Offshore high water cut oilfield subsurface-surface coupling constraint vector flow field intelligent collaborative regulation method and system

Abstract

The application discloses an offshore high-water-cut oilfield underground-ground coupling constraint vector flow field intelligent collaborative regulation method and system. The method comprises the following steps: reversely mapping the multi-dimensional processing capacity and safety constraints of an offshore platform into mathematical constraints on injection-production rates; within the ground constraint boundary, taking the injection-production rates as decision variables, establishing an integrated coupling optimization model with the underground vector flow field balance degree and the development net present value as targets; solving the model by using a hybrid intelligent algorithm combining global search and local optimization to obtain an optimal injection-production rate scheme meeting the ground constraints; and converting the optimization result into field control instructions and executing the instructions, while dynamically adjusting through a closed-loop feedback. The application realizes the collaborative optimization of underground flow field balanced displacement and ground facility safety full efficiency, and significantly improves the development benefit and operation safety of the high-water-cut oilfield without increasing the ground investment.

Description

Technical Field

[0001] This invention relates to a method and system for intelligent collaborative control of underground-surface coupled constrained vector flow fields in offshore high water-cut oilfields, belonging to the field of intelligent development and optimization control technology for oil and gas fields. Background Technology

[0002] As offshore oil and gas field development enters the medium-to-high water-cut stage, reservoir heterogeneity intensifies, crossflow channels and high-permeability bands gradually become apparent, inter-layer and intra-layer conflicts significantly worsen, and injection-production response relationships become increasingly complex. Meanwhile, platforms with longer service lives generally face a series of problems, including aging equipment, space constraints, insufficient power supply margins, and oil, gas, and water treatment capacity approaching its limit.

[0003] In the current development model, the "underground reservoir-surface platform" is often regarded as two relatively independent subsystems: the reservoir engineering side mainly uses numerical simulation to determine the injection-production well network, pressure regime, and injection allocation scheme; the surface engineering side adjusts operating parameters and loads based on the process flow and equipment capacity. This segmented decision-making architecture suffers from problems such as model fragmentation, feedback lag, and inconsistent objectives. For example, reservoir optimization schemes that reduce the production of high water-cut wells to suppress water channeling may lead to sudden changes in platform separation process conditions, increased power load fluctuations, triggering safety interlocks or even production shutdowns; conversely, when the platform adjusts the operating status of certain production wells due to power peaks or equipment maintenance, it may disrupt the reservoir injection-production balance and induce new water flooding conflicts.

[0004] Existing technologies mostly focus on optimizing single subsystems. One type emphasizes the optimization of injection and production parameters based on reservoir numerical simulation, but it does not adequately consider the dynamic response capabilities, process constraints, and safety interlocks of surface processing units. Another type focuses on optimizing platform process energy consumption or equipment scheduling, typically simplifying inputs such as wellhead flow rate to static empirical values, lacking real-time coupling with the underground seepage field. In particular, platform-side constraints are often simplified to single thresholds such as "maximum processing capacity," ignoring the coupling attributes under the superposition of multiple factors such as spatial layout, power timing allocation, emergency redundancy, and safety interlocks, making it difficult to support true underground-surface closed-loop coordinated control.

[0005] Currently, no publicly available literature systematically proposes an integrated vector flow field collaborative control method and engineering system that uses multi-dimensional platform process and safety constraints as boundary conditions, a reservoir-platform coupling model as the core, and intelligent algorithms as the solution engine. In particular, there is a lack of implementable collaborative optimization paradigms for the complex conditions of high-water-cut offshore oilfields. Therefore, it is necessary to construct an intelligent collaborative control method and system for vector flow fields that integrates underground and surface operations, has computable constraints, and unifies multi-dimensional objectives, in order to improve the overall reservoir development effect and the platform's operational safety and economic benefits. Summary of the Invention

[0006] This invention aims to overcome the problems of fragmented control between underground reservoirs and surface platforms, coarse constraint characterization, and insufficient collaborative optimization capabilities in existing technologies, and provides a method and system for intelligent collaborative control of underground-surface coupled constrained vector flow fields for high water-cut oilfields.

[0007] This invention constructs a three-in-one technical path of "systematic modeling of multi-dimensional constraints for offshore platforms – coordinated optimization of reservoir-platform coupled vector flow field – intelligent solution and closed-loop feedback execution". Based on comprehensive consideration of reservoir seepage dynamics characteristics, platform process flow and safety constraints, it achieves: equalization of flow field spatial distribution and suppression of water channeling risk; stabilization of platform processing load and power load and improvement of safety margin; and overall optimization of development costs and economic benefits under multiple objectives and constraints.

[0008] The present invention provides a method and system for coordinated optimization of offshore oil reservoir injection and production vectors based on multidimensional constraints on the ground. It transforms the physical limits of the offshore platform ground processing system into dynamic constraint boundaries for oil reservoir injection and production optimization. Under the condition of maintaining the safe operation of ground facilities, it uses the injection and production rate of oil and water wells as decision variables and achieves coordinated matching of maximizing oil reservoir development benefits and safe and controllable load on ground facilities through underground flow field vector reconstruction.

[0009] The present invention provides a method for intelligent collaborative control of underground-surface coupled constrained vector flow fields in offshore high water-cut oilfields, comprising the following steps: S1. Based on the physical processing limits and safe operation logic of offshore platforms, a mathematical model of multi-dimensional constraints on the ground is constructed. The mathematical model is used to inversely map the operating capacity of the platform into mathematical constraints on the injection and production rates of oil and water wells. S2. Within the boundary of the multidimensional constraints on the ground, a coupled optimization model integrating the underground reservoir and the surface platform is established with the injection and production rates of each oil and water well as the decision variables. The coupled optimization model takes the comprehensive optimization of the equilibrium of the underground vector flow field and the economic benefits of development as the objective function. S3. Use an intelligent optimization algorithm to solve the coupled optimization model to obtain an injection-production rate scheme that satisfies the ground multidimensional constraints and optimizes the objective function; S4. Convert the optimal injection-production rate scheme into field-executable control commands for each oil and water well and issue them for execution.

[0010] Preferably, in step S1, constructing the mathematical model of ground multidimensional constraints specifically includes: A fluid processing capacity constraint model, a power load dynamic constraint model, and a safety cutoff and pressure constraint model are constructed, and these three are encapsulated into a unified constraint verification module. This constraint verification module receives a set of injection / production rate vectors and outputs a feasibility judgment result and a penalty value for breach of contract. As input, the output includes: Feasibility determination Boolean value: True (all constraints are satisfied) or False (constraints are violated); Penalty value for breach of contract: ,in For the first c A constraint function, These are the corresponding weighting coefficients.

