A method and system for predicting hydrocarbon production profile based on quantum tracer

Abstract

This invention relates to the field of oil and gas extraction, and proposes a production profile prediction method and system based on quantum tracers. The method includes: establishing quantum tracer coding pools corresponding to well sections and oil / gas / water phases, and deploying them according to the injection schedule; collecting and testing wellhead samples according to the sampling schedule during the production period; preprocessing the observation data to form standardized data; performing inversion based on statistical fitting, physical consistency, and temporal smoothing sparsity objectives to obtain a production profile of section × phase containing uncertainty; using sampling time point, phase, sampling volume, and coding subset as joint optimization variables, and combining information gain evaluation, condition number / coherence threshold, and resource constraints to determine the optimal sampling and coding scheme; performing robust optimization of production allocation, section closure, or repressurization based on the profile and uncertainty, and writing back and updating. This scheme helps improve data comparability and verifiability, shortens the uncertainty convergence cycle, and supports production management optimization.

Description

Technical Field

[0001] This invention relates to the field of oil and gas extraction technology, and more specifically, to a method and system for predicting oil and gas production profiles based on quantum tracers. Background Technology

[0002] In the production management of multi-stage fracturing wells, it is necessary to identify the relative contribution of each well segment to oil / gas / water without interrupting production, in order to support decisions such as production allocation, segment closure, or repressurization. The quantum tracer method involves injecting distinguishable quantum tracer codes into different well segments during the proppant-carrying fluid stage, giving each segment a "tag." During the production period, samples are collected at the wellhead according to a sampling schedule and tested to obtain the observation data returned by each code, thereby reconstructing the segment × phase production profile. Existing patents (such as CN119940659A) have disclosed the use of carbon quantum dot tracers, injected during the proppant-carrying fluid stage, and providing a relatively fixed sampling schedule. In data processing, the observation data is input into an inversion framework that includes a time-series model and a physical information neural network, ensuring that the results satisfy basic physical laws. In model application, iteration and transfer are achieved through a "prediction-fact comparison-model adjustment" approach, forming a basic process from field sampling to production profile estimation.

[0003] However, the scheme represented by CN119940659A still has some problems: the quantitative mapping and verification of factual data to output profiles are insufficient; the model transfer strategy lacks robustness and uncertainty control; the sampling scheduling and coding selection are mostly fixed empirical formulas, without joint optimization of variables, without information gain assessment with the goal of operational management decision-making effect, and without setting condition number thresholds and coherence thresholds for observation mapping relationship and configuring internal standard calibration channels; ultimately, under the constraints of budget, detection delay and laboratory capacity, it is difficult to obtain the most valuable information for production allocation, closure or repressurization at the minimum sampling cost.

[0004] Therefore, it is necessary to design a method and system for predicting oil and gas production profiles based on quantum tracers to solve the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes an oil and gas production profile prediction method based on quantum tracers, which aims to solve the problems of current technology, such as fixed empirical sampling system, lack of decision-oriented optimization and lack of robustness mechanism of model.

[0006] In one aspect, the present invention proposes a method for predicting oil and gas production profiles based on quantum tracers, comprising: Establish quantum tracer coding pools corresponding to well sections and oil / gas / water phases, and complete the delivery according to the injection log during the sand-carrying fluid stage; During the production period, wellhead samples are collected and tested according to the sampling schedule to obtain observation data for each code; The observation data are subjected to background subtraction, crosstalk correction, and instrument drift correction to form standardized data; Based on the inversion objective, which includes statistical fitting terms, physical consistency terms, and time smoothing sparsity terms, a segment × phase output profile containing uncertainty is obtained from the standardized data. The sampling time point, phase, sampling volume, and coding subset of the next period are used as joint optimization variables. Information gain evaluation with the goal of improving the effectiveness of operation and management decisions is adopted. Under the constraints of budget, detection delay, and laboratory capacity, the optimal sampling and coding scheme is determined. Condition number threshold and coherence threshold are set for the observation mapping relationship, and an internal standard calibration channel is configured. Based on the output profile and its uncertainty, robust optimization of production allocation, closure, or repressurization is generated, and the execution results are written back to update the inversion target and the sampling schedule.

[0007] Furthermore, during the production period, wellhead samples are collected and tested according to the sampling schedule to obtain observation data for each code, including: Generate instructions that include sampling time, phase and sampling volume according to the sampling schedule and bind the sample number; At the wellhead, 100–250 mL of oil phase sample and water phase sample were collected in light-proof inert containers and the sample number, timestamp and operating condition information were attached. The operating condition information included wellhead pressure, temperature, instantaneous flow rate and water content. The gas phase sample was enriched by filtering through a membrane or water absorption and the enriched volume was recorded. Mechanical oscillation and ultrasonic dispersion are performed on the oil phase sample and the aqueous phase sample, and elution or desorption is performed on the gas phase enrichment sample to obtain the detection solution; the internal standard calibration channel is set up, and internal standard beads of constant count or known intensity are added to each sample and the internal standard channel reading is recorded; The quality control samples of blank samples, on-site parallel samples, and spiked samples were collected in sequence. All samples are stored under 2–8℃ light-shielding conditions and tested within a preset time limit; the test results are counted by timestamp × code, and the original results are combined with the readings of the internal standard channel to form the observation data; when the signal-to-noise ratio of any sample is lower than the threshold or the internal standard deviates from the threshold, the result of the sample is marked as invalid and resampling is triggered according to the sampling schedule.

[0008] Furthermore, when performing background subtraction, crosstalk correction, and instrument drift correction on the observed data to form standardized data, the following steps are included: When performing background subtraction on the observation data, the process includes establishing a background library using the blank sample, determining the detection limit based on the background mean and dispersion of each coding channel, marking channels below the detection limit as indefinite, and subtracting the remaining channels based on the background mean. During crosstalk correction, the process includes calibrating the crosstalk effects between each coding channel using a single coded standard sample and the spiked sample to obtain a crosstalk matrix for correction; separating the signals of each channel according to the non-negative constraint deconvolution principle; and setting the regularization intensity based on the principle of minimizing the deviation of the internal standard calibration channel. When the robustness index of the observation mapping relationship exceeds the condition number threshold, channel merging is triggered. During instrument drift correction, the relative deviation between the current reading and the reference reading of the internal standard calibration channel is used as the basis to proportionally correct the readings of each channel after crosstalk removal, and a piecewise linear method is used to smooth the long-period drift. When generating standardized data, the results of each channel are normalized based on the sample volume, dilution factor and gas phase enrichment volume.

[0009] Furthermore, based on the inversion objective, which includes statistical fitting terms, physical consistency terms, and temporal smoothing sparsity terms, when obtaining the segment × phase output profile containing uncertainty from the standardized data, the process includes: Based on the observation mapping relationship, the standardized data is mapped to the segment × phase output profile, and non-negative constraints and yield upper limit constraints are applied. The statistical fitting term is set to measure the fitting bias; the physical consistency term is set, and its weight is adjusted according to the deviation of the internal standard calibration channel, the robustness index of the observation mapping relationship, and the condition number threshold; the temporal smoothing sparsity term is set. The inversion objective is solved using an iterative method with non-negative constraints. After each iteration, a stability check is performed based on the deviation between the robustness index and the internal standard calibration channel. If the threshold is not met, the regularization intensity is adjusted. Uncertainty assessment is performed on the obtained segment × phase output profile.

[0010] Furthermore, when the sampling time point, phase, sampling size, and encoded subset of the next time period are used as joint optimization variables, the following are included: Within the time window given by the sampling schedule, a set of candidate sampling time points is generated, and a subset of candidate codes is generated by combining the quantum tracer coding pool with the injection ledger; For each candidate scheme, the corresponding phase and sampling volume range are given; the quadruple consisting of sampling time point, phase, sampling volume and encoding subset is used as joint optimization variables to form the candidate set of the optimal sampling and encoding scheme, and each candidate scheme is bound with a unique sample number and the configuration requirements of the internal standard calibration channel.

[0011] Furthermore, when using information gain assessment aimed at improving the effectiveness of operational management decisions, it includes: With the primary objective of improving the decision-making quality of production allocation, closure, or repressurization, a measure of the convergence speed of uncertainty in the output profile of the segment × phase and the reduction of water content risk are introduced as secondary objectives. For each candidate scheme, an information gain evaluation is performed based on the standardized data and the most recent execution result, taking into account the magnitude of decision improvement and uncertainty reduction brought about by sampling.

[0012] Furthermore, under constraints of budget, testing latency, and laboratory capacity, the optimal sampling and coding scheme is determined, and condition number and coherence thresholds are set for the observation mapping relationship. When configuring the internal standard calibration channel, the following steps are taken: The budget limits the cost of single and cumulative sampling, the detection delay limits the maximum allowable time from sample collection to report issuance, and the laboratory capacity limits the upper limit of the number of parallel samples, together constituting the feasible domain of candidate schemes. For each candidate scheme, the robustness index of the observation mapping relationship is calculated and compared with the condition number threshold and the coherence threshold, and only candidate schemes that meet the threshold requirements are retained. The internal standard calibration channel is configured for the selected schemes in batches, with at least one internal standard calibration channel set up for field-laboratory link calibration for each batch. When any candidate scheme triggers a threshold failure or exceeds the capacity limit, a degradation strategy of channel merging, reducing the encoding subset, or adjusting the sampling time and sampling volume is implemented.

[0013] Furthermore, the execution process for determining the optimal sampling and coding scheme includes: sorting the candidate schemes that meet the feasible region and threshold requirements according to the information gain evaluation, selecting the one with the highest score as the optimal sampling and coding scheme, and issuing an execution instruction containing the sampling time point, phase, sampling volume, coding subset and the configuration of the internal standard calibration channel. After sample submission, the detection latency and laboratory capacity utilization are monitored in real time. When a delay risk or capacity congestion is detected, a local reordering is triggered and a low-priority scheme is replaced. After the detection is completed, the results are compared with the execution cost, execution deviation and the predicted value of the information gain assessment, and the comparison results are written into the parameter library of the sampling schedule and the information gain assessment.

