Intelligent optimization decision-making method and system for oilfield water injection system

Abstract

The invention discloses an intelligent optimization decision-making method and system for an oilfield water injection system, and belongs to the technical field of oilfield development engineering. The method comprises the following steps: constructing a pipe network digital twinborn model and carrying out actual measurement verification; executing hydraulic-thermal simulation based on the model to generate a pressure-flow map; optimizing water distribution in combination with an injection allocation scheme and matching a pump set combination; generating a control instruction and executing a pressurization scheme; monitoring the state of the pipe network in real time and correcting model parameters. Through the digital twin technology and the differential evolution algorithm, global optimization decision of water distribution and pump set combination of the water injection system is achieved, and the problems that a traditional system is high in energy consumption and low in efficiency are solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of oilfield development engineering, and particularly relates to an intelligent optimization decision method and system for an oilfield water injection system. BACKGROUND

[0002] In the development process of an oilfield, water injection is a key measure to maintain formation energy and improve oil recovery. However, the traditional water injection system of an oilfield has many drawbacks, which seriously restricts the efficient development of the oilfield.

[0003] From the aspect of data collection, the traditional system relies on manual and timed collection of some key data, such as the wellhead pressure of the water injection well, the water injection volume, etc. The collection of real-time operation parameters of the pipe network node pressure, pipe section pressure loss, water injection pump (such as efficiency, lift, etc.) and dynamic change data of the oil reservoir (such as formation pressure distribution, permeability change, etc.) is seriously lacking, and the data update cycle is long (usually hours or even days), which cannot reflect the running status of the water injection system and the actual state of the oil reservoir in real time and comprehensively. According to statistics, the data coverage rate of the traditional system is less than 60%, and the real-time data delay is more than 30 minutes, which makes it difficult for workers to accurately control the actual running status of the water injection system and the dynamic change of the oil reservoir, and cannot timely find potential problems in the system, such as local pipe network blockage and water injection pump efficiency decline.

[0004] In the decision-making process, the formulation and adjustment of the traditional water injection scheme excessively rely on manual experience and lack scientific model and algorithm support. Engineers mainly set parameters such as water injection pressure and water injection volume according to historical data and personal experience, and it is difficult to formulate the optimal water injection scheme for complex oil reservoir conditions and water injection system running status. For example, when facing the problem of under-injection of water injection wells caused by changes in oil reservoir permeability, the traditional method can only take simple measures to increase pressure, but cannot accurately analyze the root cause of under-injection (such as increased pipe network resistance, formation blockage of water injection wells, etc.) and formulate targeted solutions. This leads to poor results of water injection scheme optimization, such as low water injection efficiency (the average water injection efficiency is 20%-30% lower than the theoretical value), high energy consumption (the energy consumption per unit of water injection volume is 15%-20% higher than the advanced level), and unsatisfactory oil reservoir development results (the oil recovery rate is 10%-15% lower than expected).

[0005] System coordination level: the three units of water injection station, pipe network and oil reservoir are managed independently, and there is a lack of overall optimization mechanism. The existing technology is difficult to balance the multi-objective constraints of water injection pressure, flow distribution and pump efficiency, and often appears the problem of local optimization and overall energy consumption increase.

[0006] As oilfield development deepens, complex reservoir conditions place higher demands on water injection systems. There is an urgent need for an intelligent system that can integrate multi-source real-time data, establish accurate simulation models, and achieve global optimization decisions, in order to break through the technical limitations of traditional methods in terms of data integrity, decision-making science, and system energy efficiency. Summary of the Invention

[0007] The present invention provides an intelligent optimization decision-making method and system for an oilfield water injection system, so as to solve the problem of high energy consumption and low efficiency of traditional water injection systems by achieving global optimal decision-making on water volume distribution and pump group combination of the water injection system based on water injection station equipment parameters, pipe network topology and reservoir geological data through digital twin modeling and differential evolution algorithm optimization.

[0008] In order to solve the above technical problems, the present invention provides an intelligent optimization decision-making method for an oilfield water injection system, comprising:

[0009] Obtain the water injection station equipment, geology, and pipeline network parameters for 3D modeling. After field measurement verification and error compensation, a digital twin model of the pipeline network is generated.

[0010] Perform hydraulic and thermal simulation based on the digital twin model of the pipe network, extract pressure distribution and pressure loss data, and generate pressure-flow maps;

[0011] Obtain the injection plan and set constraints on the pressure-flow map, optimize the water output distribution, match the pump library, and generate an efficient pump group ID combination plan;

[0012] Analyze the high-efficiency pump group ID combination solution to query the characteristic curve, generate the inverter control signal and convert it into the booster pump start and stop instructions and pipe valve action sequence;

[0013] Collect oil pressure and pressure loss data and combine them with maps to determine the cause of short injection, generate a multi-objective optimization boosting plan, and perform pressure fluctuation monitoring;

[0014] Collect energy consumption and efficiency indicators for target comparison, analyze recovery factor changes to calculate deviation rates, and correct pipeline roughness parameters in the digital twin model of the pipeline network.

[0015] Furthermore, the acquisition of water injection station equipment, geology, and pipe network parameters for three-dimensional modeling, and the generation of a pipe network digital twin model through field verification and error compensation, includes:

[0016] Obtain the equipment parameters of the water injection station, the geological parameters of the water injection wells, and the geometric parameters of the water injection network, perform three-dimensional digital modeling, and obtain the initial network model;

[0017] Extract the pipeline roughness coefficient and equipment resistance coefficient from the initial pipeline network model and compare and verify them with the measured data;

[0018] Error compensation is performed on the results of the measured data comparison and verification to generate a digital twin model of the pipeline network.

[0019] Furthermore, the hydraulic-thermal simulation is performed based on the pipe network digital twin model to extract pressure distribution and pressure loss data and generate a pressure-flow map, including:

[0020] Based on the digital twin model of the pipeline network, the real-time collected water injection pressure and flow rate data are injected to perform hydraulic-thermal coupling simulation calculations;

[0021] Extract the pressure distribution of pipeline nodes and the pressure loss data of pipe sections from the hydraulic-thermal coupling simulation results to perform resistance loss analysis;

[0022] Generate a pressure-flow map based on the resistance loss analysis results and mark the high pressure loss areas.

[0023] Furthermore, the method of obtaining the injection scheme and setting constraints on the pressure-flow map, optimizing the water output distribution and matching the pump library, and generating an efficient pump group ID combination scheme includes:

[0024] Obtain injection well allocation plans and pressure-flow maps, and set node pressure constraints and flow fluctuation thresholds;

[0025] The differential evolution algorithm is used to optimize the water output distribution of each injection station and calculate the minimum output power of the injection station.

[0026] Furthermore, the step of optimizing the water output distribution of each water injection station by using the differential evolution algorithm and calculating the minimum output power of the water injection station further includes:

[0027] Generate an efficient pump group ID combination solution based on the optimized water output matching pump characteristic database.

[0028] Furthermore, the method of parsing the high-efficiency pump group ID combination scheme to query the characteristic curve, generate the inverter control signal and convert it into the booster pump start and stop instructions and pipe valve action sequence, including:

[0029] Analyze the ID combination scheme of high-efficiency pump groups and query the working characteristic curve of the pump groups;

[0030] Generate inverter frequency value and valve opening control signal according to the pump group working characteristic curve;

[0031] The control signal is converted into booster pump start and stop instructions and pipe valve action sequence.

[0032] Furthermore, the collected oil pressure and pressure loss data are combined with the map to determine the cause of the short injection, generate a multi-objective optimization boosting plan, and perform pressure fluctuation monitoring, including:

[0033] Collect wellhead oil pressure data and trunk line pressure loss data, and combine them with pressure-flow graphs to determine the cause of underfilling;

[0034] Select overall boosting / local boosting / single-well boosting mode based on the cause of underfilling and generate a multi-objective optimized boosting plan;

[0035] Execute multi-objective optimization boosting plans and monitor pipeline pressure fluctuations in real time.

[0036] Furthermore, the energy consumption and efficiency indicators are collected for target comparison, the recovery rate change is analyzed to calculate the deviation rate, and the pipeline roughness parameters of the pipeline network digital twin model are corrected, including:

[0037] Collect water injection energy consumption data and water injection efficiency indicators after execution and compare them with target values;

[0038] Analyze changes in reservoir recovery and improvements in pipe network resistance loss, and calculate model prediction deviation rates;

[0039] The pipeline roughness parameters of the pipeline network digital twin model are corrected according to the predicted deviation rate.

[0040] Furthermore, the method further comprises:

[0041] The revised pipeline network digital twin model will be used in the next round of optimization decision-making process.

[0042] An intelligent optimization decision-making system for an oilfield water injection system, used to implement any of the above-mentioned intelligent optimization decision-making methods for an oilfield water injection system, comprising:

[0043] The data acquisition and modeling module is used to obtain the parameters of water injection station equipment, geology, and pipeline network for 3D modeling. After field verification and error compensation, a digital twin model of the pipeline network is generated.