[0011] Preferably, the fluid processing capacity constraint model establishes a functional relationship between the platform's total liquid production and at least one of the following fluid properties: overall water content, crude oil viscosity, and mixture density, based on the physical capacity and separation efficiency curves of the separator and / or dehydrator. The functional relationship can take the form of a linear model, a polynomial model, or a nonlinear regression model based on machine learning. As a specific modeling approach, the following functional relationship can be adopted:

[0012] in: Total liquid production (m³) 3 / d); The overall moisture content is (%). Crude oil viscosity (mPa·s); The density of the mixture (kg / m³) 3 ); This represents the maximum processing load allowed by the equipment.

[0013] More specifically, in one embodiment, the function can be represented as:

[0014] in This is the rated processing capacity of the equipment. and For reference operating parameters, , These are calibration coefficients, obtained through regression analysis of historical data.

[0015] This model is used to determine the maximum permissible wellhead production boundary under different produced fluid properties.

[0016] The dynamic constraint model for power load is based on the energy consumption characteristic curves of at least one type of equipment, including water injection pumps, external pumps, and electric submersible pumps. It establishes a correlation model between the total power consumption of the system and the injection and production intensity of each well, and is constrained by the available power supply capacity of the platform's microgrid. The power consumption of a single pump exhibits a power function, a polynomial relationship, or a functional relationship determined by the pump characteristic curve, which are related to the corresponding water injection or production volume. As a specific modeling approach, the following functional relationship can be adopted:

[0017] in: Water injection well power consumption: In the formula The density of the injected water (kg / m³) 3 ), Acceleration due to gravity (m / s²) 2 ), For the first The injection head of the injection well. This corresponds to the pump efficiency; Oil well power consumption: In the formula For the first Rated power of the wellhead electric submersible pump For the rated liquid production rate, Power index (determined based on pump type); This refers to the power consumption of the ground treatment system (including the power consumption of external pumps, sewage pumps, and the basic power consumption of the treatment system). For the platform microgrid at all times Available power supply capacity.

[0018] This model is used to limit the total injection and production rate under high energy consumption conditions.

[0019] The safety cutoff and pressure constraint model, based on the pipeline network pressure rating and emergency shutdown system trigger logic, transforms the pressure alarm threshold of key nodes into a constraint on instantaneous flow fluctuations. The relationship between the pressure of key nodes and the production rate of each well is determined through a hydraulic calculation model, which is based on the Darcy-Weisbach formula or a similar frictional pressure drop calculation formula. As a specific modeling method, the following functional relationship can be used:

[0020] in: For the first Pressure on key nodes; and These are the lower and upper limits of the pressure for that node, respectively. This represents the total number of key monitoring nodes.

[0021] The node pressure is determined through a network hydraulic calculation model. Specifically, the frictional pressure drop is calculated using the Darcy-Weisbach formula.

[0022] in This is the friction coefficient (related to the Reynolds number). For the length of the pipe section, For pipe diameter, The flow velocity is determined by dividing the sum of the fluid production from each well flowing through the pipe section by the cross-sectional area of ​​the pipe section.

[0023] This model is used to ensure that the ground system remains within safe operating boundaries under any injection-production combination.

[0024] Step S2, establishing an integrated coupled optimization model of the underground reservoir and the surface platform, specifically includes: S21: Based on the fitting results of a three-dimensional heterogeneous reservoir geological model and production history, streamline simulation technology is introduced to establish a reservoir vector flow field model. This model not only calculates the scalar distributions of pressure field, saturation field, etc., but also focuses on outputting the fluid velocity vector within each grid cell. The vector magnitude (velocity amplitude) and direction (streamline direction) are used to identify: high-velocity water channel: a region with an abnormally high velocity vector amplitude, usually corresponding to a high-permeability zone or a fractured zone; low-velocity stagnation zone: a region with a velocity vector amplitude close to zero, usually corresponding to a low-permeability dead oil zone or a displacement blind zone. S22: Set the daily water injection vector of each injection well and the daily liquid production vector of each production well as decision variables; The injection and production rates of each well were selected as optimization variables: Daily water injection volume vector of injection wells:

[0025] Daily fluid production vector of oil wells:

[0026] S23: Construct a multi-objective weighted optimization function as the objective function. The objective function includes a flow field equilibrium index and an economic index. The flow field equilibrium index is characterized by the variance of the reservoir velocity vector or the Gini coefficient, and the economic index is characterized by the net present value during the development cycle. Preferably, a multi-objective weighted optimization function is established:

[0027] in: The flow field uniformity index is characterized by minimizing the variance of the velocity vector across the entire reservoir:

[0028] in: The total number of grid cells. Let i be the velocity amplitude of the i-th grid. The mean flow velocity is the total reservoir velocity.

[0029] As an economic indicator, it represents the net present value over the development cycle:

[0030] in For oil prices, and These are cumulative oil production and water production, respectively. For energy consumption costs, This is the discount rate.

[0031] and These are weighting coefficients, adjusted according to the development strategy. S24: Use underground physical constraints and surface coupling constraints as constraints for the coupled optimization model; Subsurface constraints: Single-well injection-production rate range: , Injection-production pressure differential limit: avoid overpressure in injection wells leading to formation fracturing, and ensure that the bottom-hole flowing pressure of production wells is not lower than the saturation pressure; reservoir material balance: control the injection-production ratio within a reasonable range to avoid a continuous decline in formation pressure.

[0032] Ground coupling constraints: Invoke the constraint verification module to require the injection-production rate combination at any given time. The resulting ground loads (total fluid volume, total power consumption, and critical node pressure) strictly meet the following requirements: .

[0033] In step S3, the intelligent optimization algorithm is used to solve the coupled optimization model, specifically employing a hybrid optimization strategy of "global exploration + local gradient correction": Global exploration phase: A heuristic global search algorithm is used to search within the total solution space. The fitness function of each individual is modified according to the penalty value for breach of the ground multidimensional constraints. The heuristic global search algorithm includes particle swarm optimization algorithm, genetic algorithm or differential evolution algorithm. Preferably, a particle swarm optimization (PSO) algorithm or a genetic algorithm (GA) is used to search within the total solution space. During the algorithm iteration process: Each individual in each generation of the population (corresponding to a set of injection-progression rate schemes) must be verified by the constraint verification module of the above steps; Individuals that violate ground constraints are given a high penalty value, and the fitness function is modified as follows:

[0034] After multiple iterations, a set of feasible injection-production schemes that meet all ground constraints was selected. Local refinement stage: Taking the preferred feasible solution obtained in the global exploration stage as the initial point, a deterministic local optimization algorithm is used to locally fine-tune the injection and extraction velocity vector, and the ground multidimensional constraints are re-verified after each fine-tuning to ensure that the solution is always within the feasible region. The deterministic local optimization algorithm includes sequential quadratic programming algorithm, interior point method or adjoint gradient method.

