Oilfield injection-production parameter integrated optimization matching system and regulation and control method

Abstract

This invention relates to the field of dynamic control in oilfield development, and discloses an integrated optimization and matching system and control method for oilfield injection and production parameters. The system acquires data on the operation of injection wells, production wells, formations, equipment, and operational measures within a target block, forming a basic injection and production data table. It identifies control events caused by changes in injection volume, injection pressure, stratified water distribution ratio, production volume, or lift parameters, and eliminates operational measures, equipment anomalies, and well shutdown interference to obtain effective control events. Based on these effective control events, it constructs time-delay response samples, generating an injection-production response spectrum that includes connectivity strength, response lag time, response stability, and water-cut risk coefficient. Based on the injection-production response spectrum and the current production status, it predicts the production response of candidate injection-production parameter combinations, determines executable injection-production parameter combinations through safety constraint projection, generates control commands, and then corrects the injection-production response spectrum, prediction model, and safety constraint parameters according to the actual response.

Description

Technical Field

[0001] This invention relates to the field of dynamic regulation and control in oilfield development, specifically to an integrated optimization and matching system and regulation method for oilfield injection and production parameters. Background Technology

[0002] After an oilfield enters the medium-to-high water-cut development stage, the dynamic response relationship between injection wells, production wells, and corresponding formations changes with development time, formation pressure, inter-well connectivity, and the degree of water cut advancement. On-site adjustments to the injection-production relationship of the well group are typically required based on parameters such as injection volume, injection pressure, stratified water distribution ratio, fluid production volume, pump frequency, stroke frequency, dynamic fluid level, bottom hole flowing pressure, and water cut to maintain a reasonable injection-production balance and control the rate of water cut increase. In current production management, injection-production control is largely based on single-well production curves, well group injection-production ratios, manual experience analysis, or phased reservoir numerical simulation results. When production wells experience a decrease in production, an increase in water cut, abnormal dynamic fluid level, or pressure deviation from the target range, technicians then determine whether the corresponding injection wells need increased or decreased injection, or whether the stratified water distribution ratio needs adjustment, and coordinate adjustments to the production well lift parameters or fluid production volume.

[0003] However, in multi-well groups, multi-layered systems, and heterogeneous reservoirs, changes in water injection parameters typically have a time lag in their impact on production at the oil production end. Furthermore, the influence of the same injection well on different production wells and different formations varies. Adjusting production solely based on static injection-production ratios or single-well production curve changes makes it difficult to accurately determine the true injection-production relationship. This can easily lead to misinterpreting short-term fluctuations as effective responses and hinders timely identification of high water-cut surges, inter-layer interference, and ineffective water injection risks. Simultaneously, oilfield production data is also affected by factors such as acidizing, fracturing, water shut-off, well washing, pump inspections, well shutdowns, equipment malfunctions, and metering errors. Directly incorporating these production fluctuations into the injection-production parameter optimization process can easily misinterpret changes in production fluid, water cut, and pressure caused by operational measures or equipment malfunctions as responses after water injection adjustments, leading to subsequent injection-production parameter matching results deviating from actual reservoir dynamics.

[0004] While some existing intelligent optimization methods can incorporate machine learning, surrogate models, or optimization algorithms to calculate injection-production parameters, they typically focus on predicting indicators such as production and water cut using historical production data or numerical simulation samples. They fail to adequately distinguish between real control events and disruptive events caused by measures, and lack a unified expression for response lag time, response stability, and formation water cut risk after water injection adjustments. Furthermore, the optimized injection-production parameter results are constrained in field execution by factors such as the upper limit of injection pressure, the adjustment range of the stratified water distributor, the operating range of the lifting equipment, the variation range of the produced fluid volume, the target range of the injection-production ratio, the upper limit of water cut, and the range of formation pressure maintenance. If only production improvement or predicted revenue is used as the optimization objective, control schemes may be unenforceable by field equipment or pose risks such as water cut surges, pressure anomalies, and excessively low bottomhole flowing pressure. Therefore, it is necessary to propose an integrated optimization and matching scheme for oilfield injection-production parameters that can eliminate interfering data, identify the real injection-production time-lag response relationship, and combine machine learning prediction, safety constraint projection, and feedback correction. Summary of the Invention

[0005] To address the problems in existing injection and production control systems, such as the difficulty in eliminating interference from operational measures and equipment anomalies, the time lag in oil production response after water injection adjustments, and the lack of on-site safety constraints for candidate control parameters, this invention provides an integrated optimization and matching system and control method for oilfield injection and production parameters.

[0006] The above-mentioned technical objective of the present invention is achieved through the following technical solution: An integrated optimization and matching system for oilfield injection and production parameters includes: The system includes a data acquisition module, a regulation event identification module, a time-delay response sample construction module, an injection-machining response map generation module, a machine learning prediction module, a safety constraint optimization module, a regulation command generation module, and a feedback correction module. The data acquisition module is used to acquire data on water injection wells, oil production wells, formations, equipment operation, and maintenance measures within the target block; The control event identification module is used to identify control events from historical production data where at least one of the parameters, namely injection volume, injection pressure, stratified water distribution ratio, fluid production volume and lifting parameters, changes. It also eliminates control events that overlap with operational measures, equipment malfunctions, and well shutdown status to obtain effective control events. The time-delay response sample construction module is used to extract production response data of oil wells within different time windows based on effective control events, and generate time-delay response samples between control actions and oil production responses. The injection-production response map generation module is used to determine the connectivity strength, response lag time, response stability, and water-bearing risk coefficient between injection-production well pairs and injection-production interval pairs based on time-delay response samples. The machine learning prediction module is used to generate candidate injection-production parameter combinations based on the injection-production response spectrum, the current production status, and the preset parameter disturbance range. The injection-production response spectrum, the current production status, and the candidate injection-production parameter combinations are input into the prediction model to obtain the predicted results of fluid production, oil production, water cut, dynamic fluid level, and bottom hole flowing pressure under the corresponding candidate injection-production parameter combinations. The safety constraint optimization module is used to project the candidate injection-production parameter combinations with safety constraints according to the constraints of injection pressure, stratified water distribution ratio, liquid production volume, lift parameters, injection-production ratio, water cut and formation pressure, and determine the executable injection-production parameter combination from the projected candidate injection-production parameter combinations based on the prediction results. The control command generation module is used to generate control commands for at least one of the following parameters: water injection volume, water injection pressure, stratified water distribution ratio, liquid production volume, and lifting parameters, based on the executable injection and production parameter combination. The feedback correction module is used to correct the injection-production response map, prediction model, and safety constraint parameters based on the deviation between the actual production response and the prediction results after the control is implemented.

[0007] Preferably, the data acquisition module includes a production data acquisition unit, a hierarchical data acquisition unit, a measure operation data acquisition unit, and a time alignment unit; The production data acquisition unit is used to collect the water injection volume, water injection pressure, and stratified water distribution ratio of water injection wells, as well as the fluid production volume, oil production, water cut, dynamic fluid level, bottom hole flowing pressure, and lift parameters of oil production wells. The lift parameters include the pump frequency of electric submersible pump wells and the stroke rate of pumping unit wells. The stratigraphic data acquisition unit is used to acquire data on well number, stratigraphic segment number, perforation location, well spacing, permeability, porosity, and formation pressure. The operational data acquisition unit is used to collect records of acidizing, fracturing, water shut-off, well washing, pump inspection, and well shutdown operations. The time alignment unit is used to align production data, stratigraphic data, and operational data into a basic injection and production data table at the same time scale according to well number, stratigraphic segment number, and acquisition time.

[0008] Preferably, the regulation event identification module includes a parameter change identification unit and an event elimination unit, and the time-delay response sample construction module includes a response window division unit and a response feature extraction unit; The parameter change identification unit is used to record the corresponding well number, layer number, parameter type, change range and occurrence time when the change of any parameter among the injection volume, injection pressure, stratified water distribution ratio, liquid production volume and lifting parameters relative to the previous acquisition cycle reaches the corresponding preset control range threshold, thus forming a candidate control event. The event elimination unit is used to eliminate candidate control events whose occurrence time overlaps with the periods of acidizing, fracturing, water shut-off, well washing, pump inspection, well shutdown operations, and equipment abnormality marking, and to determine the remaining candidate control events as effective control events. The response window segmentation unit is used to extract the production response data of the associated oil wells according to at least two consecutive lag time windows, starting from the occurrence time of the effective control event. The response feature extraction unit is used to extract the rate of change of fluid production, rate of change of oil production, rate of change of water cut, change of dynamic fluid level, and change of bottom hole flowing pressure from each lag time window, and to combine the parameter type, change amplitude, well number, layer number, and production response data in the effective control event into a time-delay response sample.

[0009] Preferably, the injection-sampling response map generation module includes a node construction unit, an edge relationship construction unit, and an edge attribute calculation unit; The node construction unit is used to treat water injection wells, oil production wells, and formations as map nodes respectively; The edge relationship construction unit is used to establish response edges between water injection well nodes and oil production well nodes, and between water injection section nodes and oil production section nodes, where time-delay response samples exist; The edge attribute calculation unit is used to write connectivity strength, response lag time, response stability, and water content risk coefficient for each response edge. The response lag time is the earliest lag time window when the corresponding production response data reaches the preset response judgment threshold. The response stability is determined by the number of times the production response direction is consistent within adjacent lag time windows. The water content risk coefficient is jointly determined by the water content change rate, the oil production change rate, and the liquid production change rate.

