Pressure transient data screening method based on seepage physical constraint

Abstract

The invention discloses a pressure transient data screening method based on seepage physical constraint, comprising the following steps: S1, acquiring multi-source monitoring data of a target well, and performing standardized reconstruction processing on the multi-source monitoring data to obtain a standardized time series data set; s2, extracting a potential pressure transient response sequence from the standardized time sequence data set based on transient event identification of a rheological state; s3, monotonicity verification is carried out on the potential pressure transient response sequence based on seepage physical constraints, sequences which do not pass the monotonicity verification are removed, and a verification sequence which passes the monotonicity verification is obtained; s4, multi-dimensional quality evaluation based on response integrity is carried out on the verification sequence, optimization is carried out according to a multi-dimensional quality evaluation result, and an optimized pressure transient response sequence is obtained. According to the method, high-quality pressure transient data can be obtained through efficient screening, the objectivity and accuracy of oil reservoir dynamic analysis are remarkably improved, and a high-quality data basis is provided for reservoir parameter inversion.

Description

Technical Field

[0001] This invention relates to the fields of oil and gas field development and reservoir engineering technology, and in particular to a method for screening pressure transient data based on seepage physical constraints. Background Technology

[0002] Transient pressure analysis is a core tool in reservoir engineering for assessing reservoir dynamics and wellbore conditions. By fitting a mathematical model to the transient response signal of bottomhole pressure, key parameters such as reservoir permeability, skin factor, well-reservoir coefficient, and reservoir boundary characteristics can be obtained through inversion. These parameters are crucial for guiding oil and gas reservoir reserve assessment, production prediction, and development scheme optimization. The theoretical basis of transient pressure analysis is built upon seepage mechanics models; therefore, the accuracy of the analysis results and the uniqueness of the solution are inherently highly dependent on the signal-to-noise ratio and physical integrity of the selected well test interpretation data segment. An ideal, effective well test data segment typically needs to include a sufficiently long preceding stable flow history, followed by a shut-in response interval with a sufficiently long duration and a monotonically stable pressure recovery or decline trend.

[0003] In the modern production and operation of digital oilfields, various automated monitoring systems continuously collect and accumulate massive amounts of long-term, high-frequency dynamic production data. This data, sequenced over time and often spanning years or even decades, constitutes valuable reservoir data assets. However, actual production data is often accompanied by frequent adjustments to operating conditions, equipment noise interference, data gaps, and complex wellbore phase transition effects. Accurately extracting high-quality transient response segments that conform to Darcy flow patterns from these long-span, unstructured, and noisy massive monitoring signals is a prerequisite for large-scale reservoir dynamic evaluation and refined management.

[0004] Currently, the preprocessing and screening of well test data in the industry mainly relies on manual interactive operations by domain engineers. This traditional operating mode exposes significant technical limitations when faced with massive amounts of high-frequency data. First, for high-frequency data accumulated over several years at the second or minute level, manual visual inspection and extraction of effective data segments are extremely time-consuming, becoming a serious bottleneck restricting the efficiency of reservoir dynamic analysis. Second, different engineers have significant experience-based biases in their identification of characteristics such as flow stability, pressure stability, and sufficient test duration, resulting in a lack of unified quantitative standards for screening results, which seriously affects the consistency and comparability of subsequent reservoir parameter inversion results. At the same time, pressure and flow data from different monitoring equipment and at different times often have significant differences in storage format, sampling frequency, and time alignment, making manual data cleaning, format unification, and time alignment tedious and prone to errors. More importantly, for pseudo-smooth curves affected by severe phase separation, gravity differentiation, or weak instrument zero drift in the wellbore, manual visual inspection often fails to detect their inherent non-monotonic physical distortions. Misusing such data that does not conform to the laws of seepage physics for fitting will lead to incorrect reservoir understanding.

[0005] Furthermore, raw field monitoring data typically has an extremely high sampling frequency, resulting in a massive amount of data for a single well test event. Directly using the full dataset for nonlinear regression analysis not only leads to computational overload, but also, in many commercial well test analysis software programs, the excessively high density of data points can interfere with the calculation and smoothing of derivative curves. Existing data reduction techniques mostly employ traditional methods such as equal-interval thinning or simple threshold filtering. These methods mechanically discard data, easily losing key derivative feature points in the early, drastic change zones reflecting wellbore reservoir and skin effect, or retaining a large amount of redundant data in pressure-stabilized zones. They fail to achieve the optimal balance between data reduction and preservation of key flow characteristics.