[0014] Furthermore, when executing robust optimization commands for production allocation, phase closure, or repressurization based on the output profile and its uncertainty, and writing back the execution results to update the inversion target and the sampling schedule, the process includes: The robust and optimized feasible region is constructed, which consists of at least the safe range of wellhead pressure and temperature, the upper and lower limits of production of a single well and well section, the step size of throttling and valve position adjustment, the repressurization operation window and construction resource constraints, and environmental and safety constraints. The output profile and its uncertainty are transformed into a set of scenarios or intervals for decision-making according to a preset confidence level, and water content risk threshold and production capacity fluctuation threshold are set as risk control indicators. With the primary objective of increasing unit cost output and economic indicators, and with the secondary objective of reducing water content risk and uncertainty convergence speed, robust optimization is performed on candidate production allocation schemes, key section lists, and repressurization section combinations to output the optimal scheme that satisfies the feasible region and the risk control indicators. The optimal solution is used to generate the operation instruction, which includes at least the target output value, throttling or valve position setting, list of closed sections, repressurization section and pumping intensity, sampling encryption requirements and configuration of the internal standard calibration channel. During operation, actual output, moisture content, and pressure data are collected and compared with the observed data. The expected results of the operation instructions are compared with the actual results. When the deviation exceeds a threshold, the data is written back according to preset rules. Adjust the weights of the statistical fitting term, physical consistency term, and temporal smoothing sparsity term in the inversion objective, update the sampling time point, phase, and sampling volume of the sampling schedule, and specify the encoding subset. At the same time, write the execution cost, execution deviation, and uncertainty changes into the parameter library of the information gain evaluation.

[0015] Compared with existing technologies, the advantages of this invention are as follows: It utilizes background subtraction, crosstalk correction, and instrument drift correction to construct standardized data with a unified caliber, ensuring the quantitative comparability and verifiability of observed data and segment × phase output profiles; based on the inversion target output segment × phase output profile with uncertainty, which includes statistical fitting terms, physical consistency terms, and temporal smoothing sparsity terms, it remains stable and converges faster under multiphase transport and slow-release perturbations; it uses information gain evaluation to treat sampling time point, phase, sampling volume, and encoded subset as joint optimization variables, and considers budget, detection delay, and experimental... Under the constraint of production capacity, it automatically provides the optimal sampling and coding scheme, and sets condition number thresholds and coherence thresholds for observation mapping relationships and configures internal standard calibration channels to improve identifiability and result reliability. It obtains the most valuable information for decision-making with fewer samples and shorter testing cycles. Based on the output profile and its uncertainty, it performs robust optimization of production allocation, section closure or repressurization and generates work instructions. The execution results are written back to update the inversion target and sampling schedule, continuously reducing the risk of misadjustment and testing costs, improving unit cost output and decision consistency, and realizing quantifiable gains in production management and resource scheduling.

[0016] On the other hand, this application also provides a quantum tracer-based oil and gas production profile prediction system for applying the above-mentioned quantum tracer-based oil and gas production profile prediction method, including: The configuration unit is configured to establish a quantum tracer coding pool corresponding to the well section and oil / gas / water phase, and to complete the delivery according to the injection log during the sand-carrying fluid stage. The acquisition unit is configured to collect and test wellhead samples according to the sampling schedule during the production period to obtain observation data for each code. The preprocessing unit is configured to perform background subtraction, crosstalk correction, and instrument drift correction on the observation data to form standardized data; The processing unit is configured to obtain a segment × phase output profile containing uncertainty from the standardized data based on an inversion objective that includes statistical fitting terms, physical consistency terms, and temporal smoothing sparsity terms. The processing unit is also configured to use the sampling time point, phase, sampling volume and coding subset of the next time period as joint optimization variables, adopt information gain evaluation with the goal of improving the effectiveness of operation and management decisions, determine the optimal sampling and coding scheme under the constraints of budget, detection delay and laboratory capacity, set condition number threshold and coherence threshold for observation mapping relationship, and configure internal standard calibration channel. The execution unit is configured to execute robust optimization generation instructions for production allocation, closure, or repressurization based on the output profile and its uncertainty, and write back the execution results to update the inversion target and the sampling schedule.

[0017] It is understandable that the above-mentioned methods and systems for predicting oil and gas production profiles based on quantum tracers have the same beneficial effects, and will not be elaborated further here. Attached Figure Description

[0018] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart of a method for predicting oil and gas production profiles based on quantum tracers provided in an embodiment of the present invention; Figure 2 This is a functional block diagram of an oil and gas production profile prediction system based on quantum tracers provided in an embodiment of the present invention. Detailed Implementation

[0019] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] In traditional oil and gas production profile prediction methods, the quantitative mapping relationship between observational data and production profiles lacks an effective verification mechanism. Model transfer strategies lack robust control frameworks, and sampling scheduling and coding selection do not achieve multivariate joint optimization. Existing technologies rely on fixed, empirical sampling schemes, fail to prioritize improving operational management decision-making effectiveness as a core objective of information gain assessment, do not set condition number and coherence thresholds for observational mapping relationships, and lack internal standard calibration channels. This makes it difficult to balance data quality and sampling costs under budget and testing resource constraints.

[0021] For example, a multi-stage fractured horizontal well deployed 12 quantum tracer codes, collecting 200 mL samples of oil, gas, and water phases daily according to a fixed schedule. The detection data, after simple background subtraction, was input into the inversion model. However, due to the lack of crosstalk correction and instrument drift correction, the standardized data exhibited systematic bias. When the tracer code concentration in a certain well section approached the detection limit, the absence of a physical consistency term weighting adjustment mechanism led to the inversion results violating the law of mass conservation. In subsequent sampling scheduling, the laboratory accumulated 35% of the samples due to excessive detection delays, failing to trigger the resampling mechanism, resulting in a cumulative increase in production profile uncertainty. The decision-making level implemented production adjustments based on profile data with errors exceeding 20%, resulting in actual production fluctuations reaching 2.3 times the predicted value.

[0022] For this, please refer to Figure 1 As shown, this application proposes a method for predicting oil and gas production profiles based on quantum tracers, including: S100: Establish quantum tracer coding pools corresponding to well sections and oil / gas / water phases, and complete the delivery according to the injection log during the sand-carrying fluid stage.

[0023] S200: During the production period, wellhead samples are collected and tested according to the sampling schedule to obtain observation data for each code.

[0024] S300: Performs background subtraction, crosstalk correction, and instrument drift correction on the observation data to form standardized data.

[0025] S400: Based on the inversion objective, which includes statistical fitting terms, physical consistency terms, and time smoothing sparsity terms, a segment × phase output profile containing uncertainty is obtained from standardized data.

[0026] S500: The sampling time point, phase, sampling volume and coding subset of the next period are used as joint optimization variables. Information gain evaluation with the goal of improving the effectiveness of operation and management decisions is adopted. Under the constraints of budget, detection delay and laboratory capacity, the optimal sampling and coding scheme is determined. Condition number threshold and coherence threshold are set for the observation mapping relationship, and an internal standard calibration channel is configured.

[0027] S600: Based on the output profile and its uncertainty, execute robust optimization to generate work instructions for production allocation, closure, or repressurization, and write back the execution results to update the inversion target and sampling schedule.

[0028] Specifically, the quantum tracer coding pool refers to a distinguishable set of quantum codes corresponding to well sections and oil / gas / water phases. This can be achieved using combinations of quantum dots with different fluorescence wavelengths, lifetimes, or intensities, ensuring each well section and phase has a unique code identifier, thus addressing the insufficient label distinguishability in existing technologies. Standardized data refers to observational data after background subtraction, crosstalk correction, and instrument drift correction. This can be achieved by using blank samples to subtract background noise, crosstalk matrix to separate signals, and internal standard calibration to correct drift, eliminating interference factors during the detection process and improving data reliability. The inversion objective includes statistical fitting terms, physical consistency terms, and temporal smoothing sparsity terms. This can be achieved by using least squares fitting, physical constraint regularization, and temporal smoothing constraints to construct the objective function. By balancing data fitting and physical laws through multi-objective optimization, the problem of insufficient model robustness in existing technologies is addressed. The joint optimization variables refer to using sampling time points, phases, sampling volume, and encoded subsets as decision variables. Specifically, multi-objective optimization algorithms can be used to find the optimal combination under constraints of budget, detection latency, and laboratory capacity, achieving efficient allocation of sampling resources and improving information acquisition efficiency. Information gain assessment aims to improve the effectiveness of operational management decisions. This can be achieved by using decision trees, Bayesian networks, or value functions to quantify the improvement in production allocation, shutdown, or re-pressure decisions, ensuring that the sampling scheme directly serves production decision optimization. Condition number thresholds and coherence thresholds set limits on the numerical stability and signal independence of the observation mapping matrix. This can be achieved using matrix condition number calculation and coherence measurement methods to avoid distortion of results due to matrix ill-conditioning or signal aliasing during the inversion process. The internal standard calibration channel refers to adding standard beads of constant count or known intensity as a detection benchmark. This can be achieved using isotope markers or fluorescent standard materials, eliminating detection link bias through real-time calibration and ensuring data comparability.

[0029] This application uses sampling time point, phase, sampling volume, and encoded subset as joint optimization variables. Combining information gain evaluation and condition number threshold constraints, it dynamically formulates the optimal sampling scheme under budget and detection capability limitations. At the same time, it improves data reliability through an internal standard calibration channel, forming a closed-loop link from sampling design, data processing to decision support, and solving the problems of low sampling efficiency, insufficient data reliability, and weak decision support in the prior art.

[0030] The working process and principle of this application are as follows: First, a quantum tracer coding pool corresponding to the well section and oil / gas / water phase is established. During the sand-carrying fluid stage, the tracer is deployed according to the injection log. This step lays the foundation for subsequent production profile prediction, giving each well section a unique "tag".