[0010] Preferably, the machine learning prediction module includes a training sample generation unit, a candidate parameter generation unit, a model training unit, and a prediction output unit; The training sample generation unit is used to generate a model input sample by combining the connection strength, response lag time, response stability, water cut risk coefficient, current production status and historical control parameters of the response edge in the injection-production response map as an index, and generates a model output label by generating the production volume, oil production, water cut, dynamic fluid level and bottom hole flowing pressure corresponding to the next control cycle. The candidate parameter generation unit is used to generate candidate injection-production parameter combinations, including water injection volume, water injection pressure, stratified water distribution ratio, liquid production volume, and lift parameters, based on the current production status, injection-production response spectrum, and preset parameter disturbance range. The model training unit is used to train a prediction model based on the model input samples and the model output labels; The prediction output unit is used to input the current production status, injection-production response spectrum and candidate injection-production parameter combinations into the trained prediction model, and output the prediction results of fluid production, oil production, water cut, dynamic fluid level and bottom hole flowing pressure under the corresponding candidate injection-production parameter combinations.

[0011] Preferably, the safety constraint optimization module includes a constraint boundary setting unit, a safety constraint projection unit, and an executable combination determination unit, and the feedback correction module includes a deviation calculation unit, a deviation attribution unit, and a parameter correction unit; The constraint boundary setting unit is used to set the upper limit of water injection pressure, the upper and lower limits of stratified water distribution ratio, the upper and lower limits of liquid production, the upper and lower limits of lifting parameters, the upper and lower limits of injection-production ratio, the upper limit of water cut, and the upper and lower limits of formation pressure respectively. The safety constraint projection unit is used to correct parameters in the candidate injection-import parameter combination that exceed the corresponding constraint boundary to the corresponding constraint boundary, and to limit the parameter change amplitude between adjacent control cycles, thus forming the projected candidate injection-import parameter combination. The executable combination determination unit is used to select parameter combinations that meet the upper limit of water cut, lower limit of bottom hole flowing pressure, upper and lower limits of formation pressure and upper and lower limits of injection-production ratio from the candidate injection-production parameter combinations after projection based on the prediction results, and to use the selected parameter combinations as executable injection-production parameter combinations. The deviation calculation unit is used to calculate the deviations between the actual fluid production, actual oil production, actual water cut, actual dynamic fluid level, and actual bottom hole flowing pressure after the control is implemented and the corresponding prediction results. The deviation attribution unit is used to generate well pair edge attribute correction marks when the deviation is concentrated in a single injection-production well pair, to generate layer edge attribute correction marks when the deviation is concentrated in a single layer, and to generate well group constraint parameter correction marks when the deviation occurs simultaneously in two or more production wells in the same well group. The parameter correction unit is used to correct at least one of the connectivity strength, response lag time and response stability of the corresponding response edge according to the well-to-edge attribute correction mark, to correct the water-bearing risk coefficient of the corresponding layer according to the layer edge attribute correction mark, and to correct at least one of the upper and lower limits of the injection-production ratio and the upper and lower limits of the formation pressure of the corresponding well group according to the well group constraint parameter correction mark.

[0012] The integrated optimization and matching control method for oilfield injection and production parameters includes the following steps: S1: Obtain data on water injection wells, oil production wells, formations, equipment operation and maintenance measures within the target block, and form a basic data table of injection and production at the same time scale according to well number, formation number and acquisition time; S2: Identify control events from the injection and production basic data table where at least one of the parameters of injection volume, injection pressure, stratified water distribution ratio, production volume and lift parameters changes, and eliminate control events that overlap with the operation, equipment abnormality and well shutdown status to obtain effective control events; S3: Using effective control events as the time benchmark, extract production response data of oil wells within different lag time windows to generate time-delay response samples between control actions and oil production responses; S4: Determine the connectivity strength, response lag time, response stability, and water-bearing risk coefficient between injection-production well pairs and injection-production interval pairs based on time-delay response samples, and generate injection-production response maps; S5: Generate candidate injection-production parameter combinations based on the injection-production response spectrum, current production status, and preset parameter disturbance range. Input the injection-production response spectrum, current production status, and candidate injection-production parameter combinations into the prediction model to obtain the prediction results of fluid production, oil production, water cut, dynamic fluid level, and bottom hole flowing pressure under the corresponding candidate injection-production parameter combinations. S6: Based on the constraints of injection pressure, stratified water distribution ratio, liquid production volume, lift parameters, injection-production ratio, water cut and formation pressure, perform safety constraint projection on the candidate injection-production parameter combinations, and determine the executable injection-production parameter combinations from the projected candidate injection-production parameter combinations based on the prediction results. S7: Generate control instructions for at least one of the following parameters based on the executable injection-production parameter combination: injection volume, injection pressure, stratified water distribution ratio, production volume, and lift parameters. S8: Based on the deviation between the actual production response and the predicted results after the control measures are implemented, the injection-production response map, prediction model, and safety constraint parameters are corrected.

[0013] Preferred events that result in effective regulation include: Calculate the changes in water injection volume, water injection pressure, stratified water distribution ratio, liquid collection volume, and lifting parameters relative to the previous collection cycle; When the change in any parameter reaches the corresponding preset control amplitude threshold, the corresponding well number, layer number, parameter type, change amplitude and occurrence time are recorded to form a candidate control event; Candidate control events whose occurrence time overlaps with the periods of acidizing, fracturing, water shut-off, well washing, pump inspection, well shutdown operations, and equipment abnormality marking are eliminated, and the remaining candidate control events are determined as effective control events. The preset control amplitude threshold is determined by the upper limit of the field instrument measurement error of the corresponding parameter, the minimum adjustment step size of the equipment, and the upper limit of natural fluctuation during the historical stable production period.

[0014] Preferably, generating the injection-progression response map includes: Injection wells, production wells, and formations are respectively designated as nodes in the map; Establish response edges between water injection well nodes and oil production well nodes, and between water injection section nodes and oil production section nodes where time-delay response samples exist; Write the connectivity strength, response lag time, response stability, and water content risk coefficient for each response edge; Among them, the response lag time is the earliest lag time window when the corresponding production response data reaches the preset response judgment threshold. The response stability is determined by the number of times the production response direction is consistent within adjacent lag time windows. The water cut risk coefficient is jointly determined by the water cut change rate, the oil production change rate, and the liquid production change rate.

[0015] Preferably, the process of performing safety-constrained projection and correction on candidate injection-import parameter combinations includes: Parameters in the candidate injection-import parameter combination that exceed the corresponding constraint boundary are corrected to within the corresponding constraint boundary, and the amplitude of parameter changes between adjacent control cycles is limited. When the same parameter is subject to two or more constraints at the same time, the parameter is adjusted to the intersection range that satisfies the two or more constraints simultaneously; When there is no intersection range, the corresponding candidate injection and sampling parameter combinations are marked as unexecutable combinations; After the control measures are implemented, if the deviation between the actual production response and the predicted result is concentrated in a single injection-production well pair, then at least one of the following should be corrected: the connectivity strength of the corresponding response edge, the response lag time, and the response stability. If the deviation between the actual production response and the predicted result is concentrated in a single layer, then the water content risk coefficient of the corresponding layer should be adjusted. If the deviation between the actual production response and the predicted result occurs simultaneously in two or more production wells within the same well group, then at least one of the upper and lower limits of the injection-production ratio and the upper and lower limits of the formation pressure for the corresponding well group shall be corrected.

[0016] In summary, the present invention has the following main beneficial effects: By removing interfering data such as acidizing, fracturing, water shut-off, well washing, pump inspection, well shutdown, and equipment malfunctions from historical production data, and using effective changes in water injection volume, water injection pressure, stratified water distribution ratio, fluid production volume, and lift parameters as control events, and then extracting the fluid production, oil production, water cut, dynamic fluid level, and bottom hole flowing pressure response of oil wells according to different lag time windows, a time-lag response sample between control actions and oil production response is established. This achieves the effect of distinguishing real injection-production control behavior from production fluctuations caused by measures and equipment malfunctions, so that subsequent injection-production relationship judgment no longer relies on a single static injection-production ratio or manual experience judgment, and can more accurately reflect the lag impact of water injection end parameter adjustments on oil production dynamics.

[0017] By generating injection-production response maps based on time-delay response samples and incorporating connectivity strength, response lag time, response stability, and water cut risk coefficients between injection-production well pairs and injection-production layer pairs, the system achieves the effect of uniformly transforming the existence of inter-well responses, when responses occur, whether responses are stable, and whether there is a risk of water cut increase into calculable map relationships. Simultaneously, the machine learning prediction module does not directly generalize predictions from the original production data. Instead, it uses the injection-production response map, current production status, and candidate injection-production parameter combinations as inputs to output predicted results for fluid production, oil production, water cut, dynamic fluid level, and bottomhole flowing pressure. This allows the prediction process to correspond to the inter-well time-delay responses and layer differences in the oilfield, thereby improving the matching degree between candidate control schemes and actual reservoir dynamics.

