both control potential and the urgency of operating conditions, effectively tapping the production potential of low-yield well groups while promptly addressing production risks, achieving an organic unity of stable and increased production.

[0142] A weighted coupling function can be used to integrate the first and second control priorities to generate a final third control priority. During the coupling process, the main challenges of the current development stage of the oilfield can be considered, and the weight allocation of the two priority categories can be dynamically adjusted: increasing the weight of production-related factors during the production-increasing stage, emphasizing energy efficiency during the stable production stage, and strengthening operational risk considerations during the high water-cut period. Through this two-level priority coupling mechanism, the evaluation focus can be flexibly adjusted according to the changing needs of different development stages of the oilfield, making the priority determination method highly scalable and adaptable, meeting the decision-making needs under complex development conditions, and ultimately achieving the maximization of the benefits of limited optimal resources across the entire reservoir.

[0143] In some embodiments, step S40 above uses the control strategies of each well in the unoptimized multiple well groups as constraints, and uses an injection-production control model to optimize the control strategies of each well in the target well group. Specifically, it may include: obtaining second geological development data of multiple wells after optimization using the control strategies; using the control strategies of each well in the unoptimized multiple well groups as constraints, inputting the second geological development data into the injection-production control model to maximize the total oil production of multiple production wells in the target well group, and obtaining the optimized control strategies of multiple wells in the target well group.

[0144] The second geological development data can be obtained in real time through sensor networks after the implementation of control strategies. This data includes multi-dimensional information such as changes in formation pressure distribution, production profiles of production wells, injection profiles of injection wells, and evolution of fluid saturation fields.

[0145] Constraints can solidify the control strategies and corresponding parameters of all wells within the non-target well group into immutable boundary conditions, including but not limited to injection-production limits, pressure control ranges, and engineering operation restrictions. By solidifying unoptimized and non-target well group strategies into constraints, strategy oscillations during the optimization process are effectively prevented. While pursuing the maximization of production from the target well group, the stability of the global production system is ensured, avoiding the common drawback of traditional optimization methods that improve local conditions while harming the overall system.

[0146] The optimization objective function can be explicitly set as maximizing the total oil production of multiple producing wells in the current well group. This objective function can be expressed mathematically as follows: ,in Indicates the first well group Oil production of a production well.

[0147] A complete system of constraints can be established, including: extracting the control strategy parameters of all wells in the non-target well group, transforming these parameters into equality and inequality constraints for the optimization problem, and setting the overall material balance and energy conservation of the system as hard constraints. By transforming the fundamental principles of material balance and energy conservation in reservoir engineering into mathematical constraints, it can be ensured that all optimization results conform to the physical laws of reservoirs and are engineering-feasible.

[0148] Secondary geological development data after the implementation of the initial strategy can be acquired through a data acquisition system. Data assimilation techniques can then be used to update the parameter field of the reservoir numerical model, ensuring that the injection-production control model accurately represents the current reservoir state. Subsequently, optimization algorithms such as sequential quadratic programming or interior point methods can be employed for solution.

[0149] The new parameter set obtained from the optimization solution is transformed into specific control commands, and the physical rationality and feasibility of the new strategy can be quickly verified through the mechanism model to generate the optimized control strategy for the target well group.

[0150] In some embodiments, step S40 may further include: removing the target well group and the corresponding adjacent well group with the highest control priority from the plurality of unoptimized well groups.

[0151] After the merging operation is completed, the well group set can be updated: the original target well group and the target well group are removed from the well group set, the newly formed merged well group is added to the well group set, and the spatial adjacency relationships of all well groups are recalculated and updated. By dynamically maintaining the well group set and promptly removing processed units, the scale of the optimization problem is always kept under control, avoiding the exponential growth of computational complexity and ensuring the solvability and computational efficiency of large-scale reservoir optimization problems. In the special case where there are no adjacent well groups, the state of the target well group can be kept unchanged, and the remaining well groups that have not undergone optimization operations can be selected from the well group queue as target well groups.

[0152] In some embodiments, step S40 above iteratively performs optimization and merging operations on the target well group until the optimization termination condition is met, including: determining whether the optimized target well group meets the optimization termination condition; if it does, ending the optimization of the target well group; if it does not, determining a new control strategy for each well in the new target well group based on the optimized control strategy of each well in the target well group and the control strategy of each well in the adjacent well group with the highest control priority of the target well group; iteratively executing the optimization and merging steps until the optimization termination condition is met.