[0035] Throughout the iteration process, any solution that causes a ground constraint violation (such as separator level exceeding limit, power overload, or pipeline pressure exceeding limit) will be immediately discarded and will not participate in subsequent iterations.

[0036] In step S4, converting the optimal injection-production rate scheme into field-executable control commands specifically includes: The optimized water injection volume of the injection well is converted into the percentage of the water injection valve opening or the set value of the water injection pump frequency; the optimized production volume of the oil well is converted into the operating frequency of the electric submersible pump, the size of the downhole nozzle, or the opening of the production nozzle; the method also includes a closed-loop feedback step S5: real-time monitoring of the actual production data after the command is executed, and when the deviation between the actual production volume, water injection volume, surface pressure or power load and the model prediction value exceeds the set threshold, steps S1 to S4 are automatically triggered to be re-executed for dynamic closed-loop control.

[0037] The present invention also provides a ground-based multidimensional constraint-based offshore reservoir injection-production vector collaborative optimization system for implementing the method, comprising: Ground-based constraint inverse characterization module: used to acquire static and dynamic data of offshore platforms, construct and store multi-dimensional constraint mathematical models of fluid handling capacity, power load and safety pressure, and provide constraint verification interface; The underground vector flow field simulation module is used to store the three-dimensional geological grid model that has been historically fitted and calibrated, and calculate the flow velocity vector of each grid cell based on Darcy's law, and output the flow field uniformity index and water channeling intensity identification results of the entire reservoir. The underground-surface coupled collaborative optimization module is used to set an objective function with injection-production rate as the decision variable and flow field equilibrium and economic net present value as the objectives. It calls the hybrid intelligent solution unit to solve for the optimal injection-production working regime while calling the surface constraint inverse characterization module to perform boundary verification. Control command generation and execution module: used to convert the optimal injection and extraction working system into executable parameters of the field control system, and to monitor execution deviations through a closed-loop feedback unit to trigger system re-optimization.

[0038] Preferably, the ground constraint inverse characterization module specifically includes: Multi-source data acquisition unit: used to acquire platform static equipment parameters, pipeline topology and dynamic production data in real time; Static data: Piping and instrumentation diagrams for offshore platforms, equipment rated parameters (separator capacity, pump rated power, pipe diameter specifications, etc.), and pipeline topology; Dynamic data: Real-time production data (liquid production / injection volume of each well, flow meter readings, pressure transmitter data, power load monitoring values); Relationships: Establish topological mapping relationships between individual wells and surface facilities (manifolds, processing units, and transportation systems); Sub-constraint modeling unit: used to establish fluid handling capacity constraint model, power load constraint model and safety interlock constraint model respectively, wherein the functional form of each model is calibrated by regression analysis or machine learning method; Fluid processing capacity constraint model: established based on the rated capacity and separation efficiency curves of separators and dewaterers; Electricity load constraint model: established based on the energy consumption characteristic curves of water injection pumps, external transmission pumps, and electric submersible pumps; Safety interlock constraint model: established based on pipeline pressure rating and emergency shutdown system logic; Constraint Integration Verification Unit: Used to receive external input injection / sampling rate combinations It calls each sub-constraint model in parallel for calculation and outputs the feasibility judgment result (Boolean value (True / False)) and the default penalty signal.

[0039] The underground vector flow field simulation module is used to simulate the vectorized motion state of fluids inside the reservoir, specifically including: Geological model storage unit: Stores a 3D geological mesh model calibrated using history fitting, including: Static parameters: porosity, permeability (horizontal / vertical), rock compressibility, capillary pressure curve; Fluid properties: crude oil viscosity-temperature relationship, formation water salinity, dissolved gas-oil ratio, and relative permeability curve; Structural features: fault distribution, sand body connectivity, and description of interlayers.

[0040] Velocity vector calculation unit: configured to solve the reservoir numerical model using the finite difference or finite element method based on Darcy's law and the reservoir material balance equation, outputting: The velocity vector of each grid cell includes the vector magnitude (velocity amplitude) and direction (three-dimensional spatial angle). The overall reservoir flow field isostatic index is quantified by either the velocity vector variance or the Gini coefficient. Water channeling intensity identification: Marking the spatial location and velocity distribution of high-speed water channeling channels.

[0041] Preferably, the hybrid intelligent solution unit in the underground-surface coupled collaborative optimization module is configured to run a hybrid algorithm of particle swarm optimization and sequential quadratic programming, or a hybrid algorithm of genetic algorithm and adjoint gradient method, specifically including: Variable and target setting unit: Vector of injection and production rates of all oil and water wells in the field. Set as the sole decision variable; flow field uniformity With development net present value The weighted combination is set as the optimization objective function.

[0042] Hybrid Intelligent Solver Unit: Configured to run a hybrid intelligent algorithm combining global search and local optimization (PSO + gradient method or GA + local search). In each iteration: The "Underground Vector Flow Field Simulation Module" (Module 2) is called in the forward direction to calculate the objective function value (flow field equilibrium, oil production, energy consumption, etc.) of the current injection and production scheme. The "Ground Constraint Reverse Characterization Module" (Module 1) is called in reverse to perform boundary condition verification, and solutions that cause ground power overload, separator overflow or pipeline pressure exceeding limits are automatically eliminated.

[0043] Optimal solution filtering unit: configured to filter from the solution set where the algorithm has converged: It meets all hard constraints on the ground (processing capacity, power, safety pressure); Maximize the sweep efficiency of the underground flow field (optimal flow field uniformity). The most economically efficient (maximum net present value) injection-production rate combination .

[0044] The control command generation and execution module is used to convert the optimization calculation results into on-site executable operation commands, specifically including: Command conversion unit: configured to convert the optimal injection / import rate (Physical quantity: m) 3 Convert / d or t / d) into executable parameters for the field control system: Water injection well: water injection valve opening percentage (0-100%), water injection pump inverter frequency setting value (Hz); Oil well: EV pump inverter frequency setting (Hz), recommended downhole nozzle size setting or production nozzle opening percentage.