[0018] By setting safety constraint projections based on machine learning predictions, candidate injection-production parameter combinations are restricted to constraint boundaries such as injection pressure, stratified water distribution ratio, production volume, lift parameters, injection-production ratio, water cut, and formation pressure. When the same parameter is subject to multiple constraints, it is corrected to the range of constraint intersection. Combinations without constraint intersection are marked as unexecutable. This avoids directly adopting the scheme with the highest predicted benefit but which is unexecutable on site or has risks such as water cut surge, pressure anomaly, or equipment over-limit. At the same time, by adjusting the deviation between the actual response after execution and the prediction results, the injection-production response spectrum, prediction model, and safety constraint parameters are graded and corrected to achieve the effect of continuously updating the matching relationship of injection-production parameters with reservoir development dynamics. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1 refer to Figure 1 An integrated optimization and matching system for oilfield injection and production parameters, including: The system includes a data acquisition module, a regulation event identification module, a time-delay response sample construction module, an injection-machining response map generation module, a machine learning prediction module, a safety constraint optimization module, a regulation command generation module, and a feedback correction module. The data acquisition module is used to acquire data on water injection wells, oil production wells, formations, equipment operation, and maintenance measures within the target block; The control event identification module is used to identify control events from historical production data where at least one of the parameters, namely injection volume, injection pressure, stratified water distribution ratio, fluid production volume and lifting parameters, changes. It also eliminates control events that overlap with operational measures, equipment malfunctions, and well shutdown status to obtain effective control events. The time-delay response sample construction module is used to extract production response data of oil wells within different time windows based on effective control events, and generate time-delay response samples between control actions and oil production responses. The injection-production response map generation module is used to determine the connectivity strength, response lag time, response stability, and water-bearing risk coefficient between injection-production well pairs and injection-production interval pairs based on time-delay response samples. The machine learning prediction module is used to generate candidate injection-production parameter combinations based on the injection-production response spectrum, the current production status, and the preset parameter disturbance range. The injection-production response spectrum, the current production status, and the candidate injection-production parameter combinations are input into the prediction model to obtain the predicted results of fluid production, oil production, water cut, dynamic fluid level, and bottom hole flowing pressure under the corresponding candidate injection-production parameter combinations. The safety constraint optimization module is used to project the candidate injection-production parameter combinations with safety constraints according to the constraints of injection pressure, stratified water distribution ratio, liquid production volume, lift parameters, injection-production ratio, water cut and formation pressure, and determine the executable injection-production parameter combination from the projected candidate injection-production parameter combinations based on the prediction results. The control command generation module is used to generate control commands for at least one of the following parameters: water injection volume, water injection pressure, stratified water distribution ratio, liquid production volume, and lifting parameters, based on the executable injection and production parameter combination. The feedback correction module is used to correct the injection-production response map, prediction model, and safety constraint parameters based on the deviation between the actual production response and the prediction results after the control is implemented.

[0022] This application is used for the coordinated matching and control of injection and production parameters among water injection wells, oil production wells, and corresponding formations within a target block. The target block is a set of injection and production wells with production relationships within the same development unit, the same injection and production well group, or the same reservoir partition. The system includes a data acquisition module, a control event identification module, a time-delay response sample construction module, an injection and production response map generation module, a machine learning prediction module, a safety constraint optimization module, a control command generation module, and a feedback correction module. Each module transmits data in the following order: data acquisition, effective control event identification, time-delay response sample construction, injection and production response map generation, candidate injection and production parameter combination prediction, safety constraint projection, control command generation, and execution feedback correction.

[0023] The integrated optimization and matching system for oilfield injection and production parameters in this embodiment is also used to execute the integrated optimization and matching control method for oilfield injection and production parameters. This control method corresponds to the data processing procedures of each module in the system. Specifically, the data acquisition module corresponds to the steps of data acquisition and formation of the basic injection and production data table; the control event identification module corresponds to the step of identification of effective control events; the time-delay response sample construction module corresponds to the step of generating time-delay response samples; the injection and production response map generation module corresponds to the step of generating injection and production response maps; the machine learning prediction module corresponds to the step of generating and predicting candidate injection and production parameter combinations; the safety constraint optimization module corresponds to the step of determining safety constraint projection and executable injection and production parameter combinations; the control command generation module corresponds to the step of generating control commands; and the feedback correction module corresponds to the correction step after the control is executed.

[0024] The data acquisition module is used to acquire data on water injection wells, oil production wells, formations, equipment operation, and operational measures within the target block. Data sources include the oilfield's on-site data acquisition and monitoring system, wellhead pressure acquisition instruments, stratified water injection control equipment, oil production well metering systems, lift equipment controllers, downhole pressure gauges, dynamic fluid level test records, daily production reports, and operational measures records. For data with different acquisition frequencies, the time alignment unit aggregates data using a daily acquisition cycle. When pressure, flow rate, and lift parameters are acquired on an hourly or minute-by-minute basis, the valid acquisition values ​​within the same day are converted into daily averages, daily cumulative values, or end-of-day values. Specifically, water injection volume and production volume use daily cumulative values, water injection pressure, bottomhole flowing pressure, dynamic fluid level, and pump frequency use daily averages, and water cut uses the daily measurement value or the average of multiple measurements taken on the same day.

[0025] The data acquisition module generates a basic injection-production data table. This table includes at least the well number, formation number, acquisition time, injection volume, injection pressure, stratified water distribution ratio, produced fluid volume, oil production, water cut, dynamic fluid level, bottom hole flowing pressure, lift parameters, perforation location, well spacing, permeability, porosity, formation pressure, type of operational intervention, start time of operational intervention, end time of operational intervention, and equipment anomaly flag. Lift parameters are pump frequency in ESP wells and strokes in pumping unit wells; when other lift methods are used, the lift parameters are the adjustable operating frequency or displacement control parameters of the lift equipment. Bottom hole flowing pressure is directly collected by a downhole pressure gauge; for wells without a downhole pressure gauge, it is calculated based on the dynamic fluid level, casing pressure, oil pressure, fluid column density, and pump hanger depth, and the data source flag for the bottom hole flowing pressure is recorded in the injection-production data table.

[0026] When a water injection well has multiple layers, the stratified water allocation ratio is the proportion of water injection volume in each layer to the total water injection volume of the well, and the sum of the stratified water allocation ratios for each layer is 1. When a water injection well is a single-layer water injection well, the stratified water allocation ratio for that layer is recorded as 1. The stratigraphic data acquisition unit collects well number, stratigraphic number, perforation position, well spacing, permeability, porosity, and formation pressure data. The operational data acquisition unit collects records of acidizing, fracturing, water shut-off, well washing, pump inspection, and well shutdown operations. The time alignment unit aligns the production data, stratigraphic data, and operational data into a basic injection-production data table at the same time scale according to the well number, stratigraphic number, and acquisition time.

[0027] The regulation event identification module is used to identify valid regulation events from historical production data. The parameter change identification unit reads the injection and production basic data table and calculates the changes in injection volume, injection pressure, stratified water distribution ratio, production volume, and lift parameters relative to the previous acquisition cycle. When the change in at least one of the parameters—injection volume, injection pressure, stratified water distribution ratio, production volume, and lift parameters—reaches the corresponding preset regulation amplitude threshold, the module records the corresponding well number, layer number, parameter type, value before adjustment, value after adjustment, change amplitude, and occurrence time, forming a candidate regulation event.

[0028] The parameter change is calculated using the following formula: ; in, Indicates the first Koujing in the Parameters within a collection period The value; Indicates the first Parameters of the well in the previous acquisition cycle The value; This indicates the change in this parameter between adjacent acquisition periods; where the parameter The injection volume, injection pressure, stratified water distribution ratio, liquid production volume, or lifting parameters can be used, and the formulas can be used to calculate them respectively.

[0029] when When this happens, the parameter change identification unit records the change as a candidate control event, whereby... Indicates parameters The corresponding preset control amplitude threshold. The preset control amplitude threshold is set separately for different parameters. The preset control amplitude threshold for each parameter is jointly determined by the upper limit of the field instrument measurement error, the minimum adjustment step size of the equipment, and the upper limit of natural fluctuations during the historical stable production period. The historical stable production period is a continuous production cycle without any records of acidizing, fracturing, water shut-off, well washing, pump inspection, well shutdown, or equipment anomalies.

[0030] For parameters Its preset control amplitude threshold is determined according to the following formula: ; in, Indicates parameters The corresponding preset control amplitude threshold; Indicates parameters The upper limit of error for the corresponding data acquisition instrument or metering method; Indicates the field control equipment's parameters The minimum adjustable step size; Indicates parameters during historical stable production periods The upper limit statistical value of the absolute value of the change in adjacent collection cycles. The upper limit statistical value is the percentile statistical value of the absolute value of the change in this parameter during a historical stable production period, or the stable fluctuation boundary determined in the production management system of the target block.