[0044] A simulation analysis module is used to perform hydraulic and thermal simulation based on the digital twin model of the pipeline network, extract pressure distribution and pressure loss data, and generate pressure-flow maps;

[0045] The optimization decision module is used to obtain the injection plan and set constraints based on the pressure-flow map, optimize the water output distribution, match the pump library, and generate an efficient pump group ID combination plan;

[0046] The instruction generation module is used to analyze the high-efficiency pump group ID combination solution to query the characteristic curve, generate the inverter control signal and convert it into the booster pump start and stop instructions and pipe valve action sequence;

[0047] The execution monitoring module is used to collect oil pressure and pressure loss data and combine them with the map to determine the cause of short injection, generate a multi-objective optimization boosting plan, and perform pressure fluctuation monitoring;

[0048] The closed-loop correction module is used to collect energy consumption and efficiency indicators for target comparison, analyze the recovery rate changes to calculate the deviation rate, and correct the pipeline roughness parameters of the pipeline network digital twin model.

[0049] The key innovations of the present invention include:

[0050] (1) Integrate equipment parameters, pipe network geometry, and geological data to establish a digital twin model. By dynamically correcting the pipeline roughness parameters, the problem of insufficient accuracy of the traditional static model is solved, providing an accurate basis for subsequent optimization.

[0051] (2) Multi-objective differential evolution optimization algorithm. Aiming at the nonlinear and multi-constrained characteristics of water injection system, the differential evolution algorithm is improved: an adaptive mutation factor (dynamic adjustment of 0.3-1.0) is used to balance global / local search; hydraulic balance constraint verification is introduced to ensure the feasibility of the flow distribution scheme; a dynamic adjustment mechanism of weight coefficient (energy consumption weight 0.4-0.8) is used to adapt to different working conditions; the convergence speed is improved, and the optimization scheme reduces energy consumption.

[0052] (3) Intelligent diagnosis-execution closed-loop system; under-injection diagnosis method based on the fusion of pressure-flow map and real-time data; dynamic selection mechanism for three-level boosting strategies: overall / local / single well; 500ms response to control instructions under edge computing architecture; forming a complete closed loop of "perception-decision-execution-feedback".

[0053] The following are its main beneficial effects:

[0054] (1) This invention achieves precise perception of the operating status of the water injection system by constructing a high-precision digital twin model of the pipe network and combining it with real-time data on water injection pressure and flow. Compared with traditional manual experience-based decision-making, the use of a differential evolution algorithm to optimize water allocation reduces system energy consumption and improves water injection efficiency, effectively solving the core problem of high energy consumption and low efficiency of traditional systems.

[0055] (2) By integrating the equipment parameters of the water injection station, the pipeline network topology, and the reservoir geological data, a multi-objective optimization model is established. Under the premise of satisfying the node pressure constraints and the flow fluctuation threshold, the optimal solution with the minimum output power of each water injection station is calculated.

[0056] (3) Accurate diagnosis and dynamic optimization. The method for determining the cause of underfilling based on the pressure-flow graph can accurately identify underfilling problems such as formation blockage and pipe network resistance, improving diagnostic accuracy. Combined with the closed-loop correction mechanism, the model prediction error rate is reduced, enabling the system to have continuous self-optimization capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A flow chart of an intelligent optimization decision-making method for an oilfield water injection system provided in an embodiment of the present application;

[0058] Figure 2 A structural block diagram of an intelligent optimization decision-making system for an oilfield water injection system provided in an embodiment of the present application;

[0059] Figure 3 A flow chart of operating parameter optimization and pump start-up scheme optimization provided in an embodiment of the present application;

[0060] Figure 4 A flowchart for optimizing the short injection problem provided in an embodiment of the present application;

[0061] Figure 5 A pipeline network transformation optimization flow chart provided in an embodiment of the present application;

[0062] Figure 6 An overall workflow diagram is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0063] Example 1: Reference Figure 1 , is a flow chart of an intelligent optimization decision-making method for an oilfield water injection system provided by an embodiment of the present invention. The flow chart may include at least steps S100-S600:

[0064] S100: Obtain the water injection station equipment, geology, and pipeline network parameters for three-dimensional modeling. After field verification and error compensation, generate a digital twin model of the pipeline network.

[0065] S200 performs hydraulic and thermal simulation based on the digital twin model of the pipeline network, extracts pressure distribution and pressure loss data, and generates a pressure-flow map.

[0066] S300: Obtain the injection plan and set constraints on the pressure-flow map, optimize the water output distribution, match the pump library, and generate an efficient pump group ID combination plan.

[0067] S400, analyze the high-efficiency pump group ID combination solution to query the characteristic curve, generate the inverter control signal and convert it into the booster pump start and stop instructions and pipe valve action sequence.

[0068] S500: Collect oil pressure and pressure loss data and combine them with the map to determine the cause of short injection, generate a multi-objective optimization boosting plan, and perform pressure fluctuation monitoring.

[0069] S600 collects energy consumption and efficiency indicators for target comparison, analyzes recovery factor changes, calculates deviation rates, and corrects pipeline roughness parameters in the digital twin model of the pipeline network.

[0070] Step S100 at least includes steps S110-S130:

[0071] Construction method of digital twin model of pipeline network in intelligent optimization decision-making method of oilfield water injection system

[0072] S110, obtaining water injection station equipment parameters, water injection well geological parameters, and water injection pipe network geometric parameters, performing three-dimensional digital modeling, and obtaining an initial pipe network model.

[0073] Specifically, acquiring the water injection station equipment parameters includes collecting key performance parameters such as the rated power, head-flow characteristic curve, efficiency curve, and speed range of the water injection pump, as well as obtaining the model specifications and resistance coefficient of the valves and filters in the water injection station. Acquiring the geological parameters of the water injection wells includes collecting parameters such as the formation permeability, porosity, crude oil viscosity, and original formation pressure of the target oil reservoir, as well as the completion data, perforation parameters, and downhole string structure of each water injection well. Acquiring the geometric parameters of the water injection network includes measuring spatial geometric characteristics such as the inner diameter, wall thickness, length, direction, and elevation change of the pipeline, as well as the specific layout and quantity of pipe fittings such as elbows, tees, and reducers.

[0074] Furthermore, the 3D digital modeling process utilizes specialized pipe network modeling software, inputting the collected equipment parameters, geological parameters, and geometric parameters into the modeling system. Specifically, a 3D model of the water injection station equipment is first established, including precise modeling of key equipment such as the water injection pump unit, inlet and outlet pipelines, and valve groups. Next, a 3D topological structure of the water injection pipe network system is constructed, accurately reconstructing the spatial orientation and connection relationships of the pipes based on actual measurement data. Finally, the geological parameters of the injection wells are integrated to establish a 3D coupled model of the wellbore and formation. During this modeling process, parameters such as the pipe inner wall roughness and equipment resistance characteristics are initially assigned values ​​to form an initial pipe network model that incorporates hydraulic and thermal characteristics.

[0075] As can be understood, the initial pipeline network model includes the complete network topology, equipment performance parameters, and geological parameters, providing basic data support for subsequent simulation calculations. The model utilizes a parametric modeling approach, allowing all pipeline and equipment parameters to be dynamically adjusted based on measured data. This modeling process ensures that the model accuracy meets engineering calculation requirements, with pipeline length errors within one thousandth and equipment parameter errors within three percent.

[0076] S120. Extract the pipeline roughness coefficient and equipment resistance coefficient from the initial pipeline network model and compare and verify them with the measured data.

[0077] Specifically, extracting the pipeline roughness coefficient involves obtaining information such as the material properties, age, and surface treatment process of each pipe segment from the initial pipe network model, and determining the initial equivalent roughness value of each pipe segment based on industry standards. Extracting the equipment resistance coefficient involves obtaining the model and opening status of valves, filters, elbows, and other pipe fittings from the model, and then querying the corresponding local resistance coefficient based on fluid mechanics manuals. Furthermore, the characteristic curve of the water injection pump is digitized to extract the hydraulic loss coefficient under different operating conditions.

[0078] Furthermore, the measured data comparison and verification process involves deploying pressure sensors and flowmeters at key nodes in the on-site pipe network system to collect pressure distribution and flow data during actual operation. Specifically, representative operating conditions were selected for testing, recording the outlet pressure of each injection station, the pressure drop data for each pipe section, and the injection pressure of each injection well. The testing process covers different flow conditions, including maximum flow, minimum flow, and three intermediate flow points, to ensure comprehensive test data.

[0079] Understandably, the comparison and verification process compares and analyzes the pressure distribution calculated by the model with measured data. Specifically, under the same flow conditions, the deviation between the calculated and measured pressures at each node is compared. Any deviation exceeding 5% is marked as an abnormal pipe section. These abnormal pipe sections are then analyzed to determine if the settings for the pipe roughness coefficient or equipment resistance coefficient are appropriate. The comparison process is performed iteratively, adjusting model parameters to maintain the average deviation between the calculated and measured values ​​within 5%.

[0080] S130. Perform error compensation processing on the results of the measured data comparison and verification to generate a digital twin model of the pipeline network.