[0045] Closed-loop feedback unit: configured to monitor actual production data in real time after command execution and perform model-actual deviation analysis. When the deviation between the actual liquid production / injection volume, ground pressure, or power load and the model prediction exceeds a set threshold (e.g., ±10%); The system is automatically triggered to re-invoke the above modules to execute the optimization process, thereby achieving dynamic closed-loop control.

[0046] The present invention has the following significant beneficial effects: 1. Underground-Surface Integrated Collaboration: For the first time, surface facility constraints are inversely mapped into the dynamic boundary of reservoir optimization, achieving collaborative optimization "with surface constraints as the premise and underground regulation as the means"; 2. No increase in surface investment: No need to expand or modify surface facilities; the overall efficiency of the system can be improved simply by optimizing the downhole injection and production configuration. 3. Fine-grained control of vector flow field: Based on the flow velocity vector field, water channeling channels and dead oil areas are identified, enabling precise measures to be taken "where there is a problem, adjust accordingly"; 4. Real-time dynamic adaptation: It has a closed-loop feedback mechanism and can dynamically update and optimize the plan according to actual production deviations to adapt to the dynamic changes in the reservoir. Attached Figure Description

[0047] Figure 1 This is a flowchart of the method for coordinated optimization of injection and production vectors in offshore oil reservoirs based on ground-based multidimensional constraints, as proposed in this invention.

[0048] Figure 2 This is a schematic diagram of the offshore reservoir injection-production vector collaborative optimization system based on ground multidimensional constraints according to the present invention. Detailed Implementation

[0049] The core innovation of this invention lies in establishing a closed-loop technical framework of "inverse characterization of surface constraints—vector simulation of subsurface flow field—coupled optimization of subsurface and surface—execution of field commands." Unlike existing technologies that treat the surface system as an independent optimization object, this invention transforms the physical limits of surface facilities into dynamic boundary conditions for reservoir injection and production optimization. Without changing the operating parameters of surface facilities, it achieves a coordinated match between development benefits and surface load simply by adjusting the downhole injection and production velocity vector.

[0050] The following section uses a real-world application case of Platform A in an offshore oil field to illustrate the specific implementation of the method and system of this invention.

[0051] Example 1: A Co-optimization Method for Offshore Oil Reservoir Injection and Production Vectors Based on Ground-based Multidimensional Constraints I. Project Background and Initial Conditions Platform A in a certain offshore oil field has 12 water injection wells and 18 production wells. The surface treatment system includes: a three-phase separator (rated processing capacity of 8000 m³ / s). 3 / d), electrostatic dehydrator (rated processing capacity 5000m³) 3 The platform includes a water injection pump set (total installed power 2800kW) and an external transmission pump set (total installed power 1500kW). The available power supply capacity of the platform microgrid is 6000kW, and after deducting basic loads such as living and lighting, the power surplus available for production is 4800kW.

[0052] The main problems currently facing the platform include: 1. The water cut in some oil wells is rising rapidly (average water cut reaches 92%), causing the separator to frequently operate close to full load; 2. The high energy consumption of electric submersible pumps in high water-cut wells leads to power shortages during peak periods of platform power load; 3. The reservoir exhibits significant heterogeneity, with water channeling occurring in some injection-production well groups, while other low-permeability areas are underutilized.

[0053] To address the above problems, the method of this invention is applied for coordinated optimization of injection and sampling vectors. The specific steps are as follows: Step 1: Inverse characterization and quantification of multidimensional processing capability constraints of offshore platforms 1. Constructing a surface-wellhead topology correlation model and data foundation (1) Based on the piping and instrumentation drawings of platform A, establish a four-level topology connection model: Level 1: 18 oil wells are connected to 4 wellhead manifolds (W1, W2, W3, W4). Second stage: The four wellhead manifolds converge into the main production pipe and enter the separator group; Third stage: After separation, crude oil enters the electrostatic dehydrator, wastewater enters the wastewater treatment unit, and natural gas enters the gas treatment unit; Level 4: The treated crude oil is transported to a shuttle tanker via an external pumping unit, and the treated wastewater is partially reinjected and partially discharged.

[0054] (2) Collect static parameter data: Three-phase separator rated capacity m 3 / d, maximum allowable moisture content 95%; rated capacity of electric dehydrator m 3 / d, requires the water content of the imported crude oil. ; Four units of a water injection pump set have a rated power of 700kW each. (Efficiency curve) (Normalized representation); Three external pump sets with a rated power of 500kW each, efficiency curve ; The installed power of electric submersible pumps in each oil well ranges from 90 to 160 kW, and the actual power consumption varies dynamically with the production volume.

[0055] (3) Collect dynamic operation data (taking a typical day as an example): Real-time data on fluid production, water cut, wellhead pressure, and ESP current / frequency of 18 oil wells; Real-time data on water injection volume, wellhead pressure, and water injection pump current / frequency of 12 injection wells; Data on liquid level, pressure, temperature, and total power consumption of each ground-based processing unit.

[0056] The above data is automatically collected through the SCADA system interface at a sampling frequency of 1 minute / time. After outlier removal (such as points with pressure surges >20%) and moving average filtering, a standardized time series dataset is formed.

[0057] 2. Functional Modeling of Multidimensional Constraint Boundaries (1) Fluid handling capacity constraint model Based on statistical analysis of actual operating data from the separator and electrostatic dehydrator, a capacity constraint model is established:

[0058] in: This represents the total liquid production of the entire platform. For overall moisture content; The viscosity of crude oil is measured in real time using an online viscometer or estimated based on an empirical formula for temperature. , mPa·s is the reference operating condition parameter; , These are calibration coefficients, obtained through regression analysis of historical data.

[0059] The model shows that when the overall water content or crude oil viscosity deviates from the reference value, the actual processing capacity of the separator will decrease.

[0060] For electrostatic dehydrators, since they process dehydrated crude oil, separate constraints are required:

[0061] in This represents the amount of crude oil after dehydration.

[0062] (2) Dynamic constraint model of power load Based on the energy consumption characteristic curves of each device, a total system power consumption model is established:

[0063] in: Water injection well power consumption: ,in kg / m 3 For the density of the injected water, m / s 2 , Let be the injection head of the i-th well (determined by the difference between the wellhead pressure and the formation pressure). This corresponds to the pump efficiency; Oil well power consumption: ,in The rated power of the electric submersible pump in well j is... For the rated liquid production rate, Power index (determined based on pump type); Ground system power consumption: ,in For the power consumption of the external pump, For the power consumption of the sewage pump, To handle the system's basic power consumption (approximately 300kW); Available power capacity: kW.

[0064] This model can transform power constraints into direct limitations on injection and production rates. For example, when a high-water-cut well is producing at a high flow rate, the surge in power consumption of the electric submersible pump may cause the total power consumption to exceed the limit. In this case, it is necessary to reduce the well's production or adjust the injection rate of the injection well simultaneously.