[0031] In actual operation, a valid control event can correspond to the adjustment of a single parameter, or the joint adjustment of multiple parameters within the same well number or the same layer within the same acquisition cycle. For a single parameter adjustment, the control event identification module records the parameter type, the value before adjustment, the value after adjustment, and the magnitude of change. For the joint adjustment of multiple parameters, the control event identification module merges multiple parameters that reach the corresponding preset control magnitude threshold within the same well number, the same layer, and the same acquisition cycle into a single valid control event, and records the parameter type and magnitude of change for each parameter separately.

[0032] To avoid mistaking measurement errors, short-term fluctuations, or equipment start-ups and shutdowns for control events, if the same parameter increases and then decreases or decreases and then increases in two consecutive acquisition cycles, and the difference between the recovered value and the value before adjustment is less than a preset regression threshold, then this change will not be considered a candidate control event. The preset regression threshold is determined by the upper limit of the measurement error for this parameter and is stored in the system parameter table.

[0033] The event elimination unit removes interference from candidate control events. If the occurrence time of a candidate control event overlaps with the period of acidizing, fracturing, water shut-off, well washing, pump inspection, or well shutdown operations, or falls within the protection period after the completion of the aforementioned measures, the candidate control event is eliminated. The protection period is used to eliminate the continuous impact of the measures on the production rate, water cut, pressure, and dynamic fluid level. The length of the protection period is determined based on the historical production recovery cycle after the measures in the target block and is recorded in the system parameter table. If the occurrence time of a candidate control event corresponds to a period of flow meter malfunction, pressure sensor malfunction, lifting equipment failure, or communication interruption, the candidate control event is also eliminated. The remaining candidate control events are determined as valid control events.

[0034] The time-delay response sample construction module is used to construct time-delay response samples between control actions and oil production responses, using effective control events as the time benchmark. The response window division unit first determines the associated oil production wells based on the existing injection-production well group division, perforation layer correspondence, well spacing range, and fault boundary information of the target block. If there are conflicts between the injection-production well group division, layer correspondence, and well spacing range, the injection-production correspondence confirmed in the target block development plan or reservoir engineering interpretation results shall prevail. If a manually confirmed injection-production correspondence table already exists for the target block, this table shall be used first, and the response relationship shall be updated in subsequent feedback corrections.

[0035] The response window division unit is based on the occurrence time of the effective control event, and at least two consecutive lag time windows are divided backwards. The number and length of the response windows are determined according to the historical statistical cycle of the production response after water injection adjustment in the target block, and recorded in the system parameter table. The same response window division rule is used in the same round of control calculation for the same target block. The length of each response cycle is consistent with or an integer multiple of the target block's production data acquisition cycle. For water-driven oilfields, the response of water injection parameter adjustment to oil wells has a time lag; therefore, this embodiment does not directly use the oil well data on the day the control event occurs as the final response, but instead extracts the oil well production response separately through multiple lag time windows.

[0036] The response feature extraction unit extracts the rate of change of produced fluid volume, rate of change of produced oil volume, rate of change of water cut, change of dynamic fluid level, and change of bottom hole flowing pressure within each lag time window. The response features are compared against a previous benchmark window, which is a continuous stable production cycle prior to the occurrence of the effective control event. If the well has records of well shutdown, pump inspection, well cleaning, or equipment anomalies within the previous benchmark window, the effective control event is not used for constructing the time-lag response sample for that well.

[0037] The response characteristics are calculated according to the following formula: ; in, Indicates the first Production parameters of wells produced from the surface In the lag time window Internal response characteristics; Indicates the timing of an effective control event; Indicates the first Production parameters of wells produced from the surface In the lag time window The average value within; Indicates the first Production parameters of wells produced from the surface In the front reference window The average value within; To prevent positive numbers with a denominator of zero from having values ​​less than the measurement precision of the corresponding production parameter; production parameters This includes one of the following: fluid production, oil production, water cut, dynamic fluid level, and bottom hole flowing pressure. Formulas are used to calculate fluid production, oil production, water cut, dynamic fluid level, and bottom hole flowing pressure, respectively. For dynamic fluid level and bottom hole flowing pressure, the difference between the average value of the lag time window and the average value of the preceding reference window is used as the response characteristic.

[0038] The time-delay response sample includes the well number, layer number, parameter type, change range, and occurrence time of the effective control event, as well as the changes in fluid production, oil production, water cut, dynamic fluid level, and bottom hole flowing pressure of the associated oil wells within each lag time window. This time-delay response sample is used for subsequent injection-production response map generation and machine learning prediction model training.

[0039] The injection-production response map generation module generates injection-production response maps based on time-delay response samples. The node construction unit treats injection wells, production wells, and formations as map nodes. For stratified injection wells, the injection formations below the injection well node are designated as formation nodes; for production wells, their perforated or production formations are designated as formation nodes. The edge relationship construction unit establishes response edges between injection well nodes and production well nodes with time-delay response samples, and between injection formation nodes and production formation nodes with stratified response samples. Each response edge is associated with and stored in relation to a corresponding effective control event and response window.

[0040] The edge attribute calculation unit is used to write connectivity strength, response lag time, response stability, and water cut risk coefficient for each response edge. Preset response judgment thresholds are set separately for fluid production, oil production, water cut, dynamic fluid level, and bottom hole flowing pressure. For production parameters... Its preset response determination threshold It is determined jointly by the upper limit of the measurement error of this production parameter and the upper limit of the natural fluctuation during the historical stable production period, and Not less than the upper limit of the measurement error for this production parameter. If the production parameter within a certain lag time window... The absolute value of the response characteristics reaches If the production parameter generates a valid response within the specified lag time window, then it is determined that the response lag time is the earliest lag time window at which the corresponding production response data reaches the preset response judgment threshold.

[0041] Connectivity strength is determined by the response amplitudes of fluid production, oil production, dynamic fluid level, and bottomhole flowing pressure of the oil well after changes in injection end parameters. Connectivity strength is calculated using the following formula:

[0042] ; in, Represents the response edge The connectivity strength; This represents the set of effective lag time windows corresponding to the response edge; Indicates the lag time window The weights; Indicates the first Production volume of oil wells in the lag time window Internal response characteristics; Indicates the first Oil production from wells in the lag time window Internal response characteristics; Indicates the first Bottomhole flowing pressure of oil wells in lag time window Internal response characteristics; Indicates the first The dynamic fluid level of the oil well in the lag time window Internal response characteristics; , , , These are the weighting coefficients for the corresponding production parameters, and all are non-negative. The water cut risk coefficient is determined jointly by the rate of change in water cut, the rate of change in oil production, and the rate of change in produced fluid volume.

[0043] The water content risk factor is calculated using the following formula: ; in, Represents the response edge The water content risk coefficient; Indicates the first Water cut of oil wells in production within the target lag time window Internal response characteristics; Indicates the first Oil production from wells within the target lag time window Internal response characteristics; Indicates the first The fluid production rate of the well in the production port is within the target lag time window. Internal response characteristics; The time lag window for the response edge to reach the preset response determination threshold; , , The weights for water-cut risk are calculated, and all are non-negative. When the water cut of an oil well increases, the oil production decreases, and the fluid production increases, the water-cut risk coefficient of the corresponding response edge increases.

[0044] Weighting coefficients in connectivity strength calculation , , , Weighting coefficients in the calculation of water content risk coefficient , , and the weight of the lag time window All values ​​are stored in the system parameter table. The aforementioned weighting coefficients are determined by the verification error, development stage, and production management objectives of the historical control samples of the target block. Once determined, the weighting coefficients remain unchanged within the same training cycle; when the feedback correction module determines that prediction deviations occur consecutively in the same well group, the weighting coefficients are updated, and the values ​​before, after, and during the update are recorded.

[0045] Response stability is determined by the number of times the production response direction is consistent within adjacent lag time windows. Specifically, within consecutive lag time windows corresponding to the same response edge, if the target production parameters—fluid production, oil production, water cut, dynamic fluid level, and bottom hole flowing pressure—change in the same direction continuously, a consistent response is recorded; response stability is the ratio of the number of consistent responses to the number of valid comparisons. If the number of valid comparisons is zero, the response edge is not included in the training sample generation for the machine learning prediction module.

[0046] The machine learning prediction module is used to predict the production response under candidate injection-production parameter combinations based on the injection-production response map and the current production status. The machine learning prediction module includes a training sample generation unit, a candidate parameter generation unit, a model training unit, and a prediction output unit. The training sample generation unit uses the response edges in the injection-production response map as indices, generating model input samples by combining the connectivity strength of the response edges, response lag time, response stability, water cut risk coefficient, current production status, and historical control parameter combinations; and generating model output labels by generating production volume, oil production, water cut, dynamic fluid level, and bottom hole flowing pressure corresponding to the next control cycle.

[0047] The current production status includes the current water injection volume, water injection pressure, stratified water distribution ratio, fluid production volume, oil production volume, water cut, dynamic fluid level, bottom hole flowing pressure, lift parameters, formation pressure, and injection-production ratio within the target block. Historical control parameter combinations include adjustments to the water injection volume, water injection pressure, stratified water distribution ratio, fluid production volume, and lift parameters within the historical control period. To ensure the model corresponds to the edge attributes in the injection-production response map, each model input sample includes either the response edge number or the corresponding water injection well number, oil production well number, water injection segment number, and oil production segment number.