[0006] In summary, the oil and gas field development sector urgently needs an intelligent data processing solution based on the constraints of seepage physical characteristics. This solution should be able to efficiently and objectively identify valid pressure transient events that conform to physical laws from heterogeneous monitoring signals, and perform feature-preserving data optimization, thereby providing a high-quality, standardized data foundation for subsequent high-precision pressure transient analysis. Summary of the Invention

[0007] To address the aforementioned problems, this invention aims to provide a method for filtering pressure transient data based on seepage physical constraints.

[0008] The technical solution of the present invention is as follows: A method for filtering transient pressure data based on seepage physical constraints includes the following steps: S1: Acquire multi-source monitoring data of the target well, and perform standardized reconstruction processing on the multi-source monitoring data to obtain a standardized time series dataset; S2: Transient event identification based on rheological state: Extract potential stress transient response sequences from the standardized time-series dataset; S3: Based on the seepage physical constraints, perform monotonicity verification on the potential pressure transient response sequence, remove the sequences that fail the monotonicity verification, and obtain the verification sequences that pass the monotonicity verification. S4: Perform a multidimensional quality evaluation on the verification sequence based on response integrity, and optimize the sequence according to the multidimensional quality evaluation results to obtain an optimized pressure transient response sequence.

[0009] Preferably, in step S3, the monotonicity is checked using the following formula: (1) In the formula: It is the morphological monotonicity index; The total number of data points in the potential pressure transient response sequence; For indicator functions; For physical direction multipliers; Let be the pressure value of the (i+1)th data point; Let i be the pressure value of the i-th data point; For noise tolerance; If the morphological monotonicity index of the sequence If the value is less than the morphological monotonicity index threshold, the sequence fails the monotonicity check; otherwise, it passes.

[0010] Preferably, the physical direction multiplier The physical direction multiplier is determined based on the target well's operational type. When the target well's operational type is a production well, the physical direction multiplier... =+1, when the target well's operation type is an injection well, the physical direction multiplier =-1.

[0011] Preferably, in step S4, the quality indicators for multidimensional quality evaluation include the duration of shut-in / injection cessation, the magnitude of pressure change, the duration of flow, and the magnitude of flow pressure change.

[0012] As a preferred method, a multidimensional quality evaluation based on response completeness is performed using the following formula: (2) In the formula: For the overall quality score; The total number of quality indicators; The weight of the k-th quality indicator; It is a normalization function; This represents the original value of the k-th quality indicator for the j-th data point. When making the best selection, the sequences are sorted according to the comprehensive quality score, and the sequences with scores exceeding the comprehensive quality score threshold are selected as the preferred pressure transient response sequences.

[0013] Preferably, the following steps are also included: S5: Perform fidelity thinning on the preferred pressure transient response sequence to obtain a sparse sample set.

[0014] Preferably, in step S5, a hybrid sampling strategy combining the uniformity of logarithmic time distribution and the sensitivity of pressure derivative characteristics is used for fidelity thinning.

[0015] As a preferred method, the high-fidelity thinning process specifically includes the following sub-steps: S51: Perform a logarithmic transformation in the time domain to construct a uniform resampling grid based on a logarithmic time scale; S52: Introduces a pressure gradient sensitive mechanism to detect the pressure derivative change characteristics between adjacent sampling points. When the pressure change gradient exceeds a preset threshold, adaptive encrypted sampling is performed within the corresponding original signal interval.

[0016] The beneficial effects of this invention are: This invention significantly improves the processing efficiency and evaluation objectivity of massive monitoring data, abandons the traditional manual interactive screening mode, and achieves high-throughput processing of long-term, high-frequency monitoring signals by constructing a standardized identification process based on seepage physical characteristics. It completely eliminates subjective bias and uncertainty caused by differences in human experience, and ensures high reproducibility and consistency of screening results under different batches and different personnel operations.