[0153] In some embodiments, the optimization termination conditions include: the target well group has no adjacent well groups, or the number of injection wells contained in the target well group reaches a preset upper limit.

[0154] Candidate control strategies can be temporary optimization results calculated by the injection-production control model within the current optimization cycle, including the control methods and parameter sets optimized in this round, as candidate schemes for the target control strategy.

[0155] The optimization termination condition can include two independent conditions: first, the target well group has no adjacent well groups, indicating that there is no further room for expansion; second, the number of injection wells contained in the target well group reaches a preset upper limit to prevent the optimization unit from becoming too large and causing excessive computational complexity. By setting dual termination conditions, the natural boundary of spatial expansion is considered, and the reasonable size of the optimization unit is controlled, ensuring that the iteration process terminates at an appropriate time and avoiding over-optimization and waste of computational resources.

[0156] When further iterations are needed, the input strategy, reconstructed based on the current optimization results and merged well group information, serves as the starting point for the next round of optimization.

[0157] After each well group optimization is completed, candidate control strategies for each well within the target well group are generated. Simultaneously, a termination condition judgment module is activated to evaluate the current state from two dimensions: scanning spatial topology to confirm the existence of adjacent well groups sharing production wells with the target well group, and counting the number of injection wells within the target well group and comparing it with a preset upper limit threshold.

[0158] When any termination condition is met, the strategy solidification stage begins: all candidate control strategies for all wells in the target well group are directly upgraded to the target control strategy, timestamps and version identifiers are added to the target control strategy, and these strategies are marked as having been optimized.

[0159] When the termination condition is not met, the strategy transfer mechanism is activated: that is, candidate control strategies for each well in the target well group are extracted, and the control strategies for each well in the target well group are obtained. Through the strategy fusion algorithm, a new control strategy for each well in the target well group is generated. Through the smooth conversion mechanism from candidate strategies to new initial strategies, the optimization results are effectively accumulated in the iteration process. Each round of optimization is built on the foundation of the previous round, forming a continuously improving optimization trajectory.

[0160] An iteration counter can be maintained to record the current optimization round and compare it with the maximum number of iterations to prevent infinite loops. Simultaneously, the convergence of the optimization objective can be monitored, and the iteration can be terminated early if the improvement in optimization effect is not significant.

[0161] For any abnormal situations that occur during the optimization process, such as policy conflicts or parameter out-of-bounds errors, a rollback mechanism can be activated to restore the previous valid policy state and record the abnormal information for subsequent analysis.

[0162] In some embodiments, the injection-production control model in the above method is constructed by: acquiring third geological development data and corresponding injection-production label data of multiple wells in the reservoir; and training the injection-production control model based on the third geological development data and the corresponding injection-production label data.

[0163] Third-party geological development data can serve as input features for model training, encompassing multi-dimensional static and dynamic reservoir information. Specifically, this can include: geological attribute data, including reservoir characteristic parameters such as formation depth, permeability distribution, porosity, effective thickness, and original formation pressure; production dynamic data, including time-series monitoring data such as historical oil production, water production, water injection, bottom hole flowing pressure, and production gas-oil ratio; well connectivity data, including relational parameters such as interference coefficient and connectivity index obtained through well test interpretation; and engineering attribute data, including engineering information such as completion method, perforation location, lift technology, and downhole tool status.

[0164] Injection and production label data can be used as the target output for model training. It is derived from the experience summary of historical successful control cases and specifically includes: control type label, including control method classification based on actual effect verification; parameter adjustment label, including the optimal injection and production parameter adjustment values ​​verified on site; and effect evaluation label, including effect indicators such as pressure response, yield change rate, and water content change rate after control implementation.

[0165] The acquired third-party geological development data undergoes unified preprocessing: missing value imputation, outlier detection and correction, dimension unification and numerical normalization, extraction of statistical and trend features of time series data, and construction of structured feature vectors suitable for model input.