[0031] During production, samples are collected at the wellhead according to a pre-defined sampling schedule and tested to obtain observation data for each code. This raw data contains information about the production of each well section, but further processing is required before it can be used for profile prediction.

[0032] Next, the observed data undergoes background subtraction, crosstalk correction, and instrument drift correction to form standardized data. Background subtraction eliminates the influence of background noise, crosstalk correction addresses interference between different codes, and instrument drift correction compensates for time-varying errors in the detection equipment. These correction steps improve the accuracy and reliability of the data.

[0033] Then, based on the inversion objective, which includes statistical fitting terms, physical consistency terms, and temporal smoothing sparsity terms, segment × phase output profiles containing uncertainty are obtained from standardized data. The statistical fitting term ensures the degree of fit between the model and the observed data, the physical consistency term ensures that the results conform to basic physical laws, and the temporal smoothing sparsity term introduces time continuity constraints, which together improve the stability and reliability of the inversion results.

[0034] To optimize the sampling strategy for the next time period, sampling time point, phase, sampling volume, and coding subset are used as joint optimization variables, and an information gain evaluation method aimed at improving operational management decision-making effectiveness is adopted. The optimal sampling and coding scheme is determined under constraints of budget, testing latency, and laboratory capacity. Simultaneously, condition number and coherence thresholds are set for the observation mapping relationship, and an internal standard calibration channel is configured to ensure data quality and model stability.

[0035] Finally, robust optimizations such as production allocation, section closure, or repressurization are performed based on the output profile and its uncertainty to generate work instructions. The execution results are written back to update the inversion objective and sampling schedule, forming a closed-loop optimization process.

[0036] This series of steps constitutes a complete and adaptive oil and gas production profile prediction. Through continuous iteration and optimization, prediction accuracy and decision-making effectiveness can be continuously improved, while maximizing the use of limited sampling and detection resources.

[0037] As a preferred embodiment, the solution of this application is specifically implemented as follows: In the production management of a multi-stage fractured horizontal well, a coding pool containing 20 quantum tracer codes was established, corresponding to the well section and oil / gas / water phase. During the proppant-carrying fluid stage, these codes were allocated to different well sections and deployed according to the injection log.

[0038] Once production begins, the initial sampling schedule is set to collect 200 mL of each of the oil, gas, and water phases daily. Immediately after sample collection, coded and analyzed data are obtained for each coded phase.

[0039] When processing the observation data, background subtraction is performed first. A background database is established using blank samples to determine the detection limit for each coded channel. Channels below the detection limit are marked as indefinite, and the remaining channels are subtracted based on the mean background value.

[0040] Next, crosstalk correction is performed. The crosstalk effect between each coded channel is calibrated using a single coded standard sample and a scald sample to obtain the crosstalk matrix. The signal of each channel is then separated using a non-negative constraint deconvolution principle.

[0041] Then, instrument drift correction is performed. Based on the relative deviation between the current reading and the reference reading of the internal standard calibration channel, the readings of each channel after crosstalk removal are proportionally corrected. A piecewise linear method is used to smooth long-period drift.

[0042] After the above steps, standardized data are generated. The results of each channel are then normalized according to the sampling volume, dilution factor, and gas phase enrichment volume.

[0043] Based on an inversion objective that includes statistical fitting terms, physical consistency terms, and temporal smoothing sparsity terms, segment × phase output profiles containing uncertainty are obtained from standardized data. Statistical fitting terms are used to measure fitting bias, physical consistency terms ensure that the results conform to fundamental physical laws such as mass conservation, and temporal smoothing sparsity terms introduce time continuity constraints. An iterative method with non-negativity constraints is used to solve the inversion objective, and a stability check is performed after each iteration.

[0044] To optimize the sampling strategy for the next time period, sampling time point, phase, sampling volume, and encoding subset are used as joint optimization variables. Within a given time window, a set of candidate sampling time points is generated, and a candidate encoding subset is generated by combining the quantum tracer encoding pool and the injection log. For each candidate scheme, the corresponding phase and sampling volume range are also given.

[0045] An information gain evaluation method aimed at improving the effectiveness of operational management decisions is adopted. The primary objective is to improve the quality of decisions regarding production allocation, phase closure, or repressurization. Secondary objectives include measures of the convergence speed of uncertainty in the phase-by-phase output profile and the reduction of water content risk. Information gain is evaluated for each candidate scheme based on standardized data and the most recent execution results.

[0046] The optimal sampling and coding scheme is determined under constraints of budget, testing latency, and laboratory capacity. The budget limits the cost of single and cumulative sampling, the testing latency limits the maximum allowable time from sample collection to report generation, and the laboratory capacity limits the upper limit of the number of parallel samples. For each candidate scheme, a robustness index of the observation mapping relationship is calculated and compared with condition number thresholds and coherence thresholds; only candidate schemes that meet the threshold requirements are retained.

[0047] Finally, robust optimization of production allocation, shut-off, or repressurization is performed based on the production profile and its uncertainty, generating operational instructions. A feasible domain for robust optimization is constructed, including the safe range of wellhead pressure and temperature, upper and lower limits of production for single wells and well sections, throttling and valve position adjustment steps, repressurization operation window and construction resource constraints, as well as environmental and safety constraints. Water cut risk thresholds and production fluctuation thresholds are set as risk control indicators. With increasing unit cost production and economic indicators as the primary objective, and comprehensively reducing water cut risk and uncertainty convergence speed as secondary objectives, robust optimization is performed on candidate production allocation schemes, shut-off lists, and repressurization section combinations.

[0048] The optimal solution is used to generate work instructions, including target output value, throttling or valve position settings, a list of closed sections, repressurization sections and pumping intensity, sampling density requirements, and internal standard calibration channel configuration. During operation execution, actual output, water content, pressure, and observation data are collected, and the expected results of the work instructions are compared with the actual results. When the deviation exceeds a threshold, the weights of each item in the inversion target are adjusted according to preset rules, the sampling schedule is updated, and the execution cost, execution deviation, and uncertainty changes are written into the parameter library for information gain evaluation.

[0049] Through the above scheme, this application achieves quantitative mapping and improved verifiability of observational data to output profiles. By introducing background subtraction, crosstalk correction, and instrument drift correction, the accuracy of standardized data is improved. The inversion objective function integrates statistical fitting terms, physical consistency terms, and temporal smoothing sparsity terms, enhancing the structural stability of the solution and avoiding the solution space collapse problem.

[0050] Joint optimization of sampling scheduling and coding selection, combined with information gain assessment aimed at improving operational management decision-making effectiveness, achieved efficient utilization of sampling resources. Data quality and model stability were further improved by setting condition number and coherence thresholds for observation mapping relationships and configuring internal standard calibration channels.

[0051] Under constraints of budget, testing latency, and laboratory capacity, this solution can obtain the most valuable information for production allocation, shutdown, or repressurization decisions with minimal sampling cost. By writing back the execution results to update the inversion target and sampling schedule, an optimization closed loop is formed, continuously improving prediction accuracy and decision-making effectiveness.

[0052] This quantum tracer-based method for predicting oil and gas production profiles improves the accuracy and efficiency of production management in multi-stage fracturing wells. It provides a reliable basis for identifying the relative contributions of each well section to oil, gas, and water without interrupting production, effectively supporting the formulation of key decisions such as production allocation, section shut-off, or repressurization.

[0053] This application further proposes generating instructions during the production period according to the sampling schedule, including sampling time, phase, and sample volume, and binding sample numbers. At the wellhead, 100–250 mL each of oil and water phase samples are collected in light-proof inert containers, and sample numbers, timestamps, and operating condition information are affixed. The operating condition information includes wellhead pressure, temperature, instantaneous flow rate, and water cut. Gas phase samples are enriched through a filter membrane or water absorption, and the enriched volume is recorded. Mechanical vibration and ultrasonic dispersion are performed on the oil and water phase samples, and elution or desorption is performed on the gas phase enriched samples to obtain the detection solution. An internal standard calibration channel is set up, and internal standard beads of constant count or known strength are added to each sample, with the internal standard channel reading recorded. It should be noted that the internal standard is added and the channel reading is recorded after the detection solution is generated. Quality control collection of blank samples, field parallel samples, and spiked samples is completed sequentially. All samples are stored under light-proof conditions at 2–8℃ and tested within a preset time limit. The test obtains a count by timestamp × code, and the raw results and internal standard channel readings together form the observation data. When the signal-to-noise ratio of any sample is lower than the threshold or the internal standard deviation exceeds the threshold, the sample result is marked as invalid and resampling is triggered according to the sampling schedule.

[0054] When generating instructions, the sampling time point is adjusted dynamically based on production conditions, and the phase and sample volume are dynamically matched according to the wellhead fluid state. Oil and water phase sample volumes are controlled within the range of 100–250 mL, using light-shielding inert containers to avoid photolysis and chemical reactions. Gas phase samples are enriched through filter membranes or water absorption, with the enrichment volume correlated with the wellhead gas flow rate. Mechanical oscillation and ultrasonic dispersion are used to uniformly disperse tracer particles in the oil and water samples, and elution or desorption steps release the tracer from the gas-enriched sample. An internal standard calibration channel provides a quality control benchmark for the detection process for each sample by adding standard beads of known strength. Blank samples are used for background subtraction, field parallel samples assess operational repeatability, and spiked samples verify detection accuracy. Sample storage temperature is controlled between 2–8℃, and light-shielding conditions prevent quantum tracer degradation. The detection time limit is set based on the tracer half-life to ensure signal stability. The signal-to-noise ratio threshold and internal standard deviation threshold are determined through historical data statistics, and adjacent time points are prioritized when triggering re-extraction.