[0081] Specifically, the error compensation process involves dynamically correcting the pipeline roughness coefficient. Based on the comparison results of measured data, the roughness coefficient of pipe sections with large deviations is recalibrated. This calibration process considers practical factors such as pipeline age, corrosion conditions, and scaling, and uses an inversion algorithm to calculate the optimal roughness compensation value. For newly commissioned pipelines, the roughness compensation coefficient is controlled between 0.9 and 1.1; for pipelines in operation for more than five years, the compensation coefficient can be appropriately increased to a range of 0.8 to 1.3.

[0082] Furthermore, the error compensation process also includes optimizing and adjusting the equipment's resistance coefficient. For valves, filters, and other fittings, the local resistance coefficient is recalculated based on measured pressure drop data. This calculation takes into account the equipment's actual operating conditions, such as changes in valve opening and filter clogging. The injection pump's characteristic curve is modified, adjusting its shape based on measured head-flow data to ensure the model accurately reflects the pump's actual performance.

[0083] Understandably, the generation of the pipeline network digital twin model involves integrating all corrected parameters to create a highly accurate digital model. This model includes the complete pipeline network topology, corrected pipe roughness coefficients, optimized equipment resistance coefficients, and precise injection pump characteristic curves. The model supports real-time data access and dynamically reflects the operating status of the pipeline network system. The accuracy of the digital twin model has been rigorously verified, with pressure calculation errors exceeding 3% and flow calculation errors exceeding 2% across the full operating range, providing a reliable foundation for subsequent optimization decisions.

[0084] The pipeline network digital twin model directly connects with the subsequent S210 step, providing an accurate model foundation for coupled hydraulic and thermal simulation calculations. The pipe roughness parameters in the model are continuously updated in S630 based on operational data, forming a closed-loop optimization mechanism. The model's development process strictly adheres to engineering practice, with all parameters derived from field-measured data to ensure the model's credibility and practicality.

[0085] Step S200 at least includes steps S210-S230:

[0086] S210. Based on the digital twin model of the pipeline network, inject the real-time collected water injection pressure and water injection flow data to perform hydraulic-thermal coupling simulation calculations.

[0087] Specifically, the pipeline network digital twin model is derived from the error-compensated accurate model generated in step S130 and includes a corrected pipeline roughness coefficient, an optimized equipment resistance coefficient, and an accurate injection pump characteristic curve. The real-time collected water injection pressure data includes the outlet pressure of each water injection station, pressure monitoring values ​​at key nodes in the pipeline network, and the wellhead pressure of the injection wells. The data is collected at a frequency of seconds to ensure the timeliness of the data. The real-time collected water injection flow rate data includes the water output of each water injection station, the flow distribution of each branch of the pipeline network, and the injection volume of each injection well, and is measured by a high-precision electromagnetic flowmeter.

[0088] Furthermore, the hydraulic-thermodynamic coupling simulation calculation utilizes a professional fluid simulation software platform to import the pipeline network digital twin model into the simulation environment. Specifically, simulation boundary conditions are first set, including the injection station outlet pressure boundary, the injection well injection volume boundary, and the ambient temperature boundary. Fluid physical properties are then configured, including characteristic curves showing how the density and viscosity of the injected water change with temperature. The simulation calculation utilizes a transient solution method with a time step of 10 seconds to fully simulate the dynamic response of the pipeline network system.

[0089] Understandably, the hydraulic-thermal coupling simulation calculation process simultaneously considers the mutual influence of hydraulic characteristics and thermal characteristics. Specifically, the hydraulic calculation solves the pressure distribution of each node in the pipe network and the flow distribution of each pipe section, and the thermal calculation solves the influence of water temperature change on fluid viscosity and density. The calculation process uses an iterative solution method to perform alternating calculation of the hydraulic field and the thermal field at each time step until the convergence criterion is reached. The convergence criterion is set to a pressure change of less than 0.001 and a temperature change of less than 1% between adjacent iterations.

[0090] S220, extracting pipe network node pressure distribution and pipe section pressure loss data from the hydraulic-thermal coupling simulation calculation results for resistance loss analysis.

[0091] Specifically, the pipe network node pressure distribution data includes the pressure values of all water injection station outlet nodes, pipe network branch nodes and water injection well nodes, which are extracted and stored in time sequence. The pipe section pressure loss data includes the along-the-way pressure loss and local pressure loss of each pipe section, and the numerical value and change trend are calculated and recorded respectively. The data extraction process is realized by using an automatic script to parse the key parameters from the simulation result file and store them in a standardized format.

[0092] Further, the resistance loss analysis includes along-the-way resistance analysis and local resistance analysis. Specifically, the along-the-way resistance analysis evaluates the influence of pipe wall roughness on flow resistance by comparing the actual pressure loss of each pipe section with the theoretical calculation value. The local resistance analysis evaluates the influence of valve opening, filter blockage and other working condition changes on local resistance by comparing the pressure drop at each pipe fitting with the standard resistance coefficient calculation value. The analysis process establishes a resistance loss characteristic database to record the resistance change law under different working conditions.

[0093] Understandably, the resistance loss analysis also includes abnormal resistance identification and classification. Specifically, a resistance loss threshold is set, and when the pressure loss of a pipe section or pipe fitting exceeds 20% of the design value, it is marked as an abnormal resistance point. Feature extraction and classification are performed on the abnormal resistance points to distinguish resistance abnormalities caused by different reasons such as pipe fouling, valve failure or abnormal flow. The analysis results generate a resistance loss distribution map to visually display the resistance conditions of each region in the pipe network and provide a basis for subsequent optimization.

[0094] S230, generating a pressure-flow map according to the resistance loss analysis results and marking high pressure loss areas.

[0095] Specifically, the pressure-flow map is generated using a two-dimensional coordinate system, with the horizontal axis representing the network flow rate and the vertical axis representing the node pressure. The map includes multiple characteristic curves, each representing the pressure-flow relationship at key nodes in the network under different operating conditions. The curves are drawn using cubic spline interpolation to ensure smooth curves that accurately reflect the characteristics of the measured data. The map also includes isopressure loss lines, representing pressure-flow combinations at the same pressure loss value.

[0096] Furthermore, the high-pressure loss areas are labeled using a layered coloring method. Specifically, based on the resistance loss analysis results, the pipe network is divided into several calculation units, each assigned a different color depth based on its pressure loss value. Areas where the pressure loss exceeds 15% of the design value are labeled as yellow warning zones, and areas exceeding 30% are labeled as red alert zones. This labeling process utilizes an automated algorithm that automatically identifies and labels high-pressure loss areas based on preset rules.

[0097] As expected, the pressure-flow map also includes optimization recommendations. Specifically, near high-pressure loss areas, possible causes, such as pipe scaling and insufficient valve opening, are annotated, along with corresponding optimization recommendations, such as pipe cleaning and valve adjustment. Once generated, the map is linked to the network's digital twin model, supporting a click-to-query function to view detailed parameters and optimization recommendations for any annotated area. The map is output in a standardized format for direct use in subsequent optimization decision-making.

[0098] Step S300 at least includes steps S310-S330:

[0099] S310: Obtain the injection plan and pressure-flow map of the water injection wells, and set node pressure constraints and flow fluctuation thresholds.

[0100] Specifically, the process of acquiring the injection plan for water injection wells involves extracting the latest water injection development plan data from the oilfield development database. This data includes key parameters such as the numbers, location coordinates, target injection volume, injection pressure limit, and injection cycle for all injection wells within the block. The injection plan data undergoes an integrity check via a data verification module to ensure that the parameters of each injection well are complete and logically consistent. This verification process includes checking the matching of injection volume with formation permeability, and the compatibility of injection pressure with wellbore integrity. Once verified, the injection plan data is loaded into the optimization decision system's in-memory database for caching to improve data access speed.

[0101] Furthermore, the process of acquiring the pressure-flow map includes calling the latest map data generated in step S230 from the simulation analysis module. The map data contains the pressure values ​​of all key nodes in the pipe network system under different flow conditions, as well as the pressure loss data of each pipe section. The map adopts a hierarchical storage structure, with the base layer storing the original pressure-flow data, the middle layer storing the smoothed trend line data, and the application layer storing the analysis results marked with high pressure loss areas. The map data is transmitted to the optimization decision module via the data bus. During the transmission process, data compression technology is used to reduce the network load, and data integrity verification is performed at the receiving end.

[0102] Understandably, the setting process of the node pressure constraint adopts a hierarchical configuration method. In specific implementation, the minimum injection pressure of each water injection well is first determined according to the requirements of the reservoir engineering. This pressure value must ensure that the injected water can effectively displace crude oil and does not cause formation rupture. Then, based on the pipeline network topology, the minimum pressure requirements of each pipeline network branch node are reversely calculated from the water injection well node. The pressure constraint value is divided into three levels according to the importance of the node: the constraint deviation of key nodes (such as the outlet of the water injection station and the intersection of the main pipeline network) does not exceed plus or minus three percent; the constraint deviation of secondary nodes does not exceed plus or minus five percent; the constraint deviation of general nodes does not exceed plus or minus eight percent. The constraint value is displayed in a visual manner in the system configuration interface to support engineers to perform manual fine-tuning.