[0166] The time-series cross-validation method is adopted, and the dataset is divided into two parts according to time order: the training set is used for model parameter learning, covering 70%-80% of the historical data; the validation set is used for hyperparameter tuning, accounting for 10%-15% of the historical data; and the test set is used for final model evaluation, retaining the most recent 10%-15% of the data.

[0167] The batch training process is performed, and the model parameters are updated through the backpropagation algorithm: after each round of training, the performance index on the validation set is calculated, the training strategy is dynamically adjusted according to the performance on the validation set, the best model parameters during the training process are saved, and multiple rounds of iteration are performed until the model performance converges.

[0168] In some embodiments, the injection-import control model can be trained through the following steps:

[0169] Step X1: Taking the reservoir as the unit, and based on the order of wells within the reservoir, perform unified dimensional fusion processing on all information in the order of geological information, dynamic information, well connectivity information, and discretized information (mainly including unstructured text). In general implementation, the dimension is unified to one dimension. Among these, geological information... ,in Indicates well No. Data for each geological stratum, including time series data such as depth, permeability, porosity, effective thickness, and original formation pressure, as well as discretized information (including test reports and dynamic analysis reports). Dynamic information. ,in Indicates well No. The data collection records include time series data such as oil production, water injection, and bottom hole flowing pressure, as well as unstructured text (including test reports and dynamic analysis reports). Well connectivity information is also included. ,in Indicates well No. The data records include well completion and treatment data, such as time series of well completions, treatments, and their predicted effects, as well as unstructured text (including construction reports and post-effect evaluation reports). After data normalization, the data for each well... Usually at a certain moment Represented as The dataset for all wells is .

[0170] Step X2: For any Construct supervised learning parameters. For wells... Extract and construct based on control type Pressure response Production change rate Water content increasing trend Inter-well interference coefficient Energy consumption coefficient The key parameters are combined and normalized (using the aforementioned normalization method based on the dataset mean and standard deviation) to construct a system of multidimensional feature factors for regulation. These key parameters form the output vector. A group One corresponding group .

[0171] Step X3: Building the Model The training objective is to minimize the prediction error. ,in These are the model's predicted values.

[0172] Step X4: Update parameters using the Adam optimizer. Repeat the above training process for the data at each time step, and use the historical predictions, i.e., the results of the time step. As an additional input: A rolling window or sequence accumulation feature can be added to capture the dynamic influence between injection and production well pairs. For future time steps, the current time step is the same as the historical time step, thus enabling adaptive updates of the control model.

[0173] Step X5: After model deployment, input well Update features Prediction can be performed, and well output can be generated. The corresponding updated features include the control type, pressure response, production change rate, water cut increase trend, well interference coefficient, energy consumption coefficient, and other multi-dimensional characteristic indicators used for injection and production control.

[0174] Step X6: The model is automatically adjusted and updated at preset time intervals, but the prediction results can also be manually adjusted after step X5 to implement human-machine collaborative processing.

[0175] Through a systematic model training process, scattered oilfield data is transformed into intelligent models with predictive capabilities, providing a scientific data-driven decision-making basis for injection and production regulation, and significantly reducing the reliance on personal experience in traditional methods. The model training process essentially involves the mining and solidification of historical successful regulation experiences, enabling excellent regulation strategies to be preserved and passed down in the form of model parameters, thus forming a continuous accumulation mechanism for oilfield development knowledge.

[0176] In some embodiments, after step S60, the method may further include: acquiring the fourth geological development data and corresponding injection-production tag data of the multiple wells after being regulated by the target regulation strategy; and fine-tuning the injection-production regulation model based on the fourth geological development data and the corresponding injection-production tag data.

[0177] The fourth type of geological development data can be a dynamic response dataset that is collected in real time by a multi-source sensor network deployed in well sites, pipelines and reservoirs after the implementation of the target control strategy. Specifically, it can include: formation pressure field change data after the implementation of control, dynamic response of production well fluid production profile and injection well injection profile, actual change monitoring value of inter-well interference coefficient, monitoring data of fluid saturation field evolution trend, and equipment operation status and energy consumption change records, etc.

[0178] The labeled data is generated based on the evaluation of actual control effects. Specifically, it can cover: success / failure indicators of the actual implementation of control strategies, quantitative evaluation of the actual effects of various control parameters, stability and safety rating of the system after control, and comprehensive evaluation indicators of economic benefits and energy efficiency.