[0055] Specifically, when collecting oil and aqueous phase samples at the wellhead, operators select the corresponding collection point according to the phase information in the instructions. Oil phase samples are taken from the oil phase pipeline at the separator outlet, aqueous phase samples are taken from the aqueous phase outlet pipeline, and gas phase samples are collected through a filter membrane enrichment device. Sample numbers and timestamps are bound to QR code tags, and operating condition information is recorded in real time by sensors and synchronized to the tags. Mechanical oscillation is performed using a horizontal oscillator at a frequency of 200 times / minute for 5 minutes, followed by ultrasonic dispersion at a frequency of 40kHz for 10 minutes. Gas phase elution involves soaking the filter membrane in methanol solution for 30 minutes, and desorption is performed by heating to 60℃ and purging with nitrogen. Internal standard beads are added before sample pretreatment, and their intensity matches the expected count range of the sample. In the quality control samples, the blank sample is the wellhead fluid without tracer, the field parallel sample is two samples collected at the same time point, and the spiked sample is the sample with a known concentration of tracer added. After the test is completed, the internal standard channel readings are automatically compared with the baseline value. If the deviation exceeds 5%, the data is deemed invalid, triggering a resampling instruction and prioritizing resampling within 24 hours. Through the above steps, the entire process of sample collection, processing, and testing is standardized, ensuring that the observation data is traceable and verifiable, providing high-quality input for subsequent inversion.

[0056] As a preferred embodiment, the solution of this application is specifically implemented as follows: The sampling schedule generates instructions including sampling time, phase, and sample volume, and assigns sample numbers. At the wellhead, 150 mL each of oil and water phase samples are collected in light-proof inert containers, and each container is labeled with a sample number, timestamp, and operating condition information. Operating condition information includes wellhead pressure, temperature, instantaneous flow rate, and water cut. The gas phase sample is enriched through a filter membrane, and the enriched volume is recorded.

[0057] Mechanical oscillation and ultrasonic dispersion were performed on oil and aqueous phase samples, and elution was performed on gas-enriched samples to obtain the detection solution. An internal standard calibration channel was set up, and a constant count of internal standard beads was added to each sample, and the reading of the internal standard channel was recorded.

[0058] Quality control samples, including blank samples, field parallel samples, and spiked samples, were collected sequentially. All samples were stored at 4°C under light-proof conditions and tested within 24 hours. Counts were obtained by timestamp and encoding, and the raw results, along with internal standard channel readings, formed the observation data. If the signal-to-noise ratio of any sample was below 3 or the internal standard deviation exceeded 10%, the sample result was marked as invalid, and resampling was triggered according to the sampling schedule.

[0059] Through the above technical solutions, this application achieves standardized collection, processing, and testing of wellhead samples. By setting up an internal standard calibration channel, instrument drift and sample matrix effects during the testing process can be effectively monitored and corrected, improving data reliability. The use of quality control samples and a resampling mechanism further ensures data quality. Simultaneously, recording detailed operating information provides necessary auxiliary information for subsequent data analysis. This systematic sampling and testing method improves the accuracy and reliability of production profile prediction.

[0060] This application further proposes methods for background subtraction, crosstalk correction, and instrument drift correction of observation data.

[0061] The background subtraction process involves establishing a background library using blank samples. Detection limits are set by calculating the mean and dispersion of the background for each coded channel. Channels below the detection limit are marked as inquantitative, while the remaining channels are subtracted based on the mean background. Crosstalk correction uses a single coded standard sample and a spiked sample to calibrate the cross-influence matrix between channels. A non-negative constraint deconvolution method is used to separate the signal, and the regularization intensity is dynamically adjusted based on the deviation of the internal standard calibration channels. Instrument drift correction is based on the deviation between the current reading and the reference reading of the internal standard calibration channels. The de-crosstalk channel readings are proportionally corrected, and long-period drift is handled using a piecewise linear method. In the standardized data formation stage, the channel results are normalized according to the sample volume, dilution factor, and gas phase enrichment volume.

[0062] Specifically, the background subtraction stage utilizes a background library established using blank samples to effectively identify background signals caused by environmental noise and reagent residues. For example, when the mean background value of a certain coded channel in an oil phase sample reaches 500 counts and the dispersion is ±50, the detection limit is set to the mean plus three times the dispersion, i.e., 650 counts; signals below this value are considered invalid. In crosstalk correction, the crosstalk coefficient matrix of adjacent channels can be calibrated by injecting a single coded standard sample. For instance, if the emission spectrum of coded A generates 10% interference signal in the detection channel of coded B, this interference is proportionally subtracted from the signal of coded B during deconvolution. Instrument drift correction uses an internal standard calibration channel as a reference. When the reading of this channel drops from an initial 10,000 counts to 9,500 counts, all channel readings are adjusted upwards by 5%. The piecewise linear method divides the drift curve over 30 consecutive days into 5 intervals for local linear fitting, eliminating long-term trend deviations. During the normalization phase, the volume of the gas phase enrichment sample was corrected from 50 mL to a standard volume of 100 mL to ensure the comparability of data from different sample volumes.

[0063] As a preferred embodiment, the solution of this application is specifically implemented as follows: When subtracting background from observed data, a background library is established using blank samples. The detection limit is determined based on the mean and dispersion of the background for each coded channel. Channels below the detection limit are marked as indeterminate, and the remaining channels are subtracted based on the mean background. For example, 10 blank samples can be collected, and the mean background value and standard deviation for each coded channel can be calculated. The mean value plus three times the standard deviation is used as the detection limit. Channels below the detection limit are marked as "_LOD" (below detection limit) in subsequent analyses.

[0064] During crosstalk correction, a single coded standard sample and a spiked sample are used to calibrate the crosstalk effects between each coded channel, resulting in a crosstalk matrix for correction. The signals of each channel are separated according to the non-negative constraint deconvolution principle, and the regularization strength is set based on the principle of minimizing the deviation of the internal standard calibration channels. Channel merging is triggered when the robustness index of the observation mapping relationship exceeds the condition number threshold. The robustness index consists of the condition number threshold and the coherence threshold of the observation mapping relationship. Specifically, iterative threshold shrinkage (ISTA) algorithm can be used for deconvolution, and the regularization strength is determined through cross-validation.

[0065] When performing instrument drift correction, the relative deviation between the current reading of the internal standard calibration channel and the reference reading is used as the basis for proportionally correcting the readings of each channel after crosstalk removal. A piecewise linear method is used to smooth long-period drift. For example, the internal standard can be collected every 4 hours, the deviation relative to the reference value can be calculated, and the sample results within that time period can be linearly corrected.

[0066] When generating standardized data, the results for each channel are normalized based on the sample volume, dilution factor, and gas phase enrichment volume. For example, for liquid samples, the results can be converted to concentration per unit volume according to the actual sample volume and dilution factor. For gas samples, the enrichment volume needs to be considered, and the results converted to concentration per unit gas volume.

[0067] Through the above technical solutions, this application effectively eliminates systematic errors such as background interference, crosstalk, and instrument drift, improving the accuracy and reliability of observational data. Standardization processing makes data from different batches and phases comparable, laying the foundation for subsequent production profile inversion. The introduction of internal standards and the setting of condition number thresholds further enhance the robustness of data processing, making the results more reliable. The data preprocessing method improves the accuracy and reliability of oil and gas production profile prediction based on quantum tracers.

[0068] This application further proposes a method based on the correspondence between standardized data and segment × phase output profiles, applying non-negativity constraints and yield upper limit constraints, setting statistical fitting terms, physical consistency terms and time smoothing sparsity terms, using an iterative method with non-negativity constraints to solve the inversion objective, performing stability checks during iteration, and finally evaluating the uncertainty.

[0069] In this process, the statistical fitting term quantifies the deviation between observed data and model predictions using least squares or maximum likelihood estimation. The physical consistency term introduces wellbore flow equations or mass balance equations to constrain the yield distribution, and its weights are dynamically adjusted based on the robustness index of the internal standard calibration channel deviation and observation mapping relationship. The temporal smoothing sparsity term uses L1 regularization or total variation regularization to suppress abrupt changes in yield between adjacent time periods. During the iterative solution process, the alternating direction multiplier method or projected gradient method is used to handle non-negativity constraints, and the regularization parameters are adjusted based on the robustness index and internal standard deviation after each iteration. Uncertainty assessment uses Monte Carlo simulation or covariance matrix propagation to quantify the impact of model parameter errors on yield.

[0070] Specifically, standardized data establishes a linear or nonlinear relationship with segment × phase yields through an observation mapping matrix. Non-negativity constraints exclude negative yields, and the yield cap constraint is set based on the wellbore's sand-carrying capacity. Statistical fitting terms calculate the sum of squared residuals, physical consistency terms calculate the degree of matching between yield and wellhead flow and pressure data, and time smoothing sparsity terms calculate the sum of the absolute values ​​of yield differences between adjacent time periods. During iterative solving, the initial guessed yield gradually approaches the optimal solution by minimizing the objective function, and the regularization strength dynamically increases or decreases based on the deviation of internal standard channel readings. When stability checks are triggered, increasing the regularization strength suppresses overfitting or decreasing it improves sensitivity. Finally, uncertainty assessment involves perturbing the input data or model parameters, statistically analyzing the distribution range of yield changes, and generating confidence intervals or standard deviations as output.

[0071] As a preferred embodiment, the solution of this application is specifically implemented as follows: Based on the observation mapping relationship, the standardized data is mapped to the segment × phase output profile, and non-negative constraints and yield upper limit constraints are applied.

[0072] A statistical fit term is set to measure fit bias. A physical consistency term is set, and its weights are adjusted according to the bias of the internal standard calibration channel, the robustness index of the observation mapping relationship, and the condition number threshold. A temporal smoothing sparsity term is set.

[0073] An iterative method with non-negative constraints is used to solve the inversion objective. After each iteration, a stability check is performed based on the deviation between the robustness index and the internal standard calibration channel. If the threshold is not met, the regularization intensity is adjusted.

[0074] Uncertainty assessment is performed on the obtained segment × phase output profile.

[0075] Specifically, the least squares method can be used as the statistical fitting term to measure the fitting deviation between the standardized data and the inversion results. The physical consistency term can include constraints such as mass conservation and energy balance, and its weights can be dynamically adjusted based on indicators such as internal standard deviation and condition number. The time-smoothing sparsity term can be implemented using L1 norm regularization.