[0048] In one specific embodiment, the model training unit uses a gradient boosting regression tree to construct a multi-output prediction model, or trains single-output regression models for production volume, oil production, water cut, dynamic fluid level, and bottom hole flowing pressure respectively. Each model input sample includes response edge attributes, current production status, and candidate injection-production parameter combinations, and the model output is the predicted value of the production parameter corresponding to the next control cycle. When using a recurrent neural network model, graph neural network model, or multilayer perceptron regression model, the model input samples and model output labels remain consistent with this embodiment. The key to this embodiment is that the model input samples are formed by effective control events, time-delay response samples, and injection-production response maps, rather than directly inputting all historical production data into a general model.

[0049] The model training unit trains the prediction model based on the model input samples and model output labels, and uses a historical time period division method for both training and validation. The training data is earlier than the validation data to avoid data from before and after the same regulatory event being included in both the training and validation sets simultaneously. The following loss function is used during model training:

[0050] ; in, This represents the model training loss; Indicates the number of training samples; and They represent the first The predicted and actual fluid collection volumes for each sample. and They represent the first Predicted and actual oil production for each sample; and They represent the first The predicted moisture content and the actual moisture content of each sample; and They represent the first The predicted and actual dynamic liquid levels for each sample; and They represent the first Predicted bottomhole flowing pressure and actual bottomhole flowing pressure for each sample; These are the weighting coefficients for each output indicator, and all are non-negative. Each weighting coefficient is determined based on the control target of the target block and stored in the system parameter table; during the high water cut development stage, the weight corresponding to the water cut prediction error is increased; during the pressure maintenance stage, the weight corresponding to the bottom hole flowing pressure prediction error is increased.

[0051] The candidate parameter generation unit generates candidate injection-production parameter combinations based on the current production status, injection-production response spectrum, and preset parameter disturbance ranges. These combinations include at least one parameter to be adjusted from among injection volume, injection pressure, stratified water distribution ratio, production volume, and lift parameters. Preset parameter disturbance ranges are set separately for each parameter. For injection volume, injection pressure, stratified water distribution ratio, production volume, and lift parameters, the preset parameter disturbance range is determined by the current parameter value, the equipment's allowable adjustment range, the maximum allowable adjustment range per cycle, and the production management system. The equipment's allowable adjustment range is determined by the rated operating range of the injection control valve, stratified water distributor, ESP frequency converter, pumping unit parameter adjustment mechanism, or other field control equipment. The maximum allowable adjustment range per cycle is determined by the target block's production management system and equipment operating requirements and stored in the system parameter table.

[0052] The candidate parameter generation unit generates upward, maintain, and downward candidate values ​​within a preset parameter disturbance range, centered on the current parameter value. If the upward adjustment value of the corresponding parameter exceeds the equipment's allowable adjustment range or the maximum allowable adjustment amplitude per single cycle, the boundary of the aforementioned range is used as the upward candidate value; if the downward adjustment value of the corresponding parameter is lower than the equipment's allowable adjustment range or the maximum allowable adjustment amplitude per single cycle, the boundary of the aforementioned range is used as the downward candidate value. If the water-cut risk coefficient of a response edge is higher than a preset risk threshold, the upward candidate value of the corresponding injection layer is reduced, and the downward candidate value or maintain candidate value is increased; if the connectivity strength of a response edge reaches a preset connectivity strength threshold and the water-cut risk coefficient does not exceed the preset risk threshold, the upward, maintain, and downward candidate values ​​of the corresponding injection well or injection layer are retained.

[0053] The prediction output unit inputs the current production status, injection-production response spectrum, and candidate injection-production parameter combinations into the trained prediction model, and outputs the predicted production volume, oil production, water cut, dynamic fluid level, and bottom hole flowing pressure for each candidate injection-production parameter combination. The prediction results are saved in a one-to-one correspondence with the candidate injection-production parameter combinations and used as input to the safety constraint optimization module.

[0054] The safety constraint optimization module is used to project safety constraints onto candidate injection-production parameter combinations and determine executable combinations. The constraint boundary setting unit sets the upper and lower limits for water injection pressure, stratified water distribution ratio, fluid production volume, lift parameters, injection-production ratio, water cut, and formation pressure. These boundaries are derived from the rated parameters of the field equipment, the reservoir development plan, the single-well production system, and safety requirements. The upper limit for water injection pressure is determined by the wellhead equipment's pressure-bearing capacity, formation fracturing pressure, and the allowable pressure of the injection tubing; the upper and lower limits for stratified water distribution ratio are determined by the opening range of the stratified water distributor and the injection system for the target strata; the upper and lower limits for lift parameters are determined by the allowable operating range of the ESP or pumping unit; the upper limit for water cut is determined by the development management indicators for a single well or well group; and the upper and lower limits for formation pressure are determined by the reservoir pressure maintenance plan.

[0055] The safety constraint projection unit performs constraint correction on candidate injection-amplification parameter combinations. For upper and lower limit constraints of a single parameter, the following projection method is used:

[0056] ; in, Indicates the parameters in the candidate injection parameter combination Candidate values; Indicates parameters The projected value; Indicates parameters The lower limit; Indicates parameters Upper limit; parameters This refers to parameters such as injection volume, injection pressure, stratified water distribution ratio, liquid production volume, or lift parameters. This projection method ensures that the value of a single parameter is within its corresponding constraint boundary.

[0057] For constraints such as injection-production ratio, formation pressure, and water cut, which are influenced by multiple parameters, the safety constraint projection unit first determines whether the candidate injection-production parameter combination meets the constraints based on the prediction results. If the predicted water cut exceeds the upper limit, the water distribution ratio of the injection zone with a high water cut risk response edge to the oil well is reduced, or the candidate value of the corresponding oil well's fluid production volume is reduced. If the predicted bottomhole flowing pressure is lower than the lower limit, the candidate value of the oil well's lift parameter is reduced, or the candidate value of the water injection volume of the injection well with a stable response edge to the oil well is increased. If the predicted formation pressure is lower than the lower limit, the candidate value of the corresponding well group's water injection volume is increased, and the upward adjustment of the fluid production volume is restricted. If the predicted formation pressure is higher than the upper limit, the upward adjustment of the water injection volume is restricted, and the fluid production volume is kept from increasing.

[0058] The upper limit of equipment adjustment range is determined by the maximum allowable adjustment amount of the water injection control valve, stratified water distributor, electric submersible pump frequency converter, pumping unit parameter adjustment mechanism, or other field control equipment within a single control cycle. For different types of equipment, the upper limit of equipment adjustment range is stored separately and associated with the corresponding well number, layer number, and equipment number.

[0059] The safety constraint projection unit performs constraint correction on candidate injection-production parameter combinations in the following order: upper limit of water cut, lower limit of bottom hole flowing pressure, upper limit of injection pressure, upper limit of equipment adjustment range, and upper and lower limits of injection-production ratio. When the same parameter is subject to more than two constraints, the parameter is corrected to the intersection range that simultaneously satisfies the two or more constraints; when there is no intersection range, the corresponding candidate injection-production parameter combination is marked as an unexecutable combination and removed from subsequent screening.

[0060] The executable combination determination unit selects parameter combinations from the projected candidate injection-production parameter combinations based on the prediction results, ensuring they meet the upper limit of water cut, the lower limit of bottom hole flowing pressure, the upper and lower limits of formation pressure, and the upper and lower limits of the injection-production ratio. If multiple parameter combinations satisfy the constraints, the executable injection-production parameter combination is determined by a comprehensive ranking based on the predicted oil production value, water cut risk coefficient, parameter adjustment range, and response stability. In the comprehensive ranking, priority is given to parameter combinations where the predicted oil production does not decrease, the predicted water cut does not exceed the upper limit, the parameter adjustment range is small, and the response stability is high. If no parameter combination satisfies all constraints, the unit outputs the parameter combination that maintains the current production regime and generates a review and control instruction.

[0061] The control command generation module generates control commands based on executable injection-production parameter combinations. The control level determination unit determines the control level based on the injection-production ratio deviation, water cut risk coefficient, and response stability. The injection-production ratio deviation is the difference between the current well group's injection-production ratio and the target injection-production ratio. The water cut risk coefficient is derived from the response edges in the injection-production response map, and the response stability is derived from the edge attributes in the injection-production response map. Control levels include small-amplitude control level, coordinated control level, and verification control level.

[0062] The preset injection-production ratio deviation threshold, preset risk threshold, and preset stability threshold are all stored in the system parameter table. The preset injection-production ratio deviation threshold is determined by the allowable fluctuation range of the injection-production ratio in the target block development plan; the preset risk threshold is determined by the statistical distribution of the water cut risk coefficient in historical high water cut rise samples; and the preset stability threshold is determined by the lowest acceptable proportion of continuous and consistent response directions in historical effective control events. The control level determination unit calls the same threshold from the system parameter table each time it makes a determination, and does not temporarily change the threshold during a single calculation.