[0179] A conservative learning strategy can be adopted to update model parameters, namely, to extract small batches of samples from the incremental learning sample library, perform a limited number of retraining rounds on the basis of the pre-trained model, use a significantly reduced learning rate to prevent catastrophic forgetting, and monitor the performance of the validation set in real time to ensure a steady performance improvement.

[0180] After fine-tuning, the generalization ability of the fine-tuned model can be verified on an independent test set. The changes in prediction accuracy of key indicators before and after fine-tuning can be compared, and a model update report can be generated, including improvement points and potential risks. The evaluation results can then be fed back to the next round of regulatory decision-making process.

[0181] The fine-tuning mechanism enables the injection-production control model to adapt to dynamic changes during reservoir development, including formation parameter evolution, well network structure adjustment, and development stage transitions, ensuring that the model always maintains a high degree of matching with the current reservoir state.

[0182] The following is a specific embodiment of this specification:

[0183] There are a total of 9 water injection wells (W1 to W9) and 10 production wells (P1 to P10) in a certain water injection development block.

[0184] Injection wells W1-W9: These wells are equipped with flow meters and valve group pressure gauges to collect real-time data on injection volume and pressure. For example, W1: daily injection volume 500 m³ / d, injection pressure 12.3 MPa; W2: daily injection volume 480 m³ / d, injection pressure 11.8 MPa; W3: daily injection volume 520 m³ / d, injection pressure 12.5 MPa; W4: daily injection volume 460 m³ / d, injection pressure 11.6 MPa; W5: daily injection volume 495 m³ / d, injection pressure 12.0 MPa; W6: daily injection volume 510 m³ / d, injection pressure 12.4 MPa; W7: daily injection volume 470 m³ / d, injection pressure 11.9 MPa; W8: daily injection volume 485 m³ / d, injection pressure 12.1 MPa; W9: daily injection volume 505 m³ / d, injection pressure 12.2 MPa.

[0185] Production wells P1~P10: Online metering devices and downhole pressure sensors are installed at the wellhead to collect data on fluid production, water cut, and bottom hole flowing pressure: P1: Daily fluid production 80 m³ / d, water cut 65%, bottom hole flowing pressure 8.4 MPa; P2: Daily fluid production 72 m³ / d, water cut 70%, bottom hole flowing pressure 8.1 MPa; P3: Daily fluid production 68 m³ / d, water cut 75%, bottom hole flowing pressure 7.9 MPa; P4: Daily fluid production 74 m³ / d, water cut 72%, bottom hole flowing pressure 8.0 MPa; P5: Daily fluid production 85 m³ / d... P6: Daily production of fluid 78 m³ / d, water cut 66%, bottom hole flowing pressure 8.2 MPa; P7: Daily production of fluid 70 m³ / d, water cut 73%, bottom hole flowing pressure 7.8 MPa; P8: Daily production of fluid 82 m³ / d, water cut 68%, bottom hole flowing pressure 8.3 MPa; P9: Daily production of fluid 76 m³ / d, water cut 71%, bottom hole flowing pressure 8.0 MPa; P10: Daily production of fluid 88 m³ / d, water cut 63%, bottom hole flowing pressure 8.5 MPa.

[0186] The aforementioned data is transmitted to the oilfield production command center via the acquisition module and then integrated into the well group data integration platform. The platform cleans the data, removes outliers, synchronizes the time, and standardizes it. The system performs dimensionality reduction processing on multi-dimensional features, following the order of geological information—dynamic information—geometric information—discrete data, and then merges the data to form a unified feature dataset. Each well contains six types of features: injection volume / production volume, pressure, water cut, energy consumption coefficient, inter-well interference coefficient, and stratigraphic information.