[0076] During the iterative solution process, a maximum number of iterations and a convergence threshold can be set. After each iteration, robustness metrics and internal standard deviation are checked; if they exceed the preset threshold, the regularization strength is increased. Finally, the Monte Carlo method is used to perform uncertainty analysis on the solution results to obtain the confidence interval.

[0077] Through the above technical solutions, this application can improve the accuracy and reliability of output profile inversion. The inversion objective based on multiple regularization constraints makes the solution process more stable and effectively suppresses noise interference. The dynamically adjusted regularization strength can adapt to different operating conditions, ensuring the physical rationality of the inversion results. Uncertainty assessment provides a reliable basis for subsequent decision-making and helps reduce operational risks.

[0078] This application further proposes generating a set of candidate sampling time points within a given time window of the sampling schedule, and combining the quantum tracer encoding pool and injection log to generate a candidate encoding subset. For each candidate scheme, the corresponding phase and sampling volume range are simultaneously given. The quadruple consisting of sampling time point, phase, sampling volume, and encoding subset is used as a joint optimization variable to form a candidate set of optimal sampling and encoding schemes, and each candidate scheme is bound to a unique sample number and internal standard calibration channel configuration requirements.

[0079] The generation of the candidate sampling time point set is based on historical production data and dynamic adjustment of the time window according to wellhead operating conditions. The granularity of the time window can be set to the hourly or minute level. The candidate coding subset is selected from the quantum tracer coding pool based on the correspondence between well sections and phases in the injection log. The size of the coding combination is limited by the channel capacity of the detection equipment. The phase and sampling volume range are dynamically adjusted according to the phase distribution ratio of the wellhead fluid. The sampling volume of oil and water phases can be set to 100-250 mL, and the gas phase enrichment volume is converted from the filter membrane absorption volume or the eluent volume. The quaternion joint optimization variables generate a candidate scheme set through permutation and combination. The number of candidate schemes is limited by the size of the coding subset and the sampling time point density. Sample numbering and internal standard calibration channel configuration requirements adopt batch management, with at least one internal standard calibration channel set up for each batch for data calibration.

[0080] Specifically, candidate time points are generated at preset intervals within a time window, and an initial candidate set is formed by combining these with the codes of unexpired tracers in the coding pool. Based on real-time data of wellhead pressure, temperature, and water cut, the feasible range of sampling volume for each phase is dynamically adjusted. Quadruple variables are input into the optimization model to exclude schemes exceeding the laboratory's parallel processing capabilities. Each candidate scheme is assigned a unique number and configured with an internal standard calibration channel to ensure the traceability of the detection process. Redundant sampling points are reduced by jointly optimizing variables, prioritizing schemes with significant differences in coding combinations, comprehensive phase coverage, and moderate sampling volume to maximize information gain within budget and production capacity constraints. The final candidate set undergoes robustness screening, retaining schemes that meet the condition number threshold and coherence threshold, achieving efficient utilization of sampling resources and effective control of data quality.

[0081] As a preferred embodiment, the solution of this application is specifically implemented as follows: A set of candidate sampling time points is generated within the given time window of the sampling schedule. For example, for a multi-stage fracturing well, sampling time points can be set on days 1, 3, 7, 14, 30, 60, and 90 after production. A subset of candidate codes is generated by combining the quantum tracer coding pool and the injection log. Specifically, if there are 5 well segments and 3 codes are injected into each segment, the subset of candidate codes contains 15 codes.

[0082] For each candidate scheme, the corresponding phase and sampling volume range are given. For example, for oil wells, 200-250 mL of oil phase sample and 150-200 mL of water phase sample can be set. For gas wells, 2-5 L of gas phase enrichment sample can be set. The quadruple consisting of sampling time point, phase, sampling volume, and coding subset is used as a joint optimization variable to form a candidate set of optimal sampling and coding schemes. Furthermore, each candidate scheme is bound with a unique sample number and internal standard calibration channel configuration requirements.

[0083] Through the above technical solutions, this application achieves joint optimization of sampling time point, phase, sampling volume, and coding subset. Therefore, with limited sampling resources, more comprehensive and accurate output profile information can be obtained. Specifically, by optimizing the sampling time point, dynamic changes in output can be captured. By optimizing the phase and sampling volume, detection sensitivity can be improved. By optimizing the coding subset, cross-interference can be reduced. Furthermore, binding a unique sample number and an internal standard calibration channel improves sample traceability and data reliability.

[0084] This application further proposes an information gain assessment aimed at improving the effectiveness of operational management decisions. This includes using the improvement of decision-making quality in production allocation, phase closure, or repressurization as the primary objective, and introducing measures of the convergence speed of uncertainty in the phase-by-phase output profile and the reduction of water content risk as secondary objectives. For each candidate scheme, an information gain assessment is performed based on standardized data and the most recent execution results, comprehensively considering the magnitude of decision improvement and uncertainty reduction brought about by sampling.

[0085] The primary objective quantifies the improvement in output and economic indicators through production adjustments, shutdown operations, or re-pressurization. The efficiency of improving data confidence is reflected by the uncertainty convergence speed, and production stability is reflected by the reduction in moisture content risk. The information gain assessment of candidate schemes constructs a predictive model based on historical and current standardized data, simulating the impact of different sampling schemes on the decision-making scenario. The uncertainty convergence speed is calculated using the confidence interval width reduction rate, and the reduction in moisture content risk is measured by the probability change of moisture content exceeding a threshold.

[0086] Specifically, the information gain assessment process first sets the weights of the primary objective and secondary objectives. The weight of the primary objective is dynamically adjusted based on the economic priorities of the current production stage. After candidate schemes are generated, the changing trends of output profile uncertainty and moisture content distribution under different schemes are predicted based on standardized data and historical execution results. The uncertainty convergence rate is assessed by calculating the standard deviation reduction rate of the inversion results, and the reduction in moisture content risk is assessed by Monte Carlo simulation of the probability change of moisture content exceeding a preset threshold. The comprehensive score of each candidate scheme is obtained by a linear combination of the improvement value of the primary objective, the weighted value of the uncertainty convergence rate, and the weighted value of the reduction in moisture content risk. For example, if candidate scheme A increases the uncertainty convergence rate by 30% and reduces the probability of moisture content risk by 15%, while candidate scheme B increases the uncertainty convergence rate by 20% but reduces the probability of moisture content risk by 25%, then the scheme with the higher comprehensive score is selected based on the weight allocation. After the evaluation is completed, the scheme with the highest score is selected as the optimal sampling and coding scheme to ensure that the improvement in decision quality is maximized under budget and detection constraints.

[0087] As a preferred embodiment, the solution of this application is specifically implemented as follows: With improving the quality of decision-making regarding production allocation, phase closure, or repressurization as the primary objective, measures of the convergence speed of uncertainty in the phase-by-phase output profile and the reduction of water content risk are introduced as secondary objectives. For each candidate scheme, information gain is evaluated based on standardized data and the most recent execution results, comprehensively considering the magnitude of decision improvement and uncertainty reduction brought about by sampling.

[0088] Specifically, a decision quality assessment model is first established, comprising three sub-models: production allocation effect, shut-off effect, and repressurization effect. The production allocation effect sub-model considers yield increase, moisture content control, and production stability. The shut-off effect sub-model considers yield loss after shut-off, moisture content changes, and pressure response. The repressurization effect sub-model considers pressure recovery, yield recovery, and moisture content changes.

[0089] Secondly, an uncertainty convergence rate assessment model is constructed. This model, based on the Bayesian update principle, calculates the degree of variance reduction in the estimated output profile after each sampling. Simultaneously, a water content risk assessment model is established. This model, based on Monte Carlo simulation, calculates the probability of high water content occurring and the potential losses it may cause.

[0090] Furthermore, for each candidate solution, simulations were performed using the latest standardized data and historical execution results, employing the aforementioned model. Simulation results included indicators for decision quality improvement, uncertainty convergence speed, and water content risk reduction. These indicators were weighted and summed to obtain a comprehensive information gain score.

[0091] Finally, all candidate schemes are sorted according to their comprehensive information gain score, and the scheme with the highest score is selected as the optimal sampling and coding scheme.

[0092] Through the above technical solution, this application achieves information gain assessment oriented towards improving the effectiveness of operational management decisions. This allows for more accurate selection of the most valuable sampling scheme, improving sampling efficiency and decision quality. Simultaneously, by introducing uncertainty convergence speed and water-cut risk assessment, the robustness and risk control capabilities of the scheme are enhanced. This method avoids information redundancy or insufficiency that may result from fixed, empirical sampling, enabling optimal allocation of limited sampling resources, thereby supporting more scientific and efficient oil and gas field production management decisions.

[0093] This application further proposes a scheme to determine the optimal sampling and coding scheme under the constraints of budget, detection latency and laboratory capacity, and to set condition number thresholds and coherence thresholds for observation mapping relationships, and to configure internal standard calibration channels.

[0094] Budget constraints quantify the costs of sample collection, transportation, and testing into economic indicators by limiting the expenses of single and cumulative sampling, thus avoiding cost overruns. Testing latency constraints limit the time from sample collection to report generation within the allowable range of the process, preventing data corruption. Laboratory capacity constraints limit the upper limit of the number of samples that can be processed in parallel, ensuring the rational allocation of testing resources. Candidate schemes must simultaneously satisfy all three types of constraints to form a feasible region. The robustness index of the observation mapping relationship is calculated using the matrix condition number, reflecting the stability of data inversion. The condition number threshold is set within a preset range to screen schemes with satisfactory anti-interference capabilities. The coherence threshold is determined through correlation analysis between encoded subsets to avoid mutual interference between multiple codes. Internal standard calibration channels are configured in batches, with at least one channel per batch. Signal benchmark calibration is performed using standard beads of known strength to correct data link errors between the field and the laboratory. When a candidate scheme fails to meet the trigger threshold or exceeds capacity limits, channel merging is performed to reduce the number of codes, or the size of the encoded subset is reduced to decrease testing complexity, or the sampling time and volume are adjusted to adapt to resource constraints.