[0063] When the injection-production ratio deviation does not exceed the preset injection-production ratio deviation threshold, the water cut risk coefficient does not exceed the preset risk threshold, and the response stability reaches the preset stability threshold, the control level determination unit determines it as a small-scale control level. At this time, the instruction generation unit generates control instructions for single-well water injection volume, stratified water distribution ratio, fluid production volume, or lift parameters. When the injection-production ratio deviation exceeds the preset injection-production ratio deviation threshold, and there are two or more response edges within the same well group whose response stability reaches the preset stability threshold, it is determined as a coordinated control level. At this time, the instruction generation unit generates combined control instructions for water injection wells and oil production wells within the same well group. When the water cut risk coefficient exceeds the preset risk threshold, the response stability is lower than the preset stability threshold, or the executable combination determination unit cannot obtain a parameter combination that satisfies all constraints, it is determined as a verification control level. At this time, the instruction generation unit generates a verification control instruction containing the well number, layer number, candidate injection-production parameter combination, and triggering reason, and does not directly output an automatic execution instruction.

[0064] Control commands must include at least the well number, layer number, parameter type, current parameter value, target parameter value, allowable execution time, control level, and execution method. For stratified water injection wells, the control command includes the stratified water distribution ratio or the adjustment amount of the water distributor opening for the corresponding layer; for oil production wells, the control command includes the target value for fluid production, pump frequency, or pump stroke. Control commands are sent to the water injection control equipment, stratified water distribution control equipment, lift equipment controller, or production management terminal.

[0065] The feedback correction module is used to correct the injection-production response map, prediction model, and safety constraint parameters based on the actual production response after the control operation is executed. The deviation calculation unit collects the actual fluid production, actual oil production, actual water cut, actual dynamic fluid level, and actual bottomhole flowing pressure within the feedback observation period after the control operation is executed, and compares these data with the corresponding prediction results output by the machine learning prediction module to obtain the prediction deviation. The feedback observation period is consistent with the response lag time of the corresponding response edge in the injection-production response map, or it is a continuous observation period following the window containing the response lag time.

[0066] Prediction bias is calculated using the following formula: ; in, Indicates the first Production parameters of wells produced from the surface Prediction bias; This indicates the actual production parameter values ​​after the control measures were implemented; This represents the predicted production parameter values ​​output by the machine learning prediction module. To prevent positive numbers with a denominator of zero; production parameters This includes fluid production, oil production, water cut, dynamic fluid level, and bottom hole flowing pressure, which are then substituted into the formula for calculation. When the When the production parameters are dynamic fluid level or bottom hole flowing pressure, and Use the average value within the feedback observation period; when the first When the production parameters are liquid production rate or oil production rate, and Use the cumulative value or daily average value within the feedback observation period; when the first When the production parameter is moisture content, the average measurement value within the feedback observation period is used.

[0067] The deviation attribution unit generates different correction tags based on the distribution of predicted deviations. If the predicted deviation of a certain injection-production well pair for a corresponding production well exceeds a preset deviation threshold, while the predicted deviations of other production wells in the same well group do not exceed the preset deviation threshold, the deviation attribution unit determines that the deviation is concentrated in a single injection-production well pair and generates a well pair edge attribute correction tag. If the predicted deviations of two or more production wells corresponding to the same layer exceed the preset deviation threshold, while the predicted deviations of production wells corresponding to other layers do not exceed the preset deviation threshold, the deviation is determined to be concentrated in a single layer, and a layer edge attribute correction tag is generated. If the predicted deviations of two or more production wells in the same well group all exceed the preset deviation threshold, and the deviation directions are consistent, the deviation is determined to occur simultaneously in two or more production wells in the same well group, and a well group constraint parameter correction tag is generated. The preset deviation threshold is determined by the allowable upper limit of predicted deviations in historical verification samples and is stored in the system parameter table. If the predicted deviation overlaps with operational measures, equipment anomalies, or well shutdown events, the injection-production response spectrum is not corrected; instead, the corresponding feedback sample is marked as an invalid feedback sample.

[0068] The parameter correction unit performs corrections based on different correction markers. For well-to-edge attribute correction markers, the parameter correction unit corrects at least one of the connectivity strength, response lag time, and response stability of the corresponding response edge. For segment edge attribute correction markers, the parameter correction unit corrects the water-cut risk coefficient of the corresponding segment. For well group constraint parameter correction markers, the parameter correction unit corrects at least one of the upper and lower limits of the injection-production ratio and the upper and lower limits of formation pressure for the corresponding well group. Incremental updates are used during corrections to avoid sudden changes in the graph caused by a single abnormal feedback.

[0069] Edge attributes are updated according to the following formula: ; in, Indicates the updated edge connectivity strength; This indicates the edge connectivity strength before the update; This represents the response edge connectivity strength recalculated based on the actual production response after the implementation of this regulation. This represents the update coefficient, which takes a value greater than 0 and less than 1. The response lag time, response stability, and water content risk coefficient are corrected using the same incremental update method. Update coefficient The parameters are set based on the target block data update frequency and the degree of production fluctuation, and stored in the system parameter table; in well groups with large production data fluctuations, The value is lower than that of the stable well group in the production data. The values ​​are selected to reduce the impact of a single abnormal feedback on the response edge attributes. The corrected injection-mapping response map serves as the input to the machine learning prediction module and the safety constraint optimization module for the next regulation cycle.

[0070] In this embodiment, the update of the prediction model and the update of the injection-production response map are performed separately. When the deviation attribution unit determines that the deviation originates from a single injection-production well pair or a single layer, the corresponding response edge attribute is corrected first. When the same type of prediction deviation occurs within multiple consecutive feedback observation periods, the corresponding feedback sample is added to the training sample generation unit to retrain the prediction model. This avoids using occasional production fluctuations directly as model training samples, while ensuring that the injection-production response map is updated dynamically with reservoir development.

[0071] In this embodiment, the preset control amplitude threshold, preset regression threshold, protection period, response window length, preset response judgment threshold, preset parameter disturbance range, constraint boundary, preset injection-production ratio deviation threshold, preset risk threshold, preset stability threshold, preset deviation threshold, preset connectivity strength threshold, and update coefficient are all determined according to the instrument accuracy, equipment adjustment capability, historical stable production data, reservoir development plan, and production management system of the target block, and stored in the system parameter table. The system parameter table records at least the parameter name, applicable well number, applicable layer, parameter value, data source, determination time, and update record. Different target blocks have different system parameter tables; within the same control cycle of the same target block, the system calls the same system parameter table to perform control event identification, time-delay response sample construction, injection-production response map generation, machine learning prediction, safety constraint projection, control level determination, and feedback correction.

[0072] In another embodiment, this application provides an integrated optimization and matching control method for oilfield injection and production parameters. This control method is executed by the aforementioned integrated optimization and matching system for oilfield injection and production parameters and includes the following process.

[0073] First, acquire data on water injection wells, oil production wells, formations, equipment operation, and maintenance measures within the target block, and then create a basic injection-production data table at the same time scale according to well number, formation number, and acquisition time. The data source, time alignment, and data fields of the basic injection-production data table are consistent with those of the aforementioned data acquisition module.

[0074] Then, control events where at least one of the parameters—injection volume, injection pressure, stratified water distribution ratio, production volume, and lift parameters—changes are identified from the injection and production baseline data table. During identification, the change in each parameter relative to the previous acquisition cycle is calculated. When the change in any parameter reaches the corresponding preset control amplitude threshold, the well number, layer number, parameter type, change amplitude, and occurrence time are recorded to form candidate control events. Candidate control events whose occurrence time overlaps with acidizing, fracturing, water shut-off, well washing, pump inspection, well shutdown operation periods, and equipment anomaly marking periods are eliminated, and the remaining candidate control events are determined as valid control events. The preset control amplitude threshold is jointly determined by the upper limit of the field instrument measurement error for the corresponding parameter, the minimum equipment adjustment step size, and the upper limit of natural fluctuations during historical stable production periods.

[0075] Next, using the effective control event as the time benchmark, production response data from oil wells within different lag time windows are extracted to generate time-delay response samples between control actions and oil production responses. The production response data includes the rate of change in fluid production, the rate of change in oil production, the rate of change in water cut, the change in dynamic fluid level, and the change in bottom hole flowing pressure. The number and length of the lag time windows are determined by the historical injection-production response cycles of the target block and recorded in the system parameter table.

[0076] Subsequently, injection-production response maps are generated based on time-delay response samples. When generating the injection-production response maps, injection wells, production wells, and formations are respectively used as map nodes. Response edges are established between injection well nodes and production well nodes, and between injection formation nodes and production formation nodes, where time-delay response samples exist. For each response edge, connectivity strength, response lag time, response stability, and water-cut risk coefficient are written. The response lag time is the earliest lag time window at which the corresponding production response data reaches the preset response judgment threshold. Response stability is determined by the number of times the production response direction is consistent within adjacent lag time windows. The water-cut risk coefficient is jointly determined by the water cut change rate, oil production change rate, and fluid production change rate.

[0077] Subsequently, candidate injection-production parameter combinations are generated based on the injection-production response spectrum, current production status, and preset parameter disturbance range. These combinations are then input into the prediction model to obtain predicted results for fluid production, oil production, water cut, dynamic fluid level, and bottomhole flowing pressure under the corresponding candidate injection-production parameter combinations. The candidate injection-production parameter combinations include at least one parameter to be adjusted from among water injection volume, water injection pressure, stratified water distribution ratio, fluid production volume, and lift parameters. The input samples and output labels of the prediction model are generated in the same way as the training samples in the aforementioned machine learning prediction module.