[0187] The control methods and parameter predictions for the well group consisting of 9 injection wells were performed sequentially. Most production wells underwent 2-3 rounds of adaptive adjustments according to the above injection-production control methods, resulting in the target control strategy for each well, as shown in Tables 1 and 2. For the injection wells (W1-W9): except for W4, the injection volume of the other 8 injection wells was increased by approximately 5%-10% to improve reservoir pressure maintenance and displacement efficiency; W4 remained unchanged to avoid adverse effects on adjacent high water-cut wells. For the production wells (P1-P10): production of some high water-cut wells was reduced (not exceeding 10%): for example, P2 decreased from 72 m³ / d to 66 m³ / d, P3 from 68 m³ / d to 62 m³ / d, P7 from 70 m³ / d to 64 m³ / d, and P9 from 76 m³ / d to 69 m³ / d. Moderately increase production in some wells (around 5%): for example, increase P1 from 80 m³ / d to 84 m³ / d, P5 from 85 m³ / d to 60 m³ / d, P8 from 82 m³ / d to 86 m³ / d, and P10 from 88 m³ / d to 92 m³ / d. Keep the rest stable to avoid fluctuations in production.

[0188] Overall results: The injection-production matching relationship between well groups was optimized; the displacement front advanced more evenly, reducing early water channeling; the total fluid production remained basically the same, but the effective oil production increased, and the upward trend of water cut was controlled.

[0189] After the optimization plan is implemented, new production data is automatically collected and entered into the data integration platform, triggering a new round of iteration; high-efficiency / low-efficiency well groups are re-identified; the next stage of injection-production adjustment is predicted; and the optimal injection-production control strategy is output in a dynamically updated manner.

[0190] Table 1

[0191]

[0192] Table 2

[0193]

[0194] Based on the above injection-production control method, this specification also proposes embodiments of the injection-production control device. For example... Figure 6 As shown, the injection-production control device 600 may specifically include the following modules:

[0195] Input module 601 is used to input the first geological development data of multiple wells in the reservoir into the injection-production control model to obtain the control strategy for each well;

[0196] The partitioning module 602 is used to divide the multiple wells into multiple well groups, each well group including an injection well and multiple adjacent production wells;

[0197] Select module 603 to select the target well group with the highest control priority from multiple well groups that have not been optimized;

[0198] The optimization module 604 is used to iteratively perform optimization and merging operations on the target well group until the optimization termination condition is met. This includes: using the injection-production control model to optimize the control strategy of each well in the target well group; merging the optimized target well group with the well group with the highest control priority in its adjacent well groups to form a new target well group, wherein the adjacent well groups are well groups that have not been optimized and share at least one production well with the target well group.

[0199] Iteration module 605 is used to iteratively execute the above steps for selecting target well groups until the selection termination condition is met;

[0200] The control module 606 is used to control the injection and production parameters of the corresponding wells according to the optimized control strategy for each well in the reservoir.

[0201] In some embodiments, the input module 601 described above can be specifically used for:

[0202] Acquire geological attribute data, production dynamic data, inter-well connectivity data, and engineering attribute data from multiple wells in the reservoir;

[0203] The geological attribute data, production dynamic data, well connectivity data, and engineering attribute data are spliced ​​together;

[0204] The vectorized and stitched data yields the first geological development data for multiple wells.

[0205] In some embodiments, the input module 601 described above can also be used for:

[0206] The first geological development data is input into the injection-production control model to obtain the first injection-production control characteristic data for each well. The first injection-production control characteristic data includes multiple items from the following: control type, pressure response, production change rate, water cut increase rate, inter-well interference coefficient, and energy consumption coefficient.

[0207] Based on the first injection-production control data of each well, determine the control strategy for each well. The control strategy includes at least the control method and the corresponding multiple control parameters for that well.

[0208] In some embodiments, the input module 601 described above can also be used for:

[0209] Based on the first injection-production control data, multiple third injection-production control data are generated using multiple expert models;

[0210] Based on the injection-production control model and the confidence levels of each expert model, the first injection-production control data and multiple third injection-production control data are weighted and fused to obtain the fourth injection-production control data for each well.

[0211] Based on the fourth injection-production control data of each well, the control strategy for each well is determined.

[0212] In some embodiments, the division module 602 described above can be specifically used for:

[0213] Based on a preset spatial window, the multiple wells are divided into multiple well groups with the injection well as the spatial center.

[0214] In some embodiments, the optimization module 604 described above can be specifically used for:

[0215] Calculate the total oil production data of multiple production wells in each well group;

[0216] Based on the total oil production data of each well group, the control priority of each well group is determined, and the control priority of each well group is positively correlated with the total oil production data of that well group.