[0065] (3) Safety cut-off and pressure constraint model Five key monitoring nodes were identified in the pipeline network: four manifolds and one main production pipe inlet. Pressure constraints were established based on the pipeline pressure rating (allowable pressure 10 MPa) and emergency shutdown system settings.

[0066] in: Determined through a hydraulic calculation model of the pipeline network: ; Frictional voltage drop Calculated according to the Darcy-Weisbach formula: ,in This is the friction coefficient (related to the Reynolds number). For the length of the pipe section, For pipe diameter, For flow rate; MPa (to prevent gas evolution). MPa (below the allowable pressure, with a safety margin).

[0067] In practical applications, pressure-flow relationship curves are established for each of the four wellhead manifolds. When the total production of the oil wells collected by a certain manifold exceeds a threshold, the manifold pressure will exceed the limit, triggering an emergency shutdown. Therefore, constraints are required.

[0068] in It is obtained by reverse calculation through pressure constraints.

[0069] 3. Modular encapsulation of constraint models The above three types of constraints are encapsulated into a constraint verification function CheckConstraints, the interface of which is designed as follows: enter: Injection velocity vector (30-dimensional vector) Output: is_feasible: A boolean value, True indicates that all constraints are satisfied, False indicates that a default occurs. penalty_value: A floating-point number representing the penalty for breaching the constraint. violation_details: A dictionary that records in detail the violations of each constraint. Step 2: Co-optimization of reservoir injection and production vector flow field under surface constraints 1. Construct a detailed reservoir vector flow field model (1) Establishment of three-dimensional geological model Based on the seismic interpretation results, well logging data, and core analysis data of the reservoir where Platform A is located, a three-dimensional geological model was established: Mesh size: 120 meshes in the X direction, 80 meshes in the Y direction, and 15 layers in the Z direction, for a total of 144,000 mesh cells; Grid step size: 50m×50m in the X / Y direction, and 1-3m in the Z direction depending on the thickness of the sublayer; Porosity field: The sequential Gaussian simulation method was used, with a mean of 14.5% and a standard deviation of 3.2%. Spatial correlation reflects the distribution of sedimentary facies zones. Permeability field: The cut-off Gaussian simulation method was used, with a mean of 280 mD. Significant heterogeneity was observed, with high-permeability bands (>800 mD) accounting for approximately 12% and low-permeability areas (<50 mD) accounting for approximately 18%. Fluid properties: Crude oil density 850 kg / m³ 3 Viscosity 15 mPa·s (50℃), formation water salinity 25,000 mg / L.

[0070] (2) Historical Fit Calibration The production history of the past three years (January 2022 - December 2024) was fitted using reservoir numerical simulation software: Fitting objective: The relative error between simulated and measured values ​​of monthly oil production, water production, and bottom hole flowing pressure for each well. ; Adjust parameters: local grid permeability, endpoints of the relative permeability curve, water body size, etc. Fitting quality: The average fitting error for the cumulative oil production of the 18 wells was 6.8%, and the average fitting error for the cumulative water production was 7.2%, which met the accuracy requirements.

[0071] (3) Calculation of velocity vector field Based on the historical fitting model, streamline simulation technology is used to calculate the velocity vector field at the current time (2025.1): For each grid cell Calculate the three-dimensional flow velocity vector:

[0072] in Let be the permeability tensor of this grid. For fluid viscosity, For pressure gradient; Calculate the magnitude of the flow velocity vector: ; Statistical analysis of velocity distribution across the entire reservoir: Average flow velocity m / d; High-speed area (flow velocity) (m / d): mainly distributed in high-permeability strips extending from near water injection wells W1 and W3 to oil production wells P5 and P8, with a cumulative pore volume accounting for 9.3%; Low-velocity region (flow velocity <0.05m / d): mainly distributed in the low-permeability area in the northeast of the reservoir, accounting for 21.7% of the total pore volume.

[0073] (4) Identification of water channel and dead oil zone Based on velocity vector field analysis, three main water channeling channels were identified: Channel 1: Water injection well W1 → Oil production well P5, length approximately 850m, average flow velocity 1.35m / d, corresponding permeability >900mD; Channel 2: Water injection well W3 → Oil production well P8, length approximately 720m, average flow velocity 1.18m / d, corresponding permeability >750mD; Channel 3: Water injection well W7 → Oil production well P12, length approximately 660m, average flow velocity 0.95m / d, corresponding permeability >650mD.

[0074] Two main dead oil zones were identified at the same time: Dead oil zone A: Northeast corner of the reservoir, with a pore volume of approximately 8.5 × 10⁻⁶. 4 m 3 The remaining oil saturation is 58%, but the flow rate is <0.03m / d, resulting in low displacement mobilization. Dead oil zone B: A low-permeability interlayer between wells P14 and P16, with a pore volume of approximately 6.2 × 10⁻⁶. 4 m 3 The remaining oil saturation is 62%.

[0075] The above flow field diagnosis results provide clear optimization targets for subsequent injection and production adjustments: suppressing water channeling and activating dead oil zones.

[0076] 2. Construct a coupled optimization model with injection and extraction rate as the sole variable. (1) Setting decision variables Define a 30-dimensional decision variable vector:

[0077] in: Water injection well injection volume range: m 3 / d, a total of 12 variables; Oil well production rate range: m 3 / d, a total of 18 variables.

[0078] The upper and lower limits of the velocity for each well are determined comprehensively based on the integrity of the wellbore, the pressure-bearing capacity of the reservoir, and the capabilities of the equipment.

[0079] (2) Construction of the objective function Establish a multi-objective weighted optimization function:

[0080] Flow field uniformity target Displacement uniformity is characterized by minimizing the variance of the velocity vector.

[0081] in Number of active grids (oil saturation) (grid) For the first The velocity amplitude of each grid, This represents the average flow velocity across the entire reservoir. The larger this index (the smaller the absolute value of the negative value), the more balanced the flow field and the smaller the difference between high-speed channels and low-speed dead zones.

[0082] The economic target NPV is used to calculate the net present value of development over the next 5 years (2025-2028):

[0083] in: For month number, discount rate / year; oil price Yuan / barrel; water treatment cost Yuan / m 3 (Including costs for chemical reagents, wastewater treatment, and discharge); electricity costs Yuan / kWh; Chemical reagent cost Including the injection of polymers, corrosion inhibitors, and scale inhibitors, at a rate of 2.5 yuan / m³ based on the injection volume. 3 count.