[0103] Furthermore, the flow fluctuation threshold setting process takes into account the dynamic characteristics of the water injection system. In specific implementation, the flow fluctuation patterns of each water injection station are first statistically analyzed based on historical operating data to determine a baseline fluctuation range. The threshold is then dynamically adjusted based on the importance of the current water injection task: during the critical water injection development phase, the threshold is tightened to plus or minus 2 percent; during routine water injection, the threshold is relaxed to plus or minus 5 percent; and during system commissioning or special operations, the threshold can be temporarily set to plus or minus 10 percent. The threshold setting algorithm incorporates a built-in self-learning function that automatically optimizes the threshold parameters based on feedback from system performance.

[0104] S320. Use the differential evolution algorithm to optimize the water output distribution of each water injection station and calculate the minimum output power of the water injection station.

[0105] Specifically, the initialization process of the differential evolution algorithm adopts an intelligent population generation strategy. This strategy first analyzes the total water demand of the current water injection task and the design capacity of each water injection station to determine the approximate flow distribution ratio. Then, based on this ratio, random perturbations are introduced to generate initial population individuals. The perturbation range is dynamically adjusted based on the historical operating data of each water injection station: for water injection stations with stable operation, the perturbation range is controlled within plus or minus 5%; for water injection stations that are newly commissioned or have recently undergone maintenance, the perturbation range is relaxed to plus or minus 10%. This initialization process ensures that the population individuals have sufficient diversity and can quickly converge to the high-quality solution area.

[0106] Furthermore, the implementation of the mutation operation adopts an adaptive parameter adjustment mechanism. During specific implementation, the algorithm dynamically adjusts the size of the mutation factor according to the evolutionary generation and population diversity index: a larger mutation factor (0.8-1.0) is used in the early stage of evolution to enhance the global search capability; in the later stage of evolution, the mutation factor (0.3-0.5) is gradually reduced to improve the local optimization accuracy. The selection of the differential vector adopts an elite strategy, and individuals with high fitness values ​​are preferentially selected as benchmark individuals. At the same time, in order to ensure the rationality of the mutation direction, the algorithm will check whether the flow distribution after mutation meets the minimum operating flow requirements of each water injection station, and correct the mutation individuals that do not meet the requirements.

[0107] Understandably, the execution of the crossover operation adopts a non-uniform crossover strategy. In specific implementation, the algorithm assigns different crossover probabilities according to the importance of the parameters: for the flow parameters of the main pump of the water injection station that directly affect the energy consumption of the system, a higher crossover probability (0.9) is adopted; for the operating parameters of the auxiliary equipment, a lower crossover probability (0.6) is adopted. The crossover process also takes into account the hydraulic balance constraints of the pipeline network to ensure that the new individuals generated after the crossover will not cause pressure imbalance in the pipeline network. To this end, the algorithm has a built-in hydraulic balance verification module to quickly simulate the flow distribution plan after the crossover and eliminate individuals that do not meet the pressure balance requirements.

[0108] Furthermore, the implementation of the selection operation adopts a multi-objective optimization strategy. During specific implementation, the algorithm not only considers the main goal of the total output power of the water injection station, but also takes the pressure balance of the pipeline network, equipment operation stability, etc. as secondary optimization goals. The fitness function adopts a weighted summation method to convert multiple optimization goals into a single fitness value. The weight coefficient is dynamically adjusted according to the current system operation status: during energy-sensitive periods, the weight of the power target is increased to 0.8; during periods with high system stability requirements, the weight of the pressure balance is increased to 0.6. The selection process also adopts an elite retention strategy to ensure that the best individuals in each generation can enter the next generation population.

[0109] S330, optimize the water output to match the pump characteristics database and generate an efficient pump group ID combination solution.

[0110] Specifically, the construction of the pump characteristic database adopts multi-source data fusion technology. The database integrates various information such as technical parameters from equipment manufacturers, on-site measured performance data, and historical operation records. The data storage adopts a time-series database architecture, which can fully record the performance changes of each pump group throughout its life cycle. Each pump group in the database is equipped with a unique ID identification, which contains basic information such as equipment number, installation location, and commissioning date. The performance parameter storage adopts a multi-dimensional array form to record key indicators such as head, efficiency, and power at different speeds and different flow rates, and the sampling density reaches one data point per 100 cubic meters / hour.

[0111] Furthermore, the implementation of the matching algorithm adopts a multi-stage screening strategy. The first stage is a rough screening, and based on the flow and head requirements obtained through optimization, the pump groups that obviously do not meet the requirements are quickly eliminated to narrow the candidate range. The second stage is a precise matching, and the interpolation algorithm is used to calculate the efficiency value of each candidate pump group at the target working point, and sort them by efficiency. The third stage is a combination optimization. For water injection stations that require multiple pumps to operate in parallel, the backpack algorithm is used to find the optimal pump group combination to ensure that the total flow meets the requirements while the overall efficiency is maximized. The algorithm also takes into account the cumulative operating time of the pump group and avoids excessive use of certain pump groups by introducing a time balancing factor.

[0112] Understandably, the combination scheme generation process involves multiple verification steps. First, a hydraulic characteristic verification is performed to ensure that the selected pump combination can provide sufficient head to meet the pipeline pressure requirements. Then, an electrical characteristic verification is performed to check whether the total power of the pump combination is within the allowable range of the substation capacity. Next, an equipment status verification is performed to exclude pumps that are under maintenance or have reached the end of their maintenance cycle. Finally, an economic evaluation is performed to calculate the operating cost differences of different combination schemes. Only after all the verification processes have passed will the system generate the final pump group ID combination scheme.

[0113] Furthermore, the solution output adopts a standardized format, which includes two parts: a detailed list of pump group operating parameters and a preliminary list of control instructions. The detailed list records detailed information such as the ID number, target flow, expected efficiency, and power consumption of each pump group; the preliminary list contains control parameters such as the start and stop sequence of the pump group, the target frequency of the inverter, and the valve opening. Before the solution is transmitted to the control system through the data interface, it must go through a manual confirmation link, and the on-duty engineer will conduct a final review of the key parameters. After the review is passed, the solution data is marked as executable and waits for the control system to call. At the same time, the system will automatically generate a solution description document to record the assumptions and constraint parameters of the optimization calculation, providing a basis for subsequent analysis.

[0114] Step S400 at least includes steps S410-S430:

[0115] S410, analyze the high-efficiency pump group ID combination scheme and query the pump group working characteristic curve.

[0116] Specifically, the analysis process of the high-efficiency pump group ID combination scheme adopts a hierarchical decoding technique. The scheme data is derived from the optimal pump group combination result generated in step S330, and includes structured information such as pump group number, target operating parameters, and expected performance indicators. The analysis process first performs format checking on the data packet to check data integrity and consistency, ensuring that all mandatory fields are complete and the values are within a reasonable range. Then, according to the preset analysis rules, the composite data structure is split into independent pump group records, each record containing complete equipment identification and operating parameters. The analysis process also includes a logical verification step to verify that the sum of the flow distribution of each pump group is consistent with the total water demand of the optimization scheme. If the deviation exceeds one percent, a data anomaly alarm will be triggered.

[0117] Further, the query process of the pump group working characteristic curve adopts a distributed retrieval mechanism. The query is based on a unified equipment identification system, and is associated with the characteristic database through the pump group ID. The database uses a multi-level index structure, with the first level index partitioned by pump station number, the second level index classified by equipment type, and the third level index sorted by commissioning time, ensuring query efficiency. The query process first locates the basic information of the target pump group, including manufacturer, model, rated parameters, etc.; then obtains the head-flow characteristic curve and efficiency-flow characteristic curve of the pump group at different speeds; and finally extracts recent maintenance records and performance test data for correcting the standard characteristic curve. The data precision of the characteristic curve reaches one sampling point per 50 cubic meters / hour, and the key working interval is encrypted to one sampling point per 20 cubic meters / hour.

[0118] Understandably, the dynamic correction process of the characteristic curve considers the equipment aging factor. Specifically, according to the cumulative operating hours of the pump group, an aging coefficient is applied to adjust the standard characteristic curve: for pump groups running more than 10,000 hours, the efficiency curve is adjusted downward by 3-5%; for pump groups running more than 20,000 hours, the efficiency curve is adjusted downward by 8-12%. The correction also considers the last three maintenance records, and the performance recovery coefficient of the pump group after major repair is set to 0.95-1.05. The corrected characteristic curve is stored in the temporary working area for subsequent control parameter calculation, and a correction record is generated and stored in the equipment history library.

[0119] S420, generate frequency value and valve opening control signals of the frequency converter according to the pump group working characteristic curve.

[0120] Specifically, the frequency value of the frequency converter is calculated by an iterative approximation algorithm. The algorithm takes the target flow rate as input and finds the optimal operating point on the modified pump group characteristic curve. First, the initial frequency is estimated based on the pipe network resistance characteristics, then the expected flow rate at this frequency is calculated by characteristic curve interpolation, and the deviation from the target flow rate is compared. The iterative process adjusts the frequency value until the flow rate deviation is less than one percent or reaches a maximum of twenty iterations. For parallel operation of the pump group, a collaborative optimization strategy is used to allocate the operating frequency of each pump to achieve the highest overall efficiency while meeting the total flow requirement. The calculation process also considers the output characteristics of the frequency converter, converting the theoretical frequency value into actual control signals with a resolution of 0.1 Hz.