[0217] In some embodiments, the optimization module 604 described above can also be used for:

[0218] Using the control strategies of each well in multiple unoptimized well groups as constraints, the injection-production control model is used to optimize the control strategies of each well in the target well group.

[0219] In some embodiments, the above optimization termination conditions include: the target well group has no adjacent well groups, or the number of injection wells contained in the target well group reaches a preset upper limit;

[0220] Based on this, the aforementioned optimization module 604 can be specifically used for:

[0221] Determine whether the optimized target well group meets the optimization termination conditions;

[0222] If the conditions are met, the optimization of the target well group is terminated.

[0223] If not, based on the optimized control strategy of each well in the target well group and the control strategy of each well in the adjacent well group with the highest control priority of the target well group, a new control strategy for each well in the new target well group is determined; the optimization and merging steps are executed iteratively until the optimization termination condition is met.

[0224] In some embodiments, the above-mentioned termination condition includes: there is no unoptimized well group.

[0225] In some embodiments, the injection-progression control device 600 described above can be specifically used for:

[0226] Obtain third-party geological development data and corresponding injection-production tag data from multiple wells in the reservoir;

[0227] The injection-production control model is trained based on the third geological development data and the corresponding injection-production tag data.

[0228] In some embodiments, the injection-progression control device 600 described above can also be used for:

[0229] Obtain the fourth geological development data and corresponding injection-production tag data of the multiple wells after being regulated using the target regulation strategy;

[0230] Based on the fourth geological development data and the corresponding injection-production tag data, the injection-production control model is fine-tuned.

[0231] As can be seen from the injection-production control device provided in the embodiments of this specification above, the embodiments of this specification can input the first geological development data of multiple wells in the reservoir into the injection-production control model to obtain the control strategy for each well; divide the multiple wells into multiple well groups, each well group including an injection well and multiple adjacent production wells; select the target well group with the highest control priority from the multiple unoptimized well groups; iteratively perform optimization and merging operations on the target well group until the optimization termination condition is met; including: using the injection-production control model to optimize the control strategy of each well in the target well group; merging the optimized target well group with the highest control priority well group in its adjacent well groups to form a new target well group, wherein the adjacent well groups are unoptimized well groups that share at least one production well with the target well group; iteratively execute the above steps of selecting the target well group until the selection termination condition is met; and control the injection-production parameters of the corresponding wells according to the optimized control strategy of each well in the reservoir. Through the divide-and-conquer strategy, the complex global optimization task is decomposed into a series of sequentially executed, controllable-scale local optimization problems. This incremental approach of well group optimization -> merging -> re-optimization effectively avoids the technical difficulties, computational oscillations, or convergence problems caused by excessive dependent variables and constraints in one-time global optimization. Using the control strategies of each well in the unoptimized well group as constraints ensures that the rest of the system remains stable while optimizing the current local area, preventing strategy oscillations. Each local optimization is performed under defined boundary conditions, making the evolution of the entire system smooth and controllable, and the optimization results can be effectively solidified and transferred to the next step. Furthermore, adjacent well groups sharing at least one identical producing well with the target well group ensures that the expansion of the optimization scope follows the real dynamic interference relationships between wells. The optimization path constructed by this method is not a mathematical abstraction but closely follows the natural connectivity of underground fluids, giving the final target control strategy a solid reservoir physics foundation and strong engineering feasibility. Overall, through multiple optimization and merging iterations, the quality of the strategy can be continuously corrected and improved. The final control strategy is the result of multiple rounds of collaborative consideration from local to global perspectives, possessing both a global vision and engineering feasibility.

[0232] This specification also provides a computer device for an injection-production control method, including a processor and a memory for storing processor-executable instructions. Specifically, the processor can perform the following tasks according to the instructions: inputting first geological development data of multiple wells in the reservoir into an injection-production control model to obtain a control strategy for each well; dividing the multiple wells into multiple well groups, each well group including an injection well and several adjacent production wells; selecting the target well group with the highest control priority from the unoptimized well groups; iteratively performing optimization and merging operations on the target well group until the optimization termination condition is met; including: optimizing the control strategies of each well within the target well group using the injection-production control model; merging the optimized target well group with the highest control priority well group in its adjacent well groups to form a new target well group, where adjacent well groups are unoptimized well groups that share at least one production well with the target well group; iteratively executing the above steps of selecting the target well group until the selection termination condition is met; and controlling the injection-production parameters of the corresponding wells according to the optimized control strategy for each well in the reservoir.