[0095] Specifically, after candidate schemes are generated, the cost of a single sampling and historical cumulative consumption are first calculated based on the budget model, eliminating over-cost schemes. Detection delay constraints are calculated by superimposing logistics time and laboratory processing cycle to exclude overtime schemes. Laboratory capacity constraints are dynamically adjusted based on the current number of samples to be tested and the maximum throughput of the equipment to filter overloaded schemes. Candidate schemes that pass the constraints enter the robustness evaluation stage, calculating the condition number of their observation mapping matrix. If it exceeds a preset threshold, it is judged as an ill-conditioned problem, triggering a channel merging operation to merge codes with high spectral overlap into combined channels. At the same time, the signal coherence between code subsets is calculated. If it exceeds a threshold, the subset size is reduced, prioritizing the retention of codes with high contribution. An internal standard calibration channel is forcibly configured in the first sample of each batch, and subsequent samples are randomly inserted into the calibration channel according to a preset ratio. The calibration data is used to correct instrument drift and operational deviations. When a candidate scheme cannot be executed due to resource constraints, a degradation strategy is adopted: the sampling time point is reallocated within the time window, and the interval between adjacent samples is extended. Alternatively, the single sampling volume is reduced, and data accuracy is compensated through multiple samplings. The final selected solutions, provided that all constraints and thresholds are met, are ranked according to their information gain evaluation scores, and the best solution is selected for execution. At the same time, execution deviation and resource consumption data are recorded for subsequent solution optimization.

[0096] As a preferred embodiment, the solution of this application is specifically implemented as follows: Budget limits the cost of single and cumulative sampling, detection latency limits the maximum allowable time from sample collection to report issuance, and laboratory capacity limits the maximum number of parallel samples; these factors collectively constitute the feasible region of candidate schemes. For each candidate scheme, a robustness index of the observation mapping relationship is calculated and compared with condition number thresholds and coherence thresholds; only candidate schemes that meet the threshold requirements are retained. Internal standard calibration channels are configured for selected schemes in batches, with at least one internal standard calibration channel set up for each batch for field-laboratory link calibration. When any candidate scheme triggers a threshold failure or exceeds capacity limits, a degradation strategy is implemented, including channel merging, reducing the encoding subset, or adjusting the sampling time and sample volume.

[0097] Specifically, the budget is limited to no more than 1,000 yuan per sampling and no more than 100,000 yuan in cumulative sampling. The detection latency is set to 48 hours, and the laboratory capacity is set to 20 samples per day. The condition number threshold for the observation mapping relationship is set to 100, and the coherence threshold is set to 0.8. One internal standard calibration channel is set for every 10 samples. When the trigger threshold is not met, the system first attempts to merge similar coding channels. If this still does not work, the number of coding subsets is reduced, and finally, the sampling time interval is adjusted or the sampling volume is reduced.

[0098] Through the above technical solutions, this application ensures the feasibility of the sampling scheme while improving the reliability of observation data by setting thresholds and internal standard calibration channels. Simultaneously, a degradation strategy ensures an effective sampling scheme under various constraints, improving the robustness and adaptability of the method. Furthermore, by rationally configuring the internal standard calibration channels, the consistency and traceability of data throughout the entire process from on-site sampling to laboratory testing are improved.

[0099] This application further proposes an execution process for determining the optimal sampling and coding scheme, including: ranking candidate schemes that meet the feasible region and threshold requirements according to information gain evaluation; selecting the scheme with the highest score as the optimal sampling and coding scheme and issuing execution instructions containing sampling time point, phase, sampling volume, coding subset, and internal standard calibration channel configuration. After sample submission, the detection latency and laboratory capacity utilization are monitored in real time. When a delay risk or capacity congestion is detected, a local reordering is triggered, and lower-priority schemes are replaced. After detection, the results are compared with the execution cost, execution deviation, and predicted values ​​of the information gain evaluation, and the comparison results are written into the parameter library for sampling scheduling and information gain evaluation.

[0100] The candidate scheme ranking is based on information gain evaluation scores, with score calculation comprehensively considering both the magnitude of decision improvement and uncertainty reduction. Execution instructions include coded subsets and calibration channel configurations to ensure traceability of the detection chain. Real-time monitoring employs preset delay risk thresholds and capacity utilization thresholds, automatically triggering alarms when detection progress deviates from the preset time window or when the laboratory's parallel processing capacity exceeds limits. Local re-ranking dynamically adjusts the candidate scheme priority list, replacing low-priority schemes with suboptimal schemes that meet current constraints. The results comparison stage quantifies the actual effect of the schemes by evaluating execution deviations; the deviation data is used to optimize the parameter weights of the information gain evaluation model.

[0101] Specifically, candidate solutions are first scored using an information gain evaluation model, which integrates quantitative indicators of decision quality improvement and uncertainty convergence speed. The solution with the highest score generates execution instructions containing specific sampling parameters and calibration channel configurations, which are then distributed to the field acquisition terminal and laboratory system via a digital platform. After sample submission, detection latency is monitored via a countdown mechanism, and laboratory capacity is monitored via real-time equipment status data. When the detection progress lags behind the preset timeline or the equipment load exceeds the parallel processing limit, a solution reordering mechanism is automatically triggered, generating a new priority list based on updated constraints. Replaced solutions must be re-bound to sample numbers and calibration channel configurations to ensure the integrity of the data link. After detection, the actual detection results are compared with the predicted values ​​using a deviation analysis algorithm. The deviation data is fed back to the information gain evaluation model to correct parameter weights and update the sampling schedule's time window and volume configuration. Through a closed-loop feedback mechanism, dynamic adaptation between sampling solutions and detection resources is achieved, improving data acquisition efficiency and decision reliability.

[0102] As a preferred embodiment, the specific implementation of this application's solution is as follows: When the wellhead pressure monitoring data exceeds a preset safety threshold, the candidate scheme generation module is automatically triggered. This module generates a set of candidate sampling time points within a 24-hour time window, and combines this with the non-invalid coding sequences in the quantum tracer coding pool to generate a candidate coding subset, forming a quadruple candidate set containing the sampling time point, phase, sampling volume, and coding subset. Each candidate scheme is bound to a unique sample number and configured with at least one internal standard calibration channel, wherein the calibration channel uses europium-doped silica fluorescent beads as the reference material. Candidate schemes are sorted according to information gain scores, and the scheme that can reduce water cut uncertainty and shorten the convergence speed of the production profile is prioritized as the optimal scheme for execution. During execution, the laboratory sample processing progress is monitored in real time. When the detection delay exceeds a preset time limit, the processing requests for low-priority samples are automatically paused, and the parallel detection queue is dynamically adjusted. After the detection is completed, the actual detection results are compared with the predicted gain value. Cases with a prediction deviation of more than 15% will be marked and the model parameters will be updated. The phase priority weight and encoding subset selection rules in the sampling schedule will be corrected simultaneously.

[0103] Through the above technical solutions, this application effectively solves the problems of resource waste and insufficient decision-making information caused by static sampling schemes in existing technologies. By dynamically optimizing the candidate scheme ranking mechanism and implementing a real-time monitoring and feedback mechanism, it ensures that data from key phases are acquired preferentially under conditions of limited testing resources, reducing the uncertainty of the production profile inversion results. The detection delay risk warning and queue adjustment functions avoid sample failure caused by laboratory capacity congestion, ensuring the integrity of the data link. The parameter self-correction mechanism further enhances the adaptability to complex operating conditions, enabling the sampling scheme to continuously match the dynamic characteristics of reservoir changes.

[0104] This application further proposes an execution process for determining the optimal sampling and coding scheme, including: ranking candidate schemes that meet the feasible region and threshold requirements according to information gain evaluation; selecting the scheme with the highest score as the optimal sampling and coding scheme and issuing execution instructions containing sampling time point, phase, sampling volume, coding subset, and internal standard calibration channel configuration. After sample submission, the detection latency and laboratory capacity utilization are monitored in real time. When a delay risk or capacity congestion is detected, a local reordering is triggered, and lower-priority schemes are replaced. After detection, the results are compared with the execution cost, execution deviation, and predicted values ​​of the information gain evaluation, and the comparison results are written into the parameter library for sampling scheduling and information gain evaluation.

[0105] The candidate scheme ranking is based on information gain evaluation scores, with score calculation comprehensively considering the improvement in decision quality and the reduction in uncertainty. When execution instructions are issued, sample numbers and calibration channel configurations are linked to ensure traceability of the execution process. Real-time monitoring uses preset delay risk thresholds and capacity occupancy thresholds, with triggering conditions determined by dynamically calculating laboratory processing progress and the number of samples to be tested. Local re-ranking only applies to samples that have not yet been tested, and replacement schemes must meet feasible region and threshold requirements. During the results comparison phase, the deviations between actual and predicted costs, and between execution deviations and predicted information gain, are written into a parameter library for subsequent adaptive adjustments to the evaluation model.

[0106] Specifically, after candidate solutions are generated, they are first prioritized based on information gain evaluation scores. The solution with the highest score is selected as the optimal solution, and an execution instruction is generated. The execution instruction includes specific sampling parameters and calibration channel configurations to ensure standardization of on-site operations and laboratory testing. After sample submission, the testing progress and laboratory load are continuously tracked. When the testing time exceeds a preset threshold or the number of parallel samples approaches the capacity limit, a local re-ranking mechanism is triggered. This mechanism re-evaluates the remaining candidate solutions and selects the suboptimal solution to replace potentially delayed testing tasks, avoiding overall progress disruptions. After testing is completed, the actual test results are compared with the predicted values ​​to analyze execution cost deviations and information gain deviations. The deviation data is fed back to the sampling scheduling and evaluation model parameter library for optimizing the generation and evaluation of subsequent solutions. Through dynamic adjustment and feedback mechanisms, information acquisition efficiency can be maximized under budget and capacity constraints, while improving the prediction accuracy and adaptability of the decision-making model.