[0084] The weight coefficients in the objective function are set as follows: , This indicates that, while ensuring flow field equilibrium, greater emphasis is placed on economic benefits. The weights can be adjusted according to the development strategy phases: in the initial development stage, emphasis is placed on flow field equilibrium (…). In the later stages of development, the focus is on economic benefits. ).

[0085] (3) Loading of constraints The optimized model must satisfy the following constraints: Underground physical constraints Single-well injection-production rate range:

[0086]

[0087] The bottom pressure of the injection well should not exceed the fracturing pressure:

[0088] in For formation pressure, For skin pressure drop, For wellbore friction pressure drop, The formation fracturing pressure is determined through small-scale fracturing tests, with a safety factor of 0.9.

[0089] The bottom-hole flowing pressure of the oil well shall not be lower than the saturation pressure.

[0090] in MPa is the saturation pressure of crude oil, and the coefficient of 1.1 provides a safety margin to prevent wellbore degassing.

[0091] Injection-procurement balance constraints:

[0092] This constraint ensures that the injection-production ratio is within a reasonable range, avoiding a continuous decrease in formation pressure or excessive pressure buildup.

[0093] Ground coupling constraint Call the constraint verification function in step S1.3:

[0094] This requires the combination of injection and extraction rates at any given time. All constraints related to ground handling capacity, power load, and pipeline pressure must be met.

[0095] 3. Solution based on hybrid intelligent algorithm A hybrid strategy combining Particle Swarm Optimization (PSO) and Sequential Quadratic Programming (SQP) is employed. (1) PSO global exploration phase Algorithm parameter settings: Particle swarm size: One particle; Maximum number of iterations: generation; Inertia weight: (Linearly decreasing); Learning factors: (Balancing individual and global learning); Speed ​​limit: (To prevent particles from flying out of the feasible region).

[0096] Fitness function design:

[0097] in The objective function value, The penalty value returned by the constraint verification function.

[0098] After PSO global search, a feasible set of solutions satisfying all ground constraints is obtained, where the globally optimal solution is denoted as... .

[0099] (2) SQP Local Refinement Stage by Starting from point 1, a sequential quadratic programming algorithm is used for local optimization: Gradient calculation: The adjoint method is used to efficiently calculate the gradient of the objective function with respect to 30 decision variables. This avoids the huge computational burden of calling the simulator 31 times, which is required by the traditional finite difference method.

[0100] Quadratic programming subproblem: in the... In the next iteration, the solution is obtained as follows:

[0101]

[0102] in The Hessian matrix approximation for the objective function is obtained by updating it using the BFGS method. For the first Constraint functions.

[0103] Iterative updates: ,in This is the step size factor (determined through line search).

[0104] Constraint dynamic monitoring: CheckConstraints is called after each iteration. If the new solution violates the ground constraints, the step size is reduced. Try again to ensure the solution always lies within the feasible region.

[0105] After 15-25 SQP iterations, the objective function increment... The solution converges and the final optimal solution is obtained. .

[0106] 4. Optimize result output and execution plan generation The optimal injection-production rate scheme is obtained by solving the problem using the above hybrid optimization algorithm. The specific values ​​are as follows: Optimized water injection well allocation scheme (adjustments compared to the original scheme): Table 1 Optimized Injection Schemes from W1 to W12

[0107] Optimized oil well production allocation scheme (adjustments compared to the original scheme): Table 2 Optimized Injection Schemes from P1 to P18

[0108] Key optimization performance metrics: 1. Significantly improved flow field uniformity: Before optimization, the velocity variance (m / d) 2 ; Optimized flow velocity variance (m / d) 2 The pore volume ratio decreased by 47.6%; the pore volume ratio in the high-speed channel (flow velocity > 1.0 m / d) decreased from 9.3% to 4.8%; and the pore volume ratio in the low-speed region (flow velocity < 0.05 m / d) decreased from 21.7% to 15.2%.

[0109] 2. Ground load meets constraints and is optimized: Before optimization, the separator load rate (Near full load), after optimization There is an 11% margin; the peak power load before optimization is 4750kW, and after optimization it is 4520kW, a reduction of 4.8%, leaving a 280kW margin from the upper limit of 4800kW; the pressure of all pipeline nodes is within the safe range of [2.5, 7.8]MPa, and there is no risk of emergency shutdown triggering.

[0110] 3. Improved economic benefit forecasts: Original plan: Cumulative oil production of 485,000 tons over the next 5 years, NPV = 283 million yuan; Original plan: Cumulative oil production of 523,000 tons over the next 5 years, NPV = 316 million yuan; Incremental benefits: Increased crude oil production by 38,000 tons (+7.8%), NPV increased by 33 million yuan (+11.7%).

[0111] Convert the optimization results into on-site control commands: Based on the optimized injection-production rate The following model is used to convert it into executable device parameters: (1) Valve opening degree and pump frequency setting of water injection well For the injection well Wi, its injection volume With valve opening and pump frequency The relationship is determined by the hydraulic model:

[0112] in For valve flow coefficient, This refers to the pressure difference between the water injection pump outlet and the wellhead. In actual operation, the pump frequency is kept constant at the rated value of 50Hz, and only the valve opening is adjusted. Well W1: Valves closed from 85% to 60%; Well W5: Valves opened from 70% to 88%; Well W8: Valves opened from 75% to 92%; Other wells were adjusted similarly to ensure that the actual water injection volume reached the optimized target value.

[0113] (2) Frequency setting of ESP in oil well For oil well Pj, its production volume With electric submersible pump frequency The relationship is determined by the pump performance curve:

[0114] in For the rated liquid production rate, To increase the height, To improve pump efficiency, adjust the inverter frequency based on the optimization results. P5 well: Pump frequency decreased from 48Hz to 32Hz, and expected production decreased from 240m³ to 160m³. 3 / d; P8 well: Pump frequency decreased from 52Hz to 36Hz, and expected production decreased from 250 to 170m³. 3 / d; P14 well: pump frequency increased from 28Hz to 32Hz, in conjunction with W8 injection to increase production; other wells were adjusted synchronously according to optimization targets.

[0115] (3) Recommended sizes for oil nozzles For some flowing wells or gas lift wells, production is controlled by replacing the downhole nozzles. According to the nozzle flow rate formula:

[0116] in For flow coefficient, Let nipple cross-sectional area be the area of ​​the nozzle. For formation pressure, This refers to the wellhead pressure. For P12 wells requiring production control, it is recommended to replace the nozzle from 32 / 64 inches to 24 / 64 inches.