[0121] Further, the generation of the valve opening control signal adopts a hierarchical regulation strategy. The strategy divides the valves into two categories: key regulating valves and auxiliary balancing valves according to the pressure distribution requirements of the pipe network. The opening of the key regulating valve is determined by accurate calculation, and the operating point that meets the target pressure drop is found on the characteristic curve, with an opening regulation accuracy of one percent. The opening of the auxiliary balancing valve uses a fuzzy control method, which automatically fine-tunes according to the upstream and downstream pressure difference, with an opening change step size of five percent. The control signal generation process includes a safety verification link to ensure that the valve opening does not cause the pipe network pressure to exceed the allowed range. When the calculated opening may cause pressure to exceed the limit, a re-optimization process is automatically triggered.

[0122] Understandably, the coordinated optimization process of the control signal considers the system response characteristics. Specifically, the timing arrangement of the control instructions is established: the frequency converter frequency adjustment is performed first, and after the pump group speed stabilizes, the valve opening is adjusted in stages; the key node valve adopts a slow regulation mode with a change rate controlled at two percent opening per second; the secondary node valve adopts a fast regulation mode with a change rate of up to five percent opening per second. The coordination scheme also includes an exception handling mechanism that automatically suspends the sending of subsequent instructions and starts the diagnosis program when an abnormal response of a device is detected.

[0123] S430, convert the control signal into booster pump start-stop instructions and pipe valve action sequence.

[0124] Specifically, the generation of the booster pump start-stop instruction uses a state transition model. The model analyzes the difference between the current booster pump operating state and the target state to generate the optimal state transition path. For the booster pump to be started, the instruction sequence includes preheating, lubrication inspection, soft start and other preparatory steps; for the booster pump to be stopped, the instruction sequence includes flow migration, slow speed load reduction, post-cooling and other finishing steps. The time parameters of the instructions are dynamically set according to the characteristics of the equipment, with the start-up time of large booster pumps set to three to five minutes and that of small booster pumps set to one to two minutes. The instructions also include interlock condition checks to ensure that the start and stop of the booster pump does not cause the associated equipment to overload.

[0125] Furthermore, the arrangement of the pipe and valve action sequence adopts a topological sorting algorithm. The algorithm determines the order of operation of each valve based on the pipe network structure diagram to avoid hydraulic shock. Valves on the critical path are operated first, and branch pipeline valves are adjusted later. Each valve action in the sequence is marked with the expected execution time window and the maximum allowed operation time. Timeout will trigger an alarm. For valves at important nodes, pressure checkpoints are also inserted in the sequence. The next operation must be performed only after the previous operation achieves the expected pressure effect. The sequence file adopts a standardized format, which contains fields such as operation step number, equipment identification, target value, execution time limit, etc.

[0126] Understandably, the control instructions are packaged and transmitted using an industrial communication protocol. Specifically, discrete start / stop instructions and continuous adjustment signals are packaged separately. Start / stop instructions are transmitted using the Modbus RTU protocol to ensure reliability, while adjustment signals are transmitted using the PROFIBUSDP protocol to ensure real-time performance. The transmission process incorporates a double-check mechanism: the transmitter generates an instruction verification code, and the receiver responds with an execution confirmation code. Key instructions implement a three-way handshake protocol to ensure accurate instruction delivery. All transmitted instructions are timestamped and serialized to support post-audit tracking. Instruction execution results are fed back to the monitoring system in real time, forming a complete control closed loop.

[0127] Step 500 at least includes steps S510-S530:

[0128] S510: Collect wellhead oil pressure data and trunk line pressure loss data, and determine the cause of underfilling by combining the pressure-flow graph.

[0129] Specifically, the collection process of the wellhead oil pressure data adopts a distributed sensor network. The network is composed of intelligent pressure transmitters installed at the wellhead of each water injection well. The collection frequency is set to once per second, and the measurement accuracy reaches plus or minus 0.1 MPa. The transmitter has a built-in self-diagnosis function, which can identify abnormal conditions such as sensor drift and signal interference, and automatically mark suspicious data. The data is transmitted to the central processing unit via the industrial bus. The transmission process adopts timestamp synchronization technology to ensure the timing consistency of the data at each node. The data receiving end implements three levels of verification: format verification checks the integrity of the data packet, range verification confirms the rationality of the value, and trend verification analyzes the continuity of data changes. The data that passes the verification is stored in the real-time database. Abnormal data triggers an alarm and starts the backup collection channel.

[0130] Furthermore, the calculation of the trunk pressure loss data is based on the pipeline network topology. Specifically, high-precision differential pressure transmitters are set at the starting point and end point of each trunk line to directly measure the total pressure drop of the pipe section. At the same time, based on the pressure-flow map generated in step S230, the theoretical pressure drop curve is extracted as a reference benchmark. The calculation process takes into account the impact of fluid temperature changes on viscosity, and uses a real-time temperature compensation algorithm to correct the measured values. For pipeline sections with complex branches, intermediate pressure monitoring points are added, and the pressure loss contribution rate of each sub-section is calculated by the segmented integration method. The pressure loss data and the wellhead oil pressure data are aligned in the time dimension to establish a complete pressure distribution space-time matrix.

[0131] Understandably, the determination of the cause of underfilling adopts a multi-evidence fusion method. Specifically, first, a feature library of underfilling is established, which contains characteristic patterns of twelve common causes such as formation blockage, pipe network scaling, and valve failure. Then, the pressure data collected in real time is pattern matched with the feature library to calculate the matching probability of each cause. The matching process introduces the pressure-flow map as a spatial constraint to exclude false positive results that do not conform to the hydraulic characteristics of the pipe network. For causes with a matching probability of more than 80%, a preliminary judgment report is generated; for cases where there are conflicts in the matching results, the expert rule engine is started for secondary analysis. The judgment result contains information in three dimensions: cause type, impact range, and confidence level, which provides a basis for the formulation of subsequent boosting plans.

[0132] S520: Select the overall pressurization / local pressurization / single-well pressurization mode according to the cause of underinjection, and generate a multi-objective optimized pressurization plan.

[0133] Specifically, the selection of the boosting mode adopts a decision tree algorithm. The input of the algorithm is the underinjection cause report generated by step S510, and the output is the optimal boosting strategy. The first-level node of the decision tree determines the scope of influence of underinjection: when more than seventy percent of the injection wells have insufficient pressure, the overall boosting mode is selected; when thirty to seventy percent of the injection wells are affected, the local boosting mode is selected; when less than thirty percent of the injection wells are affected, the single-well boosting mode is selected. The second-level node of the decision tree analyzes the cause type: for underinjection caused by pipeline resistance, local boosting is preferred; for underinjection caused by decreased formation absorption capacity, single-well boosting is preferred. The third-level node of the decision tree evaluates the implementation conditions: check the availability of backup boosting equipment, evaluate the power grid expansion space, and ensure that the selected mode has an implementation basis.

[0134] Furthermore, the generation of the multi-objective optimization boosting scheme adopts the Pareto front algorithm. The algorithm simultaneously considers three optimization objectives: maximizing the boosting effect, minimizing the increase in energy consumption, and minimizing the investment cost. In the specific implementation, the decision variable space is first established, including adjustable factors such as the installation location, rated power, and operating parameters of the boosting pump. Then, hundreds of candidate schemes are generated by the parameter scanning method, and the three-dimensional objective function value of each scheme is calculated. The calculation process calls the digital twin model of the pipeline network for rapid simulation to predict the pressure distribution improvement effect after the implementation of the scheme. Finally, the non-dominated sorting algorithm is used to screen out the Pareto optimal solution set for decision makers to select the final solution according to actual needs. The scheme document contains four parts: equipment configuration table, parameter setting table, expected effect table and emergency plan table, forming a complete implementation guidance document.

[0135] Understandably, the feasibility verification of the scheme involves multiple safeguards. Specifically, hydraulic safety verification ensures that the pressure in the pipeline network after pressurization does not exceed the material strength limit; electrical safety verification calculates that the added load does not exceed the substation capacity; and operational feasibility verification checks whether the operations required by the scheme are within the skill range of existing personnel. The verification process adopts a step-by-step progressive strategy, first conducting digital simulation verification, then conducting small-scale field tests, and finally full-scale implementation. The scheme document also identifies risk control points and formulates special monitoring measures for key areas such as high-voltage areas and key equipment to ensure that the pressurization process is safe and controllable.

[0136] S530: Execute a multi-objective optimization boosting plan and monitor pipeline network pressure fluctuations in real time.

[0137] Specifically, the step-by-step implementation of the boosting plan adopts a process control method. The method decomposes the plan into a preparation stage, a main implementation stage, and a system tuning stage. The preparation stage completes the equipment placement, pipeline modification, and safety isolation measures, which takes about four to eight hours; the main implementation stage starts and stops the boosting equipment in a predetermined order and adjusts the operating parameters, which takes about two to four hours; the system tuning stage fine-tunes the control parameters to stabilize the pipeline network, which takes about eight to twelve hours. Each stage sets clear completion standards and quality checkpoints, and only after passing the stage acceptance is it allowed to enter the next stage. The execution process adopts a two-person confirmation system, and all key operations must be independently confirmed by two authorized personnel before they can be implemented.