[0107] As a preferred embodiment, the specific implementation of this application is as follows: When constructing a robust and optimized feasible domain, the safe range for wellhead pressure is set to 5-40 MPa, the safe range for temperature is set to 30-100℃, the upper limit for daily production per well is 5000 barrels, and the lower limit for well section production is 20% of the historical average. The step size for adjusting the throttle valve position is limited to no more than 10% each time, and the repressurization operation window is limited to when the formation pressure recovers to more than 80% of the initial value and the construction pump truck resources are available for dispatch. Environmental constraints require that the suspended solids concentration in the flowback fluid be less than 50 mg / L. The production profile and its uncertainty are used to generate 200 possible scenarios using the Monte Carlo method, among which the set of scenarios with a confidence level of 90% needs to cover the fluctuation range of the production profile. The water cut risk threshold is set as wellhead water cut exceeding 85%, and the production capacity fluctuation threshold is a daily production change of more than ±10% per well. The NSGA-II algorithm was used for optimization, with a primary objective of reducing the cost per barrel of oil equivalent by $0.50, and secondary objectives of reducing the probability of water cut risk to below 5% and accelerating the reduction of production profile uncertainty. Multi-objective optimization was performed on 10 candidate production allocation schemes. The optimal scheme generated operational instructions including adjusting the choke valve opening of well section A3 to 60%, closing the high water-cut section B5, implementing re-fracking operations in well section C2 with a pumping intensity of 5 m³ / min, and increasing the sampling frequency of well section D1 to twice daily. After the operations were executed, wellhead water cut data was collected in real time. When the actual water cut deviated from the expected value by more than 5%, the inversion target weight adjustment module was triggered, increasing the weight coefficient of the physical consistency term from 0.3 to 0.5, updating the phase sampling priority of well section B5 in the sampling schedule, and writing the execution deviation data into the historical database of information gain evaluation.

[0108] Through the above technical solutions, this application effectively solves the problems of insufficient robustness of decision-making schemes and inadequate model feedback mechanisms in existing technologies. By constructing a feasible domain with multi-dimensional constraints and risk control indicators, it ensures that the optimization results meet on-site safety and environmental protection requirements. A multi-scenario ensemble transformation method is adopted to fully consider the impact of output profile uncertainty on decision-making. A dynamic weight adjustment mechanism based on multi-objective optimization improves the economic efficiency and risk control capabilities of the production allocation scheme. Closed-loop feedback between execution results and model parameters enables adaptive optimization of the inversion objective and dynamic calibration of the sampling strategy, enhancing the responsiveness to changes in on-site conditions.

[0109] The above embodiments utilize background subtraction, crosstalk correction, and instrument drift correction to construct standardized data with a unified caliber, ensuring the quantitative comparability and verifiability of the observed data and the segment × phase output profile. Based on the inversion target output with uncertain segment × phase output profiles containing statistical fitting terms, physical consistency terms, and temporal smoothing sparsity terms, the output remains stable and converges faster under multiphase transport and slow-release perturbations. Information gain assessment uses sampling time point, phase, sampling volume, and coding subset as joint optimization variables, automatically providing the optimal sampling and coding scheme under constraints of budget, detection latency, and laboratory capacity. Simultaneously, condition number thresholds and coherence thresholds are set for the observation mapping relationship, and an internal standard calibration channel is configured to improve identifiability and result reliability, obtaining the most valuable information for decision-making with fewer samples and shorter testing cycles. Based on the output profile and its uncertainty, robust optimization of production allocation, shutdown or repressurization is performed and work instructions are generated. The execution results are written back to update the inversion target and sampling schedule, continuously reducing the risk of misadjustment and testing costs, improving unit cost output and decision consistency, and realizing quantifiable gains in production management and resource scheduling.

[0110] In another preferred embodiment based on the above embodiments, see [reference] Figure 2 As shown, this embodiment provides a quantum tracer-based oil and gas production profile prediction system for applying the above-mentioned quantum tracer-based oil and gas production profile prediction method, including: The configuration unit is configured to establish quantum tracer coding pools corresponding to well sections and oil / gas / water phases, and to complete the delivery according to the injection log during the sand-carrying fluid stage.

[0111] The acquisition unit is configured to collect and test wellhead samples according to the sampling schedule during the production period to obtain observation data for each code.

[0112] The preprocessing unit is configured to perform background subtraction, crosstalk correction, and instrument drift correction on the observation data to form standardized data.

[0113] The processing unit is configured to obtain a segment × phase output profile containing uncertainty from standardized data based on an inversion objective that includes statistical fitting terms, physical consistency terms, and temporal smoothing sparsity terms.

[0114] The processing unit is also configured to use the sampling time point, phase, sampling volume and coding subset of the next time period as joint optimization variables, adopt information gain evaluation with the goal of improving the effectiveness of operation and management decisions, determine the optimal sampling and coding scheme under the constraints of budget, detection latency and laboratory capacity, set condition number threshold and coherence threshold for observation mapping relationship, and configure internal standard calibration channel.

[0115] The execution unit is configured to execute robust optimization generation instructions for production allocation, section closure, or repressurization based on the output profile and its uncertainty, and write back the execution results to update the inversion target and sampling schedule.

[0116] Specifically, after the configuration unit completes the coding and delivery during the sand-carrying liquid stage, the acquisition unit generates instructions containing sampling time, phase, and volume based on the sampling schedule, binds sample numbers, and records operating information. The preprocessing unit performs standardization processing on the collected samples, where crosstalk correction uses a single coded standard sample and a spiked sample to calibrate the cross-influence matrix, and separates channel signals according to non-negative constraints through deconvolution. During the inversion process, the processing unit dynamically adjusts the weight of the physical consistency term based on the deviation of the internal standard calibration channel to ensure that the inversion results meet robustness requirements. When the optimal sampling and coding scheme is determined, the processing unit calculates the condition number and coherence index of the observation mapping relationship among the candidate schemes, retains only the schemes that meet the threshold requirements, and configures at least one internal standard calibration channel for each batch. After the work instructions are issued, the execution unit monitors the detection latency and laboratory capacity in real time. Delay risks trigger local reordering, and the detection results are compared with the predicted values ​​and written back to the sampling schedule. During the execution result writing back process, the weight parameters of the inversion target and the time and volume of the sampling schedule are dynamically updated based on the actual deviation, forming a closed-loop feedback mechanism.

[0117] As a preferred embodiment, the solution of this application is implemented as follows: The system configuration unit establishes a quantum tracer coding pool corresponding to the well section and oil / gas / water phase. The coding pool contains carbon quantum dot nanomaterials with different combinations of fluorescence wavelengths, and each well section corresponds to a specific wavelength combination. During the proppant-carrying fluid injection stage, codes are assigned according to the well section location and phase and recorded in the injection log. The proppant-carrying fluid and quantum tracer mixture is injected into the target formation through a high-pressure pumping device. During the production period, the acquisition unit receives sampling scheduling instructions and triggers an automatic sampling device according to the preset sampling time point. Oil phase samples and water phase samples are collected at the wellhead into light-proof inert containers, and gas phase samples are concentrated through a membrane enrichment device. Sample numbers, timestamps, and real-time operating parameters are synchronously uploaded to the database through an IoT module.

[0118] The preprocessing unit performs background subtraction on the acquired observation data, constructs a background signal database using blank samples, and subtracts background noise using the non-negative least squares method. Crosstalk correction is achieved by constructing a fluorescence spectral unmixing matrix, with matrix coefficients obtained through standard sample calibration. A sparse constraint optimization algorithm is used to separate overlapping signals. Instrument drift correction is dynamically adjusted based on real-time readings from the internal standard calibration channel, and a sliding window algorithm is used to smooth the drift trend. Standardized data is converted from raw signals to unit volume concentration values ​​through a normalization module.

[0119] The processing unit constructs an inversion objective function, which includes a statistical fitting term based on least squares, a physical consistency constraint term based on the mass balance equation, and a time smoothing term based on yield changes in adjacent time periods. The objective function is solved using the alternating direction multiplier method, and a stability check is performed after each iteration; if the residual exceeds a threshold, the regularization parameter is adjusted. Uncertainty assessment is achieved through Monte Carlo simulation, generating a yield probability distribution curve. The joint optimization module uses sampling time point, phase, sampling volume, and coded subset as decision variables, calculates the information gain value of each candidate scheme using an information entropy model, and constructs an integer programming model to solve for the optimal solution, combined with laboratory testing capability constraints.

[0120] The execution unit constructs a multi-objective optimization model based on the production probability distribution. The objective function includes economic benefits, water-cut risk, and operating costs, while constraints cover wellhead pressure safety thresholds and construction resource limitations. The optimization results generate production valve opening commands, section closing electrical control signals, and repressurization operation parameters. These commands are sent to field equipment via the industrial bus. The execution results are transmitted back to the system via an edge computing gateway, triggering inversion model weight updates and dynamic adjustments to the sampling schedule.

[0121] Through the above technical solutions, this application solves the problem of insufficient reliability in the mapping between observation data and output profiles in existing technologies, and realizes dynamic collaborative optimization of sampling schemes and coding strategies. Specifically, by constructing a joint optimization variable and information gain evaluation mechanism, key data with the greatest impact on decision-making is prioritized under limited detection resources, effectively improving the accuracy of production allocation and key segment decisions. Simultaneously, the introduction of internal standard calibration channels and condition number thresholds enhances the robustness of the observation system, reduces the impact of instrument drift and signal crosstalk on the inversion results, and brings the uncertainty of output profile evaluation to a convergence. The closed-loop feedback mechanism of execution results further optimizes the adaptive capability of the inversion model, forming a continuously improving intelligent decision-making chain.