[0233] To execute the above instructions more accurately, please refer to... Figure 7 As shown in the embodiments of this specification, another specific computer device 700 is also provided, wherein the computer device 700 includes a network communication port 701, a processor 702 and a memory 703, and the above structures are connected by internal cables so that the various structures can perform specific data interaction.

[0234] The processor 702 can be specifically used for: inputting the first geological development data of multiple wells in the reservoir into the injection-production control model to obtain the control strategy for each well; dividing the multiple wells into multiple well groups, each well group including an injection well and multiple adjacent production wells; selecting the target well group with the highest control priority from the multiple unoptimized well groups; iteratively performing optimization and merging operations on the target well group until the optimization termination condition is met; including: using the injection-production control model to optimize the control strategy of each well in the target well group; merging the optimized target well group with the well group with the highest control priority in its adjacent well groups to form a new target well group, wherein the adjacent well groups are unoptimized well groups that share at least one production well with the target well group; iteratively executing the above steps of selecting the target well group until the selection termination condition is met; and controlling the injection-production parameters of the corresponding wells according to the optimized control strategy of each well in the reservoir.

[0235] The memory 703 can be used to store the corresponding instruction program.

[0236] In this embodiment, the network communication port 701 can be a virtual port bound to different communication protocols, thereby enabling the sending or receiving of different data. For example, the network communication port can be a port responsible for web data communication, a port responsible for FTP data communication, or a port responsible for email data communication. Furthermore, the network communication port can also be a physical communication interface or communication chip. For example, it can be a wireless mobile network communication chip, such as GSM or CDMA; it can also be a Wi-Fi chip; or it can be a Bluetooth chip.

[0237] In this embodiment, the processor 702 can be implemented in any suitable manner. For example, the processor can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers, etc. This specification is not limiting.

[0238] In this embodiment, the memory 703 includes volatile memory and non-volatile memory. The memory 703 can include multiple layers. In digital systems, anything that can store binary data can be a memory; in integrated circuits, a circuit with storage function but no physical form is also called a memory, such as RAM, FIFO, etc.; in a system, a storage device with a physical form is also called a memory, such as a memory stick, TF card, etc.

[0239] This specification also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements... Figure 1 The method shown.

[0240] This specification also provides a computer program product, including at least one instruction or at least one program segment, wherein the at least one instruction or the at least one program segment is loaded and executed by a processor to achieve the following: Figure 1 The method shown.

[0241] It should be understood that in the various embodiments of this specification, the sequence number of each process does not imply the 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 this specification.

[0242] It should also be understood that, in the embodiments of this specification, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this specification generally indicates that the preceding and following related objects have an "or" relationship.

[0243] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0244] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0245] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0246] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational tasks to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The task is a function specified in one or more boxes.

[0247] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for injection-production control, characterized in that, include: The first geological development data of multiple wells in the reservoir are input into the injection-production control model to obtain the control strategy for each well; The multiple wells are divided into multiple well groups, each well group including an injection well and multiple adjacent production wells; Select the target well group with the highest control priority from multiple well groups that have never been optimized; The optimization and merging operations for the target well group are performed iteratively until the optimization termination condition is met. This includes: using an injection-production control model to optimize the control strategy of each well in the target well group; merging the optimized target well group with the well group with the highest control priority in its adjacent well groups to form a new target well group, where the adjacent well groups are those that have not been optimized and share at least one production well with the target well group; The above steps for selecting the target well group are executed iteratively until the selection termination condition is met; Based on the optimized control strategy for each well in the reservoir, the injection and production parameters of the corresponding wells are adjusted.

2. The method according to claim 1, characterized in that, The method further includes: Acquire geological attribute data, production dynamic data, inter-well connectivity data, and engineering attribute data from multiple wells in the reservoir; The geological attribute data, production dynamic data, well connectivity data, and engineering attribute data are spliced ​​together; The vectorized and stitched data yields the first geological development data for multiple wells.