[0117] Generate on-site execution work orders: The above parameters are compiled into a standardized "Injection and Production Adjustment Work Order", including: Well number, current working parameters, target working parameters, adjustment range; expected daily oil production, daily water production, and daily water injection after adjustment; implementation schedule: adjustments will be made in three batches, with a 3-day interval between each batch to avoid drastic fluctuations in formation pressure; key monitoring points: after adjustment, monitor wellhead pressure, production volume, water cut, and ESP current for 7 consecutive days. If the deviation is >15%, on-site fine-tuning is required.

[0118] II. Implementation Effectiveness Monitoring and Closed-Loop Feedback Comparison of actual results one month after implementation: Table 3 Comparison of results after one month of implementation

[0119] The actual results closely matched the predictions, with deviations all within ±3%, validating the model's accuracy. Daily oil production increased from 558 m³ / h. 3 / d increased to 597m 3 / d, an increase of 7.0%, compared to the predicted value of 611m 3 The deviation of / d is mainly due to the fact that the actual production increase of well P14 was slightly lower than expected (affected by wax deposition in the wellbore).

[0120] Example of closed-loop feedback mechanism triggering: Monitoring in the second month (February 2024) revealed that the actual fluid production of well P5 was 172m³. 3 / d, which is 160m higher than the optimization target. 3 The pressure drop was approximately 7.5% per day, and the water cut rose to 94% (higher than the predicted 92%). The reason for this was that a nearby shut-in well, P4A, was temporarily shut down due to casing damage, which caused changes in the local pressure field.

[0121] The system automatically triggered a re-optimization: 1. Update the geological model: Update the P4A well status to shut-in and rerun the reservoir simulator; 2. Repeat step S2: Keep the parameters of other wells unchanged, and only perform local optimization on well P5 and its neighboring wells W1, W2, and P6; 3. Output Adjustment Plan: It is recommended that the pump frequency of well P5 be further reduced to 28Hz, and the water injection rate of well W1 be reduced from 195m³ to 175m³. 3 / d; 4. On-site execution: Issue new work orders and complete adjustments within 3 days.

[0122] After adjustment, the production rate of well P5 decreased to 165m³. 3 / d, the moisture content stabilized at 92%, returning to near the optimization target.

[0123] Example 2: Offshore reservoir injection-production vector collaborative optimization system based on ground-based multidimensional constraints 1. System Overall Architecture A layered decoupling design is adopted, including: Data layer: Used to store static equipment parameters, real-time SCADA data, and geological models; Modeling layer: used to construct the ground constraint model and the underground flow velocity vector field model; Optimization layer: Used to execute coupled optimization algorithms and output the optimal injection and sampling scheme; Execution layer: Used to convert the optimal solution into equipment control commands and send them to SCADA.

[0124] Hardware configuration: Dual 48-core processors, 512GB memory, dual A100 GPU accelerator cards; Software environment: CentOS 7.9, Python 3.8, PostgreSQL 13; Communicates with the offshore platform via a 10Mbps satellite link with a data latency of <2s.

[0125] 2. Ground-constrained inverse characterization module Multi-source data acquisition unit: Reads flow rate, pressure, liquid level, and power consumption data at 1-minute intervals via OPCUA interface and stores them in InfluxDB; Topology modeling unit: Automatically extracts equipment symbols and pipe diameter, length, and material information based on P&ID drawings, generates a directed topology graph, and stores it in JSON format; Automatic calibration of the constraint model unit: Random forest regression is used to train the separator load rate, R 2 ≥0.92; Multiple regression was used to fit the power consumption of the electric submersible pump, and the mean absolute error was <5kW; Constraint Integration Verification Unit: Encapsulates fluid processing capacity, power load, pipeline pressure, and single-well velocity range into a unified verification interface, outputting feasibility, penalty value, and violation details.

[0126] 3. Underground Vector Flow Field Simulation Module Geological model storage unit: Using a corner grid format, porosity, permeability, pressure, and saturation parameters are stored in HDF5 format, with a single file size of approximately 2.8GB and a loading time of ≤15s; Velocity vector calculation unit: After obtaining the pressure field by calling the OPMFlow simulator, the three-dimensional velocity of each grid is calculated in parallel based on Darcy's law. The time taken for 140,000 grids is ≈5 seconds. Flow field equalization quantification unit: calculates velocity variance, Lorentz coefficient, sweep efficiency and high / low velocity zone ratio, and uses graph theory method to automatically identify water channeling channels and dead oil zones; Result output unit: Writes the flow field and equilibrium index into the memory database for the optimization layer to call in real time.

[0127] 4. Underground-Surface Coupled Cooperative Optimization Module Variable and objective setting unit: Define a 30-dimensional injection and extraction velocity vector and its upper and lower limits, and construct a weighted objective function of flow field equilibrium and net present value, with the weights adjustable in stages; Hybrid intelligent solution unit: First, the particle swarm algorithm is used for global exploration. Then, the optimal solution obtained is used as the initial value, and the sequential quadratic programming method is used for local refinement. During the iteration, the gradient is quickly calculated through the adjoint method. Optimal solution selection unit: The candidate solutions obtained from multiple random initializations are scored for feasibility, robustness and adjustment range, and the one with the highest comprehensive score is selected as the final recommended solution.

[0128] 5. Control command generation and execution module (1) Instruction conversion unit: Water injection well: Calculate the target valve opening based on the valve flow coefficient and pump outlet pressure, and limit it to the range of 20%-95%; ESP wells: Based on the pump similarity law, the target fluid production rate is converted into the frequency of the frequency converter, which is limited to 25-55Hz; Self-flowing / air-lift wells: Calculate the required orifice diameter based on the nozzle flow rate formula and round it to the standard specifications; (2) Batch instruction generation unit: The 30 wells are divided into 3 batches according to the adjustment range, with an interval of 3 days between each batch, and the water injection wells and oil production wells are adjusted synchronously to maintain stable formation pressure; (3) Closed-loop feedback unit: After execution, the liquid production, water content, node pressure and total power consumption are monitored for 7 consecutive days. If the deviation is greater than 15% or the constraint exceeds the limit, local re-optimization is automatically triggered, and a new instruction is issued within 3 days to achieve closed-loop control.