[0122] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for predicting oil and gas production profiles based on quantum tracers, characterized in that, include: Establish quantum tracer coding pools corresponding to well sections and oil / gas / water phases, and complete the delivery according to the injection log during the sand-carrying fluid stage; During the production period, wellhead samples are collected and tested according to the sampling schedule to obtain observation data for each code. The wellhead samples include blank samples, field parallel samples, and spiked samples. The observation data are subjected to background subtraction, crosstalk correction, and instrument drift correction to form standardized data; Based on the inversion objective, which includes statistical fitting terms, physical consistency terms, and time smoothing sparsity terms, a segment × phase output profile containing uncertainty is obtained from the standardized data. The sampling time point, phase, sampling volume, and coding subset of the next period are used as joint optimization variables. Information gain evaluation with the goal of improving the effectiveness of operation and management decisions is adopted. Under the constraints of budget, detection delay, and laboratory capacity, the optimal sampling and coding scheme is determined. Condition number threshold and coherence threshold are set for the observation mapping relationship, and an internal standard calibration channel is configured. Based on the output profile and its uncertainty, robust optimization commands for production allocation, section closure, or repressurization are generated, and the execution results are written back to update the inversion objective and the sampling schedule; When performing background subtraction, crosstalk correction, and instrument drift correction on the observed data to form standardized data, the following steps are included: When performing background subtraction on the observation data, the process includes establishing a background library using the blank sample, determining the detection limit based on the background mean and dispersion of each coding channel, marking channels below the detection limit as indefinite, and subtracting the remaining channels based on the background mean. During crosstalk correction, the process includes calibrating the crosstalk effects between each coding channel using a single coded standard sample and the spiked sample to obtain a crosstalk matrix for correction; separating the signals of each channel according to the non-negative constraint deconvolution principle; and setting the regularization intensity based on the principle of minimizing the deviation of the internal standard calibration channel. When the robustness index of the observation mapping relationship exceeds the condition number threshold, channel merging is triggered. During instrument drift correction, the relative deviation between the current reading and the reference reading of the internal standard calibration channel is used as the basis to proportionally correct the readings of each channel after crosstalk removal, and a piecewise linear method is used to smooth the long-period drift. When generating standardized data, the process also includes normalizing the results of each channel based on the sample volume, dilution factor, and gas phase enrichment volume. Based on the inversion objective, which includes statistical fitting terms, physical consistency terms, and temporal smoothing sparsity terms, when obtaining the segment × phase output profile containing uncertainty from the standardized data, the process includes: Based on the observation mapping relationship, the standardized data is mapped to the segment × phase output profile, and non-negative constraints and yield upper limit constraints are applied. The statistical fitting term is set to measure the fitting bias; the physical consistency term is set, and its weight is adjusted according to the deviation of the internal standard calibration channel, the robustness index of the observation mapping relationship, and the condition number threshold; the temporal smoothing sparsity term is set. The inversion objective is solved using an iterative method with non-negative constraints. After each iteration, a stability check is performed based on the deviation between the robustness index and the internal standard calibration channel. If the threshold is not met, the regularization intensity is adjusted. Uncertainty assessment is performed on the obtained segment × phase output profile.

2. The method for predicting oil and gas production profiles based on quantum tracers according to claim 1, characterized in that, During the production period, wellhead samples are collected and tested according to the sampling schedule to obtain observation data for each code, including: Generate instructions that include sampling time, phase and sampling volume according to the sampling schedule and bind the sample number; At the wellhead, 100–250 mL of oil phase sample and water phase sample were collected in light-proof inert containers and the sample number, timestamp and operating condition information were attached. The operating condition information included wellhead pressure, temperature, instantaneous flow rate and water content. The gas phase sample was enriched by filtering through a membrane or water absorption and the enriched volume was recorded. Mechanical oscillation and ultrasonic dispersion are performed on the oil phase sample and the aqueous phase sample, and elution or desorption is performed on the gas phase enrichment sample to obtain the detection solution; the internal standard calibration channel is set up, and internal standard beads of constant count or known intensity are added to each sample and the internal standard channel reading is recorded; The quality control samples of blank samples, on-site parallel samples, and spiked samples were collected in sequence. All samples are stored under 2–8℃ light-shielding conditions and tested within a preset time limit; the test results are counted by timestamp × code, and the original results are combined with the readings of the internal standard channel to form the observation data; when the signal-to-noise ratio of any sample is lower than the threshold or the internal standard deviates from the threshold, the result of the sample is marked as invalid and resampling is triggered according to the sampling schedule.

3. The method for predicting oil and gas production profiles based on quantum tracers according to claim 1, characterized in that, When the sampling time point, phase, sampling volume, and encoded subset of the next time period are used as joint optimization variables, the following are included: Within the time window given by the sampling schedule, a set of candidate sampling time points is generated, and a subset of candidate codes is generated by combining the quantum tracer coding pool with the injection ledger; For each candidate scheme, the corresponding phase and sampling volume range are given; the quadruple consisting of sampling time point, phase, sampling volume and encoding subset is used as joint optimization variables to form the candidate set of the optimal sampling and encoding scheme, and each candidate scheme is bound with a unique sample number and the configuration requirements of the internal standard calibration channel.

4. The method for predicting oil and gas production profiles based on quantum tracers according to claim 3, characterized in that, When using information gain assessment aimed at improving the effectiveness of operational management decisions, it includes: With the primary objective of improving the decision-making quality of production allocation, closure, or repressurization, a measure of the convergence speed of uncertainty in the output profile of the segment × phase and the reduction of water content risk are introduced as secondary objectives. For each candidate scheme, an information gain evaluation is performed based on the standardized data and the most recent execution result, taking into account the magnitude of decision improvement and uncertainty reduction brought about by sampling.

5. The method for predicting oil and gas production profiles based on quantum tracers according to claim 4, characterized in that, When determining the optimal sampling and coding scheme under constraints of budget, testing latency, and laboratory capacity, and setting condition number and coherence thresholds for observation mapping relationships, and configuring internal standard calibration channels, the following should be included: The budget limits the cost of single and cumulative sampling, the detection delay limits the maximum allowable time from sample collection to report issuance, and the laboratory capacity limits the upper limit of the number of parallel samples, together constituting the feasible domain of candidate schemes. For each candidate scheme, the robustness index of the observation mapping relationship is calculated and compared with the condition number threshold and the coherence threshold, and only candidate schemes that meet the threshold requirements are retained. The internal standard calibration channel is configured for the selected schemes in batches, with at least one internal standard calibration channel set up for field-laboratory link calibration for each batch. When any candidate scheme triggers a threshold failure or exceeds the capacity limit, a degradation strategy of channel merging, reducing the encoding subset, or adjusting the sampling time and sampling volume is implemented.

6. The method for predicting oil and gas production profiles based on quantum tracers according to claim 5, characterized in that, The execution process for determining the optimal sampling and coding scheme includes: sorting the candidate schemes that meet the feasible region and threshold requirements according to the information gain evaluation, selecting the one with the highest score as the optimal sampling and coding scheme, and issuing an execution instruction containing the sampling time point, phase, sampling volume, coding subset and the configuration of the internal standard calibration channel. After sample submission, the detection latency and laboratory capacity utilization are monitored in real time. When a delay risk or capacity congestion is detected, a local reordering is triggered and a low-priority scheme is replaced. After the detection is completed, the results are compared with the execution cost, execution deviation and the predicted value of the information gain assessment, and the comparison results are written into the parameter library of the sampling schedule and the information gain assessment.

7. The method for predicting oil and gas production profiles based on quantum tracers according to claim 6, characterized in that, When generating robust optimization instructions for production allocation, shutdown, or repressurization based on the output profile and its uncertainty, and writing back the execution results to update the inversion objective and the sampling schedule, the process includes: The robust and optimized feasible region is constructed, which consists of at least the safe range of wellhead pressure and temperature, the upper and lower limits of production of a single well and well section, the step size of throttling and valve position adjustment, the repressurization operation window and construction resource constraints, and environmental and safety constraints. The output profile and its uncertainty are transformed into a set of scenarios or intervals for decision-making according to a preset confidence level, and water content risk threshold and production capacity fluctuation threshold are set as risk control indicators. With the primary objective of increasing unit cost output and economic indicators, and with the secondary objective of reducing water content risk and uncertainty convergence speed, robust optimization is performed on candidate production allocation schemes, key section lists, and repressurization section combinations to output the optimal scheme that satisfies the feasible region and the risk control indicators. The optimal solution is used to generate the operation instruction, which includes at least the target output value, throttling or valve position setting, list of closed sections, repressurization section and pumping intensity, sampling encryption requirements and configuration of the internal standard calibration channel. During operation, actual output, moisture content, and pressure data are collected and compared with the observed data. The expected results of the operation instructions are compared with the actual results. When the deviation exceeds a threshold, the data is written back according to preset rules. Adjust the weights of the statistical fitting term, physical consistency term, and temporal smoothing sparsity term in the inversion objective, update the sampling time point, phase, and sampling volume of the sampling schedule, and specify the encoding subset. At the same time, write the execution cost, execution deviation, and uncertainty changes into the parameter library of the information gain evaluation.

8. A quantum tracer-based oil and gas production profile prediction system, used to apply the quantum tracer-based oil and gas production profile prediction method as described in any one of claims 1-7, characterized in that, include: The configuration unit is configured to establish a quantum tracer coding pool corresponding to the well section and oil / gas / water phase, and to complete the delivery according to the injection log during the sand-carrying fluid stage. The acquisition unit is configured to collect and test wellhead samples according to the sampling schedule during the production period to obtain observation data for each code. The preprocessing unit is configured to perform background subtraction, crosstalk correction, and instrument drift correction on the observation data to form standardized data; The processing unit is configured to obtain a segment × phase output profile containing uncertainty from the standardized data based on an inversion objective that includes statistical fitting terms, physical consistency terms, and temporal smoothing sparsity terms. The processing unit is also configured to use the sampling time point, phase, sampling volume and coding subset of the next time period as joint optimization variables, adopt information gain evaluation with the goal of improving the effectiveness of operation and management decisions, determine the optimal sampling and coding scheme under the constraints of budget, detection delay and laboratory capacity, set condition number threshold and coherence threshold for observation mapping relationship, and configure internal standard calibration channel. The execution unit is configured to execute robust optimization generation instructions for production allocation, closure, or repressurization based on the output profile and its uncertainty, and write back the execution results to update the inversion target and the sampling schedule.

Images