[0078] Next, based on constraints such as injection pressure, stratified water distribution ratio, production volume, lift parameters, injection-production ratio, water cut, and formation pressure, a safety constraint projection is performed on the candidate injection-production parameter combinations. During safety constraint projection, parameters in the candidate injection-production parameter combinations that exceed the corresponding constraint boundaries are corrected to within the corresponding constraint boundaries, and the parameter variation amplitude between adjacent control cycles is limited. When the same parameter is subject to two or more constraints simultaneously, the parameter is corrected to the intersection range that simultaneously satisfies the two or more constraints; when the intersection range does not exist, the corresponding candidate injection-production parameter combination is marked as an unexecutable combination. Based on the prediction results, parameter combinations that satisfy the upper limit of water cut, lower limit of bottomhole flowing pressure, upper and lower limits of formation pressure, and upper and lower limits of injection-production ratio are selected from the projected candidate injection-production parameter combinations, and the selected parameter combinations are taken as executable injection-production parameter combinations.

[0079] Then, based on the executable injection-production parameter combination, control instructions are generated for at least one of the following parameters: injection volume, injection pressure, stratified water distribution ratio, production volume, and lift parameters. The control instructions include well number, stratum number, parameter type, current parameter value, target parameter value, allowable execution time, control level, and execution method.

[0080] Finally, based on the deviation between the actual production response and the predicted results after the control measures are implemented, the injection-production response map, prediction model, and safety constraint parameters are revised. If the deviation between the actual production response and the predicted results is concentrated in a single injection-production well pair, at least one of the following parameters is corrected: connectivity strength, response lag time, and response stability of the corresponding response edge. If the deviation is concentrated in a single layer, the water-cut risk coefficient of the corresponding layer is corrected. If the deviation occurs simultaneously in two or more production wells within the same well group, at least one of the upper and lower limits of the injection-production ratio and the upper and lower limits of the formation pressure of the corresponding well group is corrected. The revised injection-production response map, prediction model, or safety constraint parameters are then used in the next control cycle.

[0081] Each step in the above method embodiment can be executed by the corresponding module in the aforementioned system embodiment. Specifically, the formation of the injection-progression basic data table is executed by the data acquisition module, the identification of effective control events is executed by the control event identification module, the generation of time-delay response samples is executed by the time-delay response sample construction module, the generation of injection-progression response maps is executed by the injection-progression response map generation module, the generation of candidate injection-progression parameter combinations and production response prediction are executed by the machine learning prediction module, the determination of safety constraint projection and executable injection-progression parameter combinations is executed by the safety constraint optimization module, the generation of control instructions is executed by the control instruction generation module, and the calculation, attribution, and parameter correction of deviations after control are executed by the feedback correction module. Therefore, the system embodiment and the method embodiment maintain consistency in data flow, control flow, and parameter update logic.

[0082] In a specific operational process, the system first generates a basic injection-production data table for the target block. The control event identification module identifies an adjustment in the stratified water distribution ratio of a certain injection well and confirms that this adjustment does not overlap with periods of acidizing, fracturing, water shut-off, well washing, pump inspection, well shutdown, or equipment malfunction. Therefore, this adjustment is identified as a valid control event. The time-delay response sample construction module uses this valid control event as a time reference to extract changes in production volume, oil production, water cut, dynamic fluid level, and bottom hole flowing pressure of associated wells within the corresponding well group over multiple lag time windows, forming a time-delay response sample. The injection-production response map generation module updates the response edges between the injection segment node and the production segment node accordingly, and writes the connectivity strength, response lag time, response stability, and water cut risk coefficient. The machine learning prediction module generates multiple candidate injection-production parameter combinations based on the current production status and preset parameter disturbance range, and predicts the production response under each candidate combination. The safety constraint optimization module performs safety constraint projection on the candidate combinations, eliminates unexecutable combinations, and determines the executable injection-production parameter combinations. The control command generation module generates control commands based on the executable injection and extraction parameter combinations. After the control is executed, the feedback correction module collects the actual production response and corrects the injection and extraction response spectrum, prediction model, or safety constraint parameters based on the deviation between the actual response and the prediction results, thus completing one closed-loop control cycle.

[0083] This embodiment differs from methods that output injection-production parameters solely based on static injection-production ratios, single-well connectivity, numerical simulation surrogate models, or general reinforcement learning strategies. This embodiment first removes operational interventions, equipment malfunctions, and well shutdown interference from historical production data to obtain effective control events. Then, using these effective control events as a time base, it constructs time-delay response samples and generates an injection-production response map containing connectivity strength, response lag time, response stability, and water-cut risk coefficients. The machine learning prediction module does not directly use raw production data for generalization prediction; instead, it uses the response edge attributes of the injection-production response map, the current production state, and candidate injection-production parameter combinations as model inputs. The safety constraint optimization module does not directly adopt the candidate scheme with the highest predicted benefit; instead, it projects safety constraints based on water cut, bottomhole flowing pressure, injection pressure, equipment adjustment range, injection-production ratio, and formation pressure constraints, marking candidate schemes as unexecutable combinations when no constraint intersection exists. Therefore, the control scheme can correspond to the inter-well time-delay response, formation water-cut risk, and equipment executable boundaries in the oilfield.

[0084] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An integrated optimization and matching system for oilfield injection and production parameters, characterized in that, Includes the following steps: The system includes a data acquisition module, a regulation event identification module, a time-delay response sample construction module, an injection-machining response map generation module, a machine learning prediction module, a safety constraint optimization module, a regulation command generation module, and a feedback correction module. The data acquisition module is used to acquire data on water injection wells, oil production wells, formations, equipment operation, and maintenance measures within the target block; The control event identification module is used to identify control events from historical production data where at least one of the parameters, namely injection volume, injection pressure, stratified water distribution ratio, fluid production volume and lifting parameters, changes. It also eliminates control events that overlap with operational measures, equipment malfunctions, and well shutdown status to obtain effective control events. The time-delay response sample construction module is used to extract production response data of oil wells within different time windows based on effective control events, and generate time-delay response samples between control actions and oil production responses. The injection-production response map generation module is used to determine the connectivity strength, response lag time, response stability, and water-bearing risk coefficient between injection-production well pairs and injection-production interval pairs based on time-delay response samples. The machine learning prediction module is used to generate candidate injection-production parameter combinations based on the injection-production response spectrum, the current production status, and the preset parameter disturbance range. The injection-production response spectrum, the current production status, and the candidate injection-production parameter combinations are input into the prediction model to obtain the predicted results of fluid production, oil production, water cut, dynamic fluid level, and bottom hole flowing pressure under the corresponding candidate injection-production parameter combinations. The safety constraint optimization module is used to project the candidate injection-production parameter combinations with safety constraints according to the constraints of injection pressure, stratified water distribution ratio, liquid production volume, lift parameters, injection-production ratio, water cut and formation pressure, and determine the executable injection-production parameter combination from the projected candidate injection-production parameter combinations based on the prediction results. The control command generation module is used to generate control commands for at least one of the following parameters: water injection volume, water injection pressure, stratified water distribution ratio, liquid production volume, and lifting parameters, based on the executable injection and production parameter combination. The feedback correction module is used to correct the injection-production response map, prediction model, and safety constraint parameters based on the deviation between the actual production response and the prediction results after the control is implemented.

2. The integrated optimization and matching system for oilfield injection and production parameters according to claim 1, characterized in that, The data acquisition module includes a production data acquisition unit, a hierarchical data acquisition unit, a work operation data acquisition unit, and a time alignment unit; The production data acquisition unit is used to collect the water injection volume, water injection pressure, and stratified water distribution ratio of water injection wells, as well as the fluid production volume, oil production, water cut, dynamic fluid level, bottom hole flowing pressure, and lift parameters of oil production wells. The lift parameters include the pump frequency of electric submersible pump wells and the stroke rate of pumping unit wells. The stratigraphic data acquisition unit is used to acquire data on well number, stratigraphic segment number, perforation location, well spacing, permeability, porosity, and formation pressure. The operational data acquisition unit is used to collect records of acidizing, fracturing, water shut-off, well washing, pump inspection, and well shutdown operations. The time alignment unit is used to align production data, stratigraphic data, and operational data into a basic injection and production data table at the same time scale according to well number, stratigraphic segment number, and acquisition time.

3. The integrated optimization and matching system for oilfield injection and production parameters according to claim 2, characterized in that, The regulation event identification module includes a parameter change identification unit and an event elimination unit; the time-delay response sample construction module includes a response window division unit and a response feature extraction unit. The parameter change identification unit is used to record the corresponding well number, layer number, parameter type, change range and occurrence time when the change of any parameter among the injection volume, injection pressure, stratified water distribution ratio, liquid production volume and lifting parameters relative to the previous acquisition cycle reaches the corresponding preset control range threshold, thus forming a candidate control event. The event elimination unit is used to eliminate candidate control events whose occurrence time overlaps with the periods of acidizing, fracturing, water shut-off, well washing, pump inspection, well shutdown operations, and equipment abnormality marking, and to determine the remaining candidate control events as effective control events. The response window segmentation unit is used to extract the production response data of the associated oil wells according to at least two consecutive lag time windows, starting from the occurrence time of the effective control event. The response feature extraction unit is used to extract the rate of change of fluid production, rate of change of oil production, rate of change of water cut, change of dynamic fluid level, and change of bottom hole flowing pressure from each lag time window, and to combine the parameter type, change amplitude, well number, layer number, and production response data in the effective control event into a time-delay response sample.