Claims

1. A method for intelligent collaborative control of underground-surface coupled constrained vector flow field in offshore high water-cut oilfields, comprising the following steps: S1. Based on the physical processing limits and safe operation logic of offshore platforms, a mathematical model of multi-dimensional constraints on the ground is constructed. The mathematical model is used to inversely map the operating capacity of the platform into mathematical constraints on the injection and production rates of oil and water wells. S2. Within the boundary of the multidimensional constraints on the ground, a coupled optimization model integrating the underground reservoir and the surface platform is established with the injection and production rates of each oil and water well as the decision variables. The coupled optimization model takes the comprehensive optimization of the equilibrium of the underground vector flow field and the economic benefits of development as the objective function. S3. Use an intelligent optimization algorithm to solve the coupled optimization model to obtain an injection-production rate scheme that satisfies the ground multidimensional constraints and optimizes the objective function; S4. Convert the optimal injection-production rate scheme into field-executable control commands for each oil and water well and issue them for execution.

2. The method according to claim 1, characterized in that: In step S1, constructing the mathematical model of ground multidimensional constraints specifically includes: A fluid processing capacity constraint model, a power load dynamic constraint model, and a safety cutoff and pressure constraint model are constructed, and the three are encapsulated into a unified constraint verification module. The constraint verification module is used to receive a set of injection and extraction velocity vectors and output the feasibility judgment result and the default penalty value.

3. The method according to claim 2, characterized in that: The fluid processing capacity constraint model is based on the physical capacity and separation efficiency curves of the separator and / or dehydrator, establishing a functional relationship between the platform's total liquid production and at least one of the fluid properties parameters, namely, comprehensive water content, crude oil viscosity, and mixed liquid density; the functional relationship may take the form of a linear model, a polynomial model, or a nonlinear regression model based on machine learning. The power load dynamic constraint model is based on the energy consumption characteristic curves of at least one type of equipment, including water injection pumps, external pumps, and electric submersible pumps. It establishes a correlation model between the total power consumption of the system and the injection and production intensity of each well, and is limited by the available power supply capacity of the platform microgrid. Among them, the power consumption of a single pump has a power function relationship, a polynomial relationship, or a functional relationship determined by the pump characteristic curve with respect to the corresponding water injection volume or liquid production volume. The safety cut-off and pressure constraint model is based on the pipeline pressure rating and emergency shutdown system trigger logic, which transforms the pressure alarm threshold of key nodes into a constraint on instantaneous flow fluctuations; the relationship between the pressure of the key nodes and the production of each well is determined by a hydraulic calculation model, which is based on the Darcy-Weisbach formula or a similar frictional pressure drop calculation formula.

4. The method according to any one of claims 1-3, characterized in that: Step S2, establishing an integrated coupled optimization model of the underground reservoir and the surface platform, specifically includes: S21: Based on the fitting results of the three-dimensional heterogeneous reservoir geological model and production history, streamline simulation technology is introduced to establish a reservoir vector flow field model. The model outputs the fluid velocity vector in each grid cell to identify high-speed water channeling channels and low-speed stagnation zones. S22: Set the daily water injection vector of each injection well and the daily liquid production vector of each production well as decision variables; S23: Construct a multi-objective weighted optimization function as the objective function. The objective function includes a flow field equilibrium index and an economic index. The flow field equilibrium index is characterized by the variance of the reservoir velocity vector or the Gini coefficient, and the economic index is characterized by the net present value during the development cycle. S24: Use underground physical constraints and surface coupling constraints as constraints for the coupled optimization model.

5. The method according to any one of claims 1-4, characterized in that: In step S3, the intelligent optimization algorithm is used to solve the coupled optimization model, specifically employing a hybrid optimization strategy of "global exploration + local gradient correction": Global exploration phase: A heuristic global search algorithm is used to search within the total solution space. The fitness function of each individual is modified according to the penalty value for breach of the ground multidimensional constraints. The heuristic global search algorithm includes particle swarm optimization algorithm, genetic algorithm or differential evolution algorithm. Local refinement stage: Taking the preferred feasible solution obtained in the global exploration stage as the initial point, a deterministic local optimization algorithm is used to locally fine-tune the injection and extraction velocity vector, and the ground multidimensional constraints are re-verified after each fine-tuning to ensure that the solution is always within the feasible region. The deterministic local optimization algorithm includes sequential quadratic programming algorithm, interior point method or adjoint gradient method.

6. The method according to any one of claims 1-5, characterized in that: In step S4, converting the optimal injection-production rate scheme into field-executable control commands specifically includes: The optimized water injection volume of the injection well is converted into the percentage of the water injection valve opening or the set value of the water injection pump frequency; the optimized production volume of the oil well is converted into the operating frequency of the electric submersible pump, the size of the downhole nozzle, or the opening of the production nozzle; the method also includes a closed-loop feedback step S5: real-time monitoring of the actual production data after the command is executed, and when the deviation between the actual production volume, water injection volume, surface pressure or power load and the model prediction value exceeds the set threshold, steps S1 to S4 are automatically triggered to be re-executed for dynamic closed-loop control.

7. A ground-based multidimensional constraint-based offshore reservoir injection-production vector collaborative optimization system for implementing the method of any one of claims 1-6, comprising: Ground-based constraint inverse characterization module: used to acquire static and dynamic data of offshore platforms, construct and store multi-dimensional constraint mathematical models of fluid handling capacity, power load and safety pressure, and provide constraint verification interface; The underground vector flow field simulation module is used to store the three-dimensional geological grid model that has been historically fitted and calibrated, and calculate the flow velocity vector of each grid cell based on Darcy's law, and output the flow field uniformity index and water channeling intensity identification results of the entire reservoir. The underground-surface coupled collaborative optimization module is used to set an objective function with injection-production rate as the decision variable and flow field equilibrium and economic net present value as the objectives. It calls the hybrid intelligent solution unit to solve for the optimal injection-production working regime while calling the surface constraint inverse characterization module to perform boundary verification. Control command generation and execution module: used to convert the optimal injection and extraction working system into executable parameters of the field control system, and to monitor execution deviations through a closed-loop feedback unit to trigger system re-optimization.

8. The system according to claim 7, characterized in that: The ground constraint inverse characterization module specifically includes: Multi-source data acquisition unit: used to acquire platform static equipment parameters, pipeline topology and dynamic production data in real time; Sub-constraint modeling unit: used to establish fluid handling capacity constraint model, power load constraint model and safety interlock constraint model respectively, wherein the functional form of each model is calibrated by regression analysis or machine learning method; Constraint Integration Verification Unit: Used to receive the injection and sampling speed combination from external input, call each sub-constraint model in parallel for verification, and output the feasibility judgment result and the default penalty signal.

9. The system according to claim 8, characterized in that: The hybrid intelligent solution unit in the underground-surface coupled collaborative optimization module is configured to run a hybrid algorithm of particle swarm optimization algorithm and sequential quadratic programming algorithm, or a hybrid algorithm of genetic algorithm and adjoint gradient method.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.

Images