3. The method according to claim 1, characterized in that, The process involves inputting the first geological development data from multiple wells in the reservoir into the injection-production control model to obtain the control strategy for each well, including: The first geological development data is input into the injection-production control model to obtain the first injection-production control characteristic data for each well. The first injection-production control characteristic data includes multiple items from the following: control type, pressure response, production change rate, water cut increase rate, inter-well interference coefficient, and energy consumption coefficient. Based on the first injection-production control data of each well, determine the control strategy for each well. The control strategy includes at least the control method and the corresponding multiple control parameters for that well.

4. The method according to claim 3, characterized in that, The process of determining the control strategy for each well based on the first injection-production control data of each well includes: Based on the first injection-production control data, multiple third injection-production control data are generated using multiple expert models; Based on the injection-production control model and the confidence levels of each expert model, the first injection-production control data and multiple third injection-production control data are weighted and fused to obtain the fourth injection-production control data for each well. Based on the fourth injection-production control data of each well, the control strategy for each well is determined.

5. The method according to claim 1, characterized in that, The process of dividing the multiple wells into multiple well groups includes: Based on a preset spatial window, the multiple wells are divided into multiple well groups with the injection well as the spatial center.

6. The method according to claim 1, characterized in that, The method further includes: Calculate the total oil production data of multiple production wells in each well group; Based on the total oil production data of each well group, the control priority of each well group is determined, and the control priority of each well group is positively correlated with the total oil production data of that well group.

7. The method according to claim 1, characterized in that, The optimization of the control strategy for each well within the target well group using the injection-production control model includes: Using the control strategies of each well in multiple unoptimized well groups as constraints, the injection-production control model is used to optimize the control strategies of each well in the target well group.

8. The method according to claim 7, characterized in that, The optimization of the control strategies for each well within a target well group, using the control strategies of each well in multiple unoptimized well groups as constraints, and employing an injection-production control model, includes: Obtain the second geological development data of multiple wells optimized using the aforementioned control strategy; Using the control strategies of each well in multiple unoptimized well groups as constraints, the second geological development data is input into the injection-production control model to maximize the total oil production of multiple production wells in the target well group, thereby obtaining the optimized control strategies of multiple wells in the target well group.

9. The method according to claim 1, characterized in that, The optimization termination conditions include: the target well group has no adjacent well groups, or the number of injection wells contained in the target well group reaches a preset upper limit; The iterative optimization and merging operation on the target well group until the optimization termination condition is met includes: Determine whether the optimized target well group meets the optimization termination conditions; If the conditions are met, the optimization of the target well group is terminated. If not, based on the optimized control strategy of each well in the target well group and the control strategy of each well in the adjacent well group with the highest control priority of the target well group, a new control strategy for each well in the new target well group is determined; the optimization and merging steps are executed iteratively until the optimization termination condition is met.

10. The method according to claim 1, characterized in that, The termination criteria include: there are no unoptimized well groups.

11. The method according to claim 1, characterized in that, The injection-production control model is constructed in the following way: Obtain third-party geological development data and corresponding injection-production tag data from multiple wells in the reservoir; The injection-production control model is trained based on the third geological development data and the corresponding injection-production tag data.

12. The method according to claim 11, characterized in that, The method further includes: Obtain the fourth geological development data and corresponding injection-production tag data of the multiple wells after being regulated using the target regulation strategy; Based on the fourth geological development data and the corresponding injection-production tag data, the injection-production control model is fine-tuned.

13. An injection-production control device, characterized in that, The device includes: The input module is used to input the first geological development data of multiple wells in the reservoir into the injection-production control model to obtain the control strategy for each well; A partitioning module is used to divide the multiple wells into multiple well groups, each well group including an injection well and multiple adjacent production wells; The selection module is used to select the target well group with the highest control priority from multiple well groups that have not been optimized; The optimization module is used to iteratively perform optimization and merging operations on the target well group until the optimization termination condition is met. This includes: using the injection-production control model to optimize the control strategy of each well in the target well group; merging the optimized target well group with the well group with the highest control priority in its adjacent well groups to form a new target well group, where the adjacent well groups are those that have not been optimized and share at least one production well with the target well group. The iteration module is used to iteratively execute the above steps for selecting the target well group until the selection termination condition is met; The control module is used to adjust the injection and production parameters of each well in the reservoir according to the optimized control strategy.

14. A computer device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the method of any one of claims 1-12.

15. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 12.

Images