[0138] Furthermore, the monitoring of pressure fluctuations adopts a multi-scale analysis method. Specifically, second-level monitoring focuses on rapid pressure fluctuations to identify emergencies such as water hammer; minute-level monitoring tracks medium-speed pressure changes to evaluate the regulation effect; hour-level monitoring records long-term trends to analyze system stability. The monitoring network consists of fixed measuring points and mobile measuring points: fixed measuring points are installed at key nodes of the pipeline network to provide continuous monitoring data; mobile measuring points are temporarily deployed through portable devices to carry out key monitoring of specific areas. The data analysis adopts the wavelet transform method to decompose the pressure signal into components of different frequency bands, and evaluate their harmfulness and causes respectively. The monitoring results are displayed in real time on the large screen of the control center, and important alarm information is pushed to the mobile terminals of relevant personnel at the same time.

[0139] Understandably, the emergency handling of the abnormal fluctuations adopts a graded response mechanism. Specifically, the pressure fluctuation amplitude is divided into four levels: normal fluctuation (less than 5 percent), warning fluctuation (5 to 10 percent), alarm fluctuation (10 to 20 percent) and emergency fluctuation (greater than 20 percent). For warning fluctuations, the system automatically records the event and prompts attention; for alarm fluctuations, the system starts the preset adjustment program to try to recover; for emergency fluctuations, the system immediately executes the emergency shutdown procedure and notifies the emergency team. The response mechanism includes comprehensive communication guarantee measures to ensure that the alarm information is delivered to all relevant positions within 30 seconds and the emergency instructions are confirmed and fed back within 60 seconds. A detailed event analysis report is generated after each emergency response to optimize the subsequent boosting scheme design.

[0140] Step S600 at least includes steps S610-S630:

[0141] S610: Collect the water injection energy consumption data and water injection efficiency index after execution and compare them with the target values.

[0142] Specifically, the collection process of the water injection energy consumption data adopts a combination of a smart meter network and a distributed metering system. The smart meter is installed on the power supply circuit of each water injection pump, and records electrical parameters such as voltage, current, and power factor at a frequency of once per minute, and transmits the data to the energy consumption monitoring center through power carrier communication technology. The distributed metering system sets electromagnetic flowmeters at the total water outlet of the water injection station and each branch distribution point to measure the water injection volume in real time, with a measurement accuracy of plus or minus 0.5 percent. The data acquisition system has a built-in clock synchronization module to ensure that the energy consumption data and flow data are strictly aligned in the time dimension. After the raw data is filtered for outliers, filled with missing values ​​and unified in units, it is stored in a time series database to form a complete energy consumption data set.

[0143] Furthermore, the calculation of the water injection efficiency index adopts a multi-level evaluation method. Specifically, the first level calculates the equipment efficiency, and calculates the efficiency of a single machine based on the input power and effective water power of the pump group; the second level calculates the system efficiency, and considers the overall energy efficiency after the pipeline transmission loss; the third level calculates the development efficiency, and evaluates the water injection development effect in combination with the reservoir dynamic data. The calculation process calls the pump group characteristic data generated by step S330 and the pipeline pressure loss data obtained by step S230 to ensure the accuracy of the calculation results. The efficiency index is generated in the form of a daily report, which contains statistical features such as maximum value, minimum value, average value and trend analysis, providing a comprehensive basis for subsequent comparison.

[0144] Understandably, the target value comparison process adopts a dynamic deviation analysis method. The method first extracts the energy consumption baseline value and efficiency target value corresponding to the current water injection task from the optimization target library. The baseline value is determined based on the design parameters in the water injection development plan and the historical optimal operation data. Then, the absolute deviation and relative deviation between the actual value and the target value are calculated, and the deviation is divided into three levels according to the size of the deviation: normal range (less than 5%), attention range (5% to 10%), and abnormal range (greater than 10%). The comparison results are displayed in a visual form, abnormal deviation items are marked in red, and the change curves of related operating parameters are automatically associated to assist in analyzing the cause of the deviation.

[0145] S620. Analyze the changes in reservoir recovery rate and the improvement of pipeline network resistance loss, and calculate the model prediction deviation rate.

[0146] Specifically, the analysis of changes in reservoir recovery rate is based on a method that combines dynamic monitoring data with numerical simulation. The monitoring data includes injection pressure and injection volume of water injection wells, and production dynamic indicators such as liquid production and water content of production wells, and the collection frequency is once a day. The numerical simulation calls the reservoir geological model, inputs the current water injection parameters, and predicts the recovery rate change trend in the next three months. The analysis method establishes a correlation curve between the actual recovery rate and the predicted recovery rate, calculates the correlation coefficient and deviation amplitude between the two, and evaluates the degree of consistency between the water injection effect and the expectation. The analysis process takes into account the influence of reservoir heterogeneity and fluid phase changes, and uses multiple regression technology to eliminate the influence of interference factors.

[0147] Furthermore, the evaluation of the improvement of the resistance loss of the pipeline network adopts the time-space comparison method. The method extracts the pipeline pressure loss data before and after optimization from the resistance loss database generated in step S220, and establishes a time series comparison chart and a spatial distribution comparison chart. The evaluation indicators include three dimensions: the total resistance loss reduction rate, the reduction ratio of the high pressure loss area, and the resistance distribution uniformity index. The evaluation process calls the pipeline digital twin model established in step S130 for simulation verification to ensure the reliability of the evaluation results. For pipe sections where the improvement effect is not obvious, the system automatically marks them as key focus objects, suggesting that additional transformation measures may be required.

[0148] Understandably, the model prediction deviation rate is calculated using a weighted synthesis method. The method first determines three types of key deviation indicators: energy consumption prediction deviation, pressure distribution prediction deviation, and recovery rate prediction deviation. Then, a weight coefficient is assigned according to the importance of each indicator, with the energy consumption prediction deviation weight being 0.4, the pressure distribution prediction deviation weight being 0.3, and the recovery rate prediction deviation weight being 0.3. The calculation process uses a sliding time window technique to calculate the average deviation rate over the past thirty days to reflect the current prediction capability of the model. The deviation rate result is expressed in percentages, with scores of 85 or above being excellent, 70 to 85 being good, and scores below 70 requiring model correction.

[0149] S630. Correct the pipeline roughness parameters of the pipeline network digital twin model according to the predicted deviation rate.

[0150] Specifically, the correction parameters are determined using an inverse optimization algorithm. The algorithm uses the minimum model prediction deviation rate as the objective function and the pipeline roughness coefficient as the decision variable to search for the optimal parameter combination within the allowable correction range. The correction range is determined based on the pipeline material and service life. The correction range is allowed to be plus or minus ten percent for newly commissioned steel pipes, and plus or minus twenty percent for pipelines that have been in operation for more than five years. The optimization process adopts a hybrid strategy that combines genetic algorithms with local search to ensure convergence speed while seeking global optimization. The algorithm sets the maximum number of iterations to two hundred times, and terminates the calculation early when the fitness value changes by less than one percent.

[0151] Furthermore, the update of the model parameters adopts a progressive adjustment strategy. Specifically, the total correction amount is decomposed into ten adjustment steps, and each step is implemented at intervals of twenty-four hours. After each parameter update, the system automatically runs a verification test to compare the deviation changes between the model prediction value and the actual monitoring value. If the deviation improvement effect meets expectations, the next round of adjustment will continue; if the deviation expands or the fluctuation intensifies, the parameters will be rolled back to the previous step and re-optimized. The strategy effectively controls the correction risk and avoids model inaccuracy due to parameter mutations. The update process records complete version information, including modification time, modification personnel, modification content and verification results, to form a traceable model maintenance file.

[0152] Understandably, the verification of the correction effect adopts the cross-validation method. The method divides the monitoring data into two parts: a training set and a test set. The training set is used for parameter correction, and the test set is used to verify the correction effect. The verification indicators include three dimensions: the average absolute error of pressure prediction, the correlation coefficient of flow prediction, and the relative deviation of energy consumption prediction. The verification process not only checks the overall accuracy improvement, but also focuses on analyzing the prediction accuracy of special parts such as high pressure loss areas and key nodes. After the verification is passed, the system automatically generates a model correction report, updates the error compensation parameters of step S130, and deploys the corrected model to the production environment, completing a complete closed loop from data acquisition to model optimization.