4. The integrated optimization and matching system for oilfield injection and production parameters according to claim 3, characterized in that, The injection-collection response graph generation module includes a node construction unit, an edge relationship construction unit, and an edge attribute calculation unit; The node construction unit is used to treat water injection wells, oil production wells, and formations as map nodes respectively; The edge relationship construction unit is used to establish response edges between water injection well nodes and oil production well nodes, and between water injection section nodes and oil production section nodes, where time-delay response samples exist; The edge attribute calculation unit is used to write connectivity strength, response lag time, response stability, and water content risk coefficient for each response edge. The response lag time is the earliest lag time window when the corresponding production response data reaches the preset response judgment threshold. The response stability is determined by the number of times the production response direction is consistent within adjacent lag time windows. The water content risk coefficient is jointly determined by the water content change rate, the oil production change rate, and the liquid production change rate.

5. The integrated optimization and matching system for oilfield injection and production parameters according to claim 4, characterized in that, The machine learning prediction module includes a training sample generation unit, a candidate parameter generation unit, a model training unit, and a prediction output unit. The training sample generation unit is used to generate a model input sample by combining the connection strength, response lag time, response stability, water cut risk coefficient, current production status and historical control parameters of the response edge in the injection-production response map as an index, and generates a model output label by generating the production volume, oil production, water cut, dynamic fluid level and bottom hole flowing pressure corresponding to the next control cycle. The candidate parameter generation unit is used to generate candidate injection-production parameter combinations, including water injection volume, water injection pressure, stratified water distribution ratio, liquid production volume, and lift parameters, based on the current production status, injection-production response spectrum, and preset parameter disturbance range. The model training unit is used to train a prediction model based on the model input samples and the model output labels; The prediction output unit is used to input the current production status, injection-production response spectrum and candidate injection-production parameter combinations into the trained prediction model, and output the prediction results of fluid production, oil production, water cut, dynamic fluid level and bottom hole flowing pressure under the corresponding candidate injection-production parameter combinations.

6. The integrated optimization and matching system for oilfield injection and production parameters according to claim 5, characterized in that, The safety constraint optimization module includes a constraint boundary setting unit, a safety constraint projection unit, and an executable combination determination unit; the feedback correction module includes a deviation calculation unit, a deviation attribution unit, and a parameter correction unit. The constraint boundary setting unit is used to set the upper limit of water injection pressure, the upper and lower limits of stratified water distribution ratio, the upper and lower limits of liquid production, the upper and lower limits of lifting parameters, the upper and lower limits of injection-production ratio, the upper limit of water cut, and the upper and lower limits of formation pressure respectively. The safety constraint projection unit is used to correct parameters in the candidate injection-import parameter combination that exceed the corresponding constraint boundary to the corresponding constraint boundary, and to limit the parameter change amplitude between adjacent control cycles, thus forming the projected candidate injection-import parameter combination. The executable combination determination unit is used to select parameter combinations that meet the upper limit of water cut, lower limit of bottom hole flowing pressure, upper and lower limits of formation pressure and upper and lower limits of injection-production ratio from the candidate injection-production parameter combinations after projection based on the prediction results, and to use the selected parameter combinations as executable injection-production parameter combinations. The deviation calculation unit is used to calculate the deviations between the actual fluid production, actual oil production, actual water cut, actual dynamic fluid level, and actual bottom hole flowing pressure after the control is implemented and the corresponding prediction results. The deviation attribution unit is used to generate well pair edge attribute correction marks when the deviation is concentrated in a single injection-production well pair, to generate layer edge attribute correction marks when the deviation is concentrated in a single layer, and to generate well group constraint parameter correction marks when the deviation occurs simultaneously in two or more production wells in the same well group. The parameter correction unit is used to correct at least one of the connectivity strength, response lag time and response stability of the corresponding response edge according to the well-to-edge attribute correction mark, to correct the water-bearing risk coefficient of the corresponding layer according to the layer edge attribute correction mark, and to correct at least one of the upper and lower limits of the injection-production ratio and the upper and lower limits of the formation pressure of the corresponding well group according to the well group constraint parameter correction mark.

7. An integrated optimization and matching control method for oilfield injection and production parameters, applicable to the integrated optimization and matching system for oilfield injection and production parameters as described in any one of claims 1-6, characterized in that, Includes the following steps: S1: Obtain data on water injection wells, oil production wells, formations, equipment operation and maintenance measures within the target block, and form a basic data table of injection and production at the same time scale according to well number, formation number and acquisition time; S2: Identify control events from the injection and production basic data table where at least one of the parameters of injection volume, injection pressure, stratified water distribution ratio, production volume and lift parameters changes, and eliminate control events that overlap with the operation, equipment abnormality and well shutdown status to obtain effective control events; S3: Using effective control events as the time benchmark, extract production response data of oil wells within different lag time windows to generate time-delay response samples between control actions and oil production responses; S4: Determine the connectivity strength, response lag time, response stability, and water-bearing risk coefficient between injection-production well pairs and injection-production interval pairs based on time-delay response samples, and generate injection-production response maps; S5: Generate candidate injection-production parameter combinations based on the injection-production response spectrum, current production status, and preset parameter disturbance range. Input the injection-production response spectrum, current production status, and candidate injection-production parameter combinations into the prediction model to obtain the prediction results of fluid production, oil production, water cut, dynamic fluid level, and bottom hole flowing pressure under the corresponding candidate injection-production parameter combinations. S6: Based on the constraints of injection pressure, stratified water distribution ratio, liquid production volume, lift parameters, injection-production ratio, water cut and formation pressure, perform safety constraint projection on the candidate injection-production parameter combinations, and determine the executable injection-production parameter combinations from the projected candidate injection-production parameter combinations based on the prediction results. S7: Generate control instructions for at least one of the following parameters based on the executable injection-production parameter combination: injection volume, injection pressure, stratified water distribution ratio, production volume, and lift parameters. S8: Based on the deviation between the actual production response and the predicted results after the control measures are implemented, the injection-production response map, prediction model, and safety constraint parameters are corrected.

8. The integrated optimization and matching control method for oilfield injection and production parameters according to claim 7, characterized in that, Events that can be effectively regulated include: Calculate the changes in water injection volume, water injection pressure, stratified water distribution ratio, liquid collection volume, and lifting parameters relative to the previous collection cycle; When the change in any parameter reaches the corresponding preset control amplitude threshold, the corresponding well number, layer number, parameter type, change amplitude and occurrence time are recorded to form a candidate control event; Candidate control events whose occurrence time overlaps with the periods of acidizing, fracturing, water shut-off, well washing, pump inspection, well shutdown operations, and equipment abnormality marking are eliminated, and the remaining candidate control events are determined as effective control events. The preset control amplitude threshold is determined by the upper limit of the field instrument measurement error of the corresponding parameter, the minimum adjustment step size of the equipment, and the upper limit of natural fluctuation during the historical stable production period.

9. The integrated optimization and matching control method for oilfield injection and production parameters according to claim 8, characterized in that, The generation of injection-progression response maps includes: Injection wells, production wells, and formations are respectively designated as nodes in the map; Establish response edges between water injection well nodes and oil production well nodes, and between water injection section nodes and oil production section nodes where time-delay response samples exist; Write the connectivity strength, response lag time, response stability, and water content risk coefficient for each response edge; Among them, the response lag time is the earliest lag time window when the corresponding production response data reaches the preset response judgment threshold. The response stability is determined by the number of times the production response direction is consistent within adjacent lag time windows. The water cut risk coefficient is jointly determined by the water cut change rate, the oil production change rate, and the liquid production change rate.

10. The integrated optimization and matching control method for oilfield injection and production parameters according to claim 9, characterized in that, The process of performing safety-constrained projection and correction on candidate injection-production parameter combinations includes: Parameters in the candidate injection-import parameter combination that exceed the corresponding constraint boundary are corrected to within the corresponding constraint boundary, and the amplitude of parameter changes between adjacent control cycles is limited. When the same parameter is subject to two or more constraints at the same time, the parameter is adjusted to the intersection range that satisfies the two or more constraints simultaneously; When there is no intersection range, the corresponding candidate injection and sampling parameter combinations are marked as unexecutable combinations; After the control measures are implemented, if the deviation between the actual production response and the predicted result is concentrated in a single injection-production well pair, then at least one of the following should be corrected: the connectivity strength of the corresponding response edge, the response lag time, and the response stability. If the deviation between the actual production response and the predicted result is concentrated in a single layer, then the water content risk coefficient of the corresponding layer should be adjusted. If the deviation between the actual production response and the predicted result occurs simultaneously in two or more production wells within the same well group, then at least one of the upper and lower limits of the injection-production ratio and the upper and lower limits of the formation pressure for the corresponding well group shall be corrected.

Images