[0153] Example 2: Figure 2 FIG. 1 shows a block diagram of an intelligent optimization decision system for an oilfield water injection system according to an embodiment of the present invention. Figure 2 As shown, the structure may include:

[0154] The data acquisition and modeling module 10 is used to obtain the parameters of the water injection station equipment, geology, and pipeline network for three-dimensional modeling, and generate a digital twin model of the pipeline network after field verification and error compensation. This module connects various sensors of the water injection station, pipeline network, and oil wells through industrial Internet of Things technology to collect equipment parameters, geological parameters, and pipeline network geometric parameters in real time. Specifically, the module includes a three-dimensional scanning unit, a data verification unit, and a model correction unit. The three-dimensional scanning unit uses lidar and photogrammetry technology to establish an initial three-dimensional model of the water injection system; the data verification unit identifies parameter deviations by comparing and analyzing the on-site measured data with the model prediction value; the model correction unit applies an error compensation algorithm to dynamically adjust key parameters such as pipeline roughness to ensure that the accuracy error of the digital twin model is controlled within 5%. The module is directly connected to the simulation analysis module to provide an accurate model basis for hydraulic calculations.

[0155] The simulation analysis module 20 is used to perform hydraulic and thermal simulation based on the digital twin model of the pipeline network, extract pressure distribution and pressure loss data, and generate a pressure-flow map. This module builds a multi-physics field coupling simulation engine based on the principles of computational fluid dynamics. During specific implementation, the module obtains the digital twin model from the data acquisition and modeling module, injects real-time operation data, and performs transient simulation calculations. The module includes a preprocessing unit, a solver unit, and a post-processing unit. The preprocessing unit completes the grid division and boundary condition setting; the solver unit uses parallel computing technology to solve the Navier-Stokes equations; the post-processing unit extracts pressure distribution and pressure loss data and generates a pressure-flow map marked with high pressure loss areas. The map is transmitted to the optimization decision module through a standardized interface to provide constraints for water distribution optimization.

[0156] The optimization decision module 30 is used to obtain the injection plan and set constraints from the pressure-flow map, optimize the water output distribution, match the pump library, and generate an efficient pump group ID combination plan. This module uses a hybrid intelligent algorithm, combining mathematical optimization and expert experience to make decisions. Specifically, the module obtains the pressure-flow map from the simulation analysis module, calls the injection well injection plan from the production database, and establishes a multi-objective optimization model. The module includes a constraint management unit, an algorithm solving unit, and a solution evaluation unit. The constraint management unit sets the node pressure constraint and flow fluctuation threshold; the algorithm solving unit applies the improved differential evolution algorithm to find the optimal water distribution plan; the solution evaluation unit selects the best pump group combination through cost-benefit analysis. The optimization plan generated by the module is transmitted to the instruction generation module via the data bus to realize the transition from decision-making to execution.

[0157] The instruction generation module 40 is used to parse the high-efficiency pump group ID combination scheme to query the characteristic curve, generate the inverter control signal and convert it into the booster pump start and stop instruction and pipe valve action sequence. The module has a built-in equipment characteristic database to store the working curves and performance parameters of various pump groups. During specific implementation, the module parses the high-efficiency pump group ID combination scheme sent by the optimization decision module and queries the characteristic curve of the corresponding equipment. The module includes an instruction calculation unit, a safety verification unit and a protocol conversion unit. The instruction calculation unit calculates the optimal output frequency of the inverter based on the target flow and head requirements; the safety verification unit ensures that the control parameters are within the allowable range of the equipment; the protocol conversion unit converts the standardized instructions into the communication protocol of the specific control system. The instructions generated by the module are sent to the execution monitoring module through the industrial network to achieve precise control.

[0158] The execution monitoring module 50 is used to collect oil pressure and pressure loss data and combine them with the map to determine the cause of underfilling, generate a multi-objective optimization boosting plan, and perform pressure fluctuation monitoring. This module is deployed on-site at the oil field and is directly connected to actuators such as boosting pumps and regulating valves. Specifically, the module receives control instructions from the instruction generation module and executes the boosting plan in steps. The module includes a task scheduling unit, a status monitoring unit, and an exception handling unit. The task scheduling unit arranges the execution order of instructions to avoid hydraulic shock; the status monitoring unit collects pipeline pressure fluctuation data in real time; and the exception handling unit activates the emergency plan when a pressure anomaly is detected. The module feeds back the execution effect data to the closed-loop correction module in real time, forming a complete closed loop of decision-making-execution-feedback.

[0159] The closed-loop correction module 60 is used to collect energy consumption and efficiency indicators for target comparison, analyze the changes in recovery rate to calculate the deviation rate, and correct the pipeline roughness parameters of the digital twin model of the pipeline network. This module continuously improves the accuracy of the model by analyzing historical operation data. During specific implementation, the module collects water injection energy consumption, efficiency indicators and reservoir dynamic data to evaluate the actual effect of the optimization plan. The module includes a performance evaluation unit, a deviation analysis unit and a parameter adjustment unit. The performance evaluation unit calculates the deviation between each indicator and the target; the deviation analysis unit identifies the source of the model prediction error; and the parameter adjustment unit corrects the parameter settings of the digital twin model. The correction results of the module are fed back to the data acquisition and modeling module to start a new round of optimization cycle.

[0160] In another embodiment, Figure 3The figure shows a flow chart for operating parameter optimization and pump start-up scheme optimization provided by this embodiment. This embodiment is implemented based on the complete workflow of the operating parameter optimization and pump start-up scheme optimization modules. After the system is started, multi-source data collection is first performed to obtain basic information such as station production data. Subsequently, the two core links of data processing and system optimization are executed in parallel: in the data processing link, a three-level progressive data cleaning strategy is adopted. First, significant outliers are removed by combining Z-scores and IQR for cleaning, and then a clustering algorithm is applied to clean and remove local outliers. Finally, abnormal data that does not conform to physical laws is screened out based on the trend law of the characteristic curve. The cleaned data is combined with the equipment factory coefficient to generate a standardized discrete data set. After single pump curve fitting correction, a high-precision pump characteristic database is established, and different pump combinations are analyzed to construct a combined characteristic curve. In the system optimization link, a basic hydraulic model is established based on the simulation calculation of the water injection network. After setting the water volume optimization target, an improved differential evolution algorithm is used for global optimization calculation. The algorithm ensures the feasibility of the solution through adaptive mutation factors and constraint verification mechanisms. The water allocation plan generated by the optimized calculation interacts in real time with the pump characteristics database, intelligently matching the optimal pump combination based on actual water temperature and head requirements. These two links form a closed-loop connection between data processing and plan generation, ultimately outputting a complete pump startup plan, including execution instructions such as inverter control parameters, valve opening, and pump start and stop sequences. This entire process coordinates the various modules through a data bus, ensuring high consistency between optimized operating parameters and the planned pump startup plan.

[0161] In another embodiment, Figure 4The figure shows a flowchart for optimizing the under-injection problem provided by this embodiment. This embodiment is implemented based on the closed-loop workflow of the under-injection problem optimization module. After the system is started, the water injection wellhead pressure, flow data and pipeline pressure limit data are first collected. The collected data is subjected to real-time quality inspection through the data validity verification module. When abnormal data is found, the supplementary collection mechanism is automatically triggered or the pressure transmitter calibration program is started. For valid data that passes the verification, the system performs a two-level pressure loss judgment: first, it judges whether the main line pressure loss exceeds the preset threshold (1.0MPa), and then verifies whether the actual water injection pressure and water injection volume at the wellhead meet the design requirements. When it is determined to be an under-injection condition, the system automatically starts the multi-source data acquisition program to obtain geological dynamic parameters including reservoir permeability, well group connectivity, and formation water absorption index, and constructs a multi-dimensional decision domain through the big data analysis engine. Based on the decision analysis results, differentiated processing modes such as overall pressure increase, local pressure regulation or single well injection are intelligently selected, and supporting transformation plans are generated for infrastructure problems such as pipeline scaling and insufficient pipe diameter. During the implementation phase, digital twin technology was used for prior simulation verification. Once the feasibility of the solution was ensured, execution instructions were issued through the edge control unit. During execution, wellhead pressure fluctuations and injection volume trends were monitored in real time. Injection parameters were dynamically adjusted using intelligent prediction algorithms. Finally, a multi-metric evaluation system (including injection efficiency, oil production increase, and input-output ratio) was used to verify the effectiveness of the measures in a closed loop. Optimization experience was fed back to the system knowledge base, forming a continuously self-improving intelligent optimization mechanism. The entire process seamlessly connects all links through a data bus and event-driven architecture, ensuring optimization of the entire process from diagnosis to treatment of underinjection issues.

[0162] In another embodiment, Figure 5The figure shows a pipeline network transformation and optimization flow chart provided by this embodiment. The closed-loop workflow of this embodiment based on the pipeline network transformation and optimization module is specifically implemented as follows: After the system is started, the pipeline network node parameters (including pressure, flow, temperature, etc.) are first monitored in real time, and the collected data is filtered for outliers and compensated for missing values ​​through the quality verification module. The processed monitoring data is input into the digital twin pipeline network model, and hydraulic and thermal coupling simulation analysis is performed. By comparing the deviation between the design parameters and the actual operating parameters, the system automatically identifies four typical problems such as unreasonable layout, pipe diameter mismatch, internal leakage of pipe fittings and leakage points. Based on the identification results, the system calls the pipeline network layout diagram, reservoir distribution data and water injection demand parameters to construct a three-dimensional connectivity model, and uses the connectivity analysis algorithm based on graph theory to calculate the topological importance and hydraulic contribution of each pipe section. Based on the analysis results, renovation priorities are determined, and a comprehensive renovation plan is intelligently generated, encompassing layout adjustments, pipe diameter optimization, and fitting replacement. For the main pipeline network, a genetic algorithm-based pipe diameter optimization method is used to minimize renovation costs while meeting flow and pressure constraints. For localized pipe sections, resistance loss analysis results are combined to determine which redundant pipe sections need to be removed or which high-resistance fittings need to be replaced. During the implementation phase, BIM technology is used for construction simulation, and construction drawings and bills of materials are generated after feasibility is verified. Following the completion of the renovation, a 72-hour trial run is conducted, verifying the effectiveness of the renovation through real-time monitoring of pressure fluctuations and flow distribution. Operating data is compared and analyzed against expected targets, and a secondary optimization process is initiated for pipe sections that do not meet standards, forming a complete closed-loop optimization system of "monitoring-analysis-renovation-verification." The entire process is interconnected through the Industrial Internet of Things platform, ensuring that the renovation plan closely matches actual operational needs.

[0163] In another embodiment, Figure 6The figure shows a flowchart based on the overall workflow provided by this embodiment. The closed-loop optimization process based on the overall workflow diagram of this embodiment is specifically implemented as follows: After the system is started, the water injection well pressure, water injection station operating parameters and pipeline flow data are first collected in real time through the distributed sensor network. The collected data undergoes a three-level data cleaning process, including outlier removal, missing value compensation and unit standardization, to form a standardized data set. The cleaned data is synchronously transmitted to three core analysis modules: in the water injection station decision analysis module, operating parameter optimization and equipment reliability evaluation are performed based on the digital twin model to generate an optimization plan including variable frequency control parameters and pump group start and stop sequences; in the under-injection problem cause analysis module, through pressure-flow map matching and big data analysis technology, the causes of under-injection such as formation blockage and pipeline resistance are accurately identified, and pressurization measures are intelligently selected; in the pipeline network transformation research module, graph theory algorithms are used to analyze the pipeline network topology, and transformation suggestions are proposed for pipe performance and pipe diameter matching problems. After the optimization solutions generated by each module are collaboratively integrated by the dynamic registration control unit, the solution evaluation phase begins. This phase verifies technical feasibility through hydraulic simulation and assesses economic viability using cost models. The final output is a comprehensive strategy plan with implementation priorities. During implementation, a real-time monitoring system continuously collects operational data and dynamically adjusts control parameters using a deviation analysis algorithm, forming a complete closed loop of "data collection - analysis and decision-making - solution generation - implementation feedback." The entire process utilizes a microservices architecture to decouple modules and ensures real-time data exchange through message-based middleware, ultimately achieving the technical benefits of improving water injection system operational efficiency and reducing energy consumption.

[0164] Obviously, the embodiments described above are only some of the embodiments of the present application, rather than all of the embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application has been described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions described in the aforementioned specific embodiments, or to make equivalent replacements for some of the technical features therein. Any equivalent structure made using the contents of the present application specification and the accompanying drawings, directly or indirectly used in other related technical fields, is also within the scope of patent protection of the present application.

Claims

1. An intelligent optimization decision-making method for an oilfield water injection system, characterized in that: include: Obtain the water injection station equipment, geology, and pipeline network parameters for 3D modeling. After field measurement verification and error compensation, a digital twin model of the pipeline network is generated. Perform hydraulic and thermal simulation based on the digital twin model of the pipe network, extract pressure distribution and pressure loss data, and generate pressure-flow maps; Obtain the injection plan and set constraints on the pressure-flow map, optimize the water output distribution, match the pump library, and generate an efficient pump group ID combination plan; Analyze the high-efficiency pump group ID combination solution to query the characteristic curve, generate the inverter control signal and convert it into the booster pump start and stop instructions and pipe valve action sequence; Collect oil pressure and pressure loss data and combine them with maps to determine the cause of short injection, generate a multi-objective optimization boosting plan, and perform pressure fluctuation monitoring; Collect energy consumption and efficiency indicators for target comparison, analyze recovery factor changes to calculate deviation rates, and correct pipeline roughness parameters in the digital twin model of the pipeline network.

2. The intelligent optimization decision-making method according to claim 1, characterized in that: The acquisition of water injection station equipment, geology, and pipe network parameters for three-dimensional modeling, and the generation of a pipe network digital twin model through field verification and error compensation, include: Obtain the equipment parameters of the water injection station, the geological parameters of the water injection wells, and the geometric parameters of the water injection network, perform three-dimensional digital modeling, and obtain the initial network model; Extract the pipeline roughness coefficient and equipment resistance coefficient from the initial pipeline network model and compare and verify them with the measured data; Error compensation is performed on the results of the measured data comparison and verification to generate a digital twin model of the pipeline network.

3. The intelligent optimization decision-making method according to claim 1, characterized in that: The hydraulic-thermal simulation is performed based on the pipe network digital twin model to extract pressure distribution and pressure loss data and generate a pressure-flow map, including: Based on the digital twin model of the pipeline network, the real-time collected water injection pressure and flow rate data are injected to perform hydraulic-thermal coupling simulation calculations; Extract the pressure distribution of pipeline nodes and the pressure loss data of pipe sections from the hydraulic-thermal coupling simulation results to perform resistance loss analysis; Generate a pressure-flow map based on the resistance loss analysis results and mark the high pressure loss areas.

4. The intelligent optimization decision-making method according to claim 1, characterized in that: The method of obtaining the injection scheme and setting constraints on the pressure-flow map, optimizing the water output distribution and matching the pump library, and generating an efficient pump group ID combination scheme includes: Obtain injection well allocation plans and pressure-flow maps, and set node pressure constraints and flow fluctuation thresholds; The differential evolution algorithm is used to optimize the water output distribution of each injection station and calculate the minimum output power of the injection station.

5. The intelligent optimization decision-making method according to claim 4, characterized in that: The step of optimizing the water output distribution of each water injection station by using the differential evolution algorithm and calculating the minimum output power of the water injection station also includes: Generate an efficient pump group ID combination solution based on the optimized water output matching pump characteristic database.

6. The intelligent optimization decision-making method according to claim 1, characterized in that: The method of parsing the high-efficiency pump group ID combination scheme to query the characteristic curve, generate the inverter control signal and convert it into the booster pump start and stop instructions and pipe valve action sequence, including: Analyze the ID combination scheme of high-efficiency pump groups and query the working characteristic curve of the pump groups; Generate inverter frequency value and valve opening control signal according to the pump group working characteristic curve; The control signal is converted into booster pump start and stop instructions and pipe valve action sequence.

7. The intelligent optimization decision-making method according to claim 1, characterized in that: The collected oil pressure and pressure loss data are combined with the map to determine the cause of the short injection, generate a multi-objective optimization boosting plan, and perform pressure fluctuation monitoring, including: Collect wellhead oil pressure data and trunk line pressure loss data, and combine them with pressure-flow graphs to determine the cause of underfilling; Select overall boosting / local boosting / single-well boosting mode based on the cause of underfilling and generate a multi-objective optimized boosting plan; Execute multi-objective optimization boosting plans and monitor pipeline pressure fluctuations in real time.

8. The intelligent optimization decision-making method according to claim 1, characterized in that: The collected energy consumption and efficiency indicators are compared with the target, the recovery rate change is analyzed to calculate the deviation rate, and the pipeline roughness parameters of the pipeline network digital twin model are corrected, including: Collect water injection energy consumption data and water injection efficiency indicators after execution and compare them with target values; Analyze changes in reservoir recovery and improvements in pipe network resistance loss, and calculate model prediction deviation rates; The pipeline roughness parameters of the pipeline network digital twin model are corrected according to the predicted deviation rate.

9. The intelligent optimization decision-making method according to claim 1, characterized in that: The method further comprises: The revised pipeline network digital twin model will be used in the next round of optimization decision-making process.

10. An intelligent optimization decision-making system for an oilfield water injection system, used to implement the intelligent optimization decision-making method for an oilfield water injection system according to any one of claims 1 to 9, characterized in that: include: The data acquisition and modeling module is used to obtain the parameters of water injection station equipment, geology, and pipeline network for 3D modeling. After field verification and error compensation, a digital twin model of the pipeline network is generated. A simulation analysis module is used to perform hydraulic and thermal simulation based on the digital twin model of the pipeline network, extract pressure distribution and pressure loss data, and generate pressure-flow maps; The optimization decision module is used to obtain the injection plan and set constraints based on the pressure-flow map, optimize the water output distribution, match the pump library, and generate an efficient pump group ID combination plan; The instruction generation module is used to analyze the high-efficiency pump group ID combination solution to query the characteristic curve, generate the inverter control signal and convert it into the booster pump start and stop instructions and pipe valve action sequence; The execution monitoring module is used to collect oil pressure and pressure loss data and combine them with the map to determine the cause of short injection, generate a multi-objective optimization boosting plan, and perform pressure fluctuation monitoring; The closed-loop correction module is used to collect energy consumption and efficiency indicators for target comparison, analyze the recovery rate changes to calculate the deviation rate, and correct the pipeline roughness parameters of the pipeline network digital twin model.

Images