[0017] Meanwhile, this invention enhances the physical authenticity and broad applicability of the screening results, and innovatively introduces an adaptive seepage physical constraint mechanism, which can intelligently identify and be compatible with two completely different physical response processes of production wells and injection wells, and has strong versatility; through the quantitative verification of the morphological monotonicity index, it effectively eliminates distorted data affected by wellbore phase change, gravity differentiation or instrument drift, and ensures that the selected data strictly follows the Darcy flow law.

[0018] Furthermore, this invention can achieve the best balance between data size reduction and preservation of key seepage characteristics. In response to the shortcomings of traditional downsampling methods that are prone to losing key information, an adaptive downsampling strategy is adopted. This strategy can significantly reduce the data size while fully preserving the key pressure derivative shape that reflects wellbore reservoir, skin effect and boundary characteristics. This provides a high-quality and standardized data foundation for subsequent high-precision reservoir parameter inversion, thereby directly improving the accuracy and reliability of well test interpretation results. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the pressure transient data screening method based on seepage physical constraints of the present invention. Detailed Implementation

[0021] The present invention will be further described below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and technical features described in this application can be combined with each other. It should also be pointed out that, unless otherwise indicated, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terms "comprising" or "including" and similar words used in this invention refer to elements or objects preceding the word that encompass the elements or objects listed following the word and their equivalents, without excluding other elements or objects.

[0022] like Figure 1 As shown, this invention provides a method for filtering pressure transient data based on seepage physical constraints, comprising the following steps: S1: Obtain multi-source monitoring data of the target well, and perform standardized reconstruction processing on the multi-source monitoring data to obtain a standardized time series dataset.

[0023] In the modern production and operation of digital oilfields, the same production well often has various automated monitoring systems continuously collecting and accumulating massive amounts of long-term, high-frequency production dynamic data (including bottomhole pressure time-series signals and surface flow time-series signals for oil and gas wells). Therefore, the monitoring data for this well may come from multiple sources. These multi-source monitoring data often exhibit significant differences in storage format, sampling frequency, and time alignment. This step, through standardization and reconstruction, generates a standardized time-series dataset with a unified format suitable for subsequent analysis.

[0024] S2: Transient event identification based on rheological state extracts potential stress transient response sequences from the standardized time series dataset.

[0025] Transient event identification based on rheological state identifies the response interval of bottom hole fluid switching from a flowing state to a shut-in / injection state based on the step change characteristics of the flow rate time series signal. Based on a preset minimum duration and differential pressure response threshold, potential pressure transient response sequences are extracted. Specifically, based on the comparison between the flow rate value and the preset flow rate threshold, the time series data is discretized into flowing and non-flowing states. Then, based on the state switching pattern, all candidate data sequences that meet the basic duration and pressure change constraints are scanned and identified.

[0026] S3: Based on the seepage physical constraints, perform monotonicity verification on the potential pressure transient response sequence, remove the sequences that fail the monotonicity verification, and obtain the verification sequences that pass the monotonicity verification.

[0027] In a specific embodiment, monotonicity is checked using the following formula: (1) In the formula: It is the morphological monotonicity index; The total number of data points in the potential pressure transient response sequence; For indicator functions; For physical direction multipliers; Let be the pressure value of the (i+1)th data point; Let i be the pressure value of the i-th data point; For noise tolerance; If the morphological monotonicity index of the sequence If the value is less than the morphological monotonicity index threshold, the sequence fails the monotonicity check; otherwise, it passes.

[0028] In the above embodiments, when the indicator function When the conditional expression within the parentheses is true, the indicator function... The value is 1 if it is not 1, and 0 otherwise. The physical direction multiplier The determination is based on the target well's operational type. When the target well's operational type is "production well," =+1, when the target well's operation type is "injection well", =-1; By matching the corresponding physical direction multiplier with the target well's operation type, an adaptive physical constraint criterion can be constructed, including the production well pressure recovery model and the injection well pressure drop model, thereby achieving seepage-based physical constraints. The noise tolerance... is a preset small positive tolerance for resisting data noise, which is a configurable parameter. This value is preset by technicians according to the noise level of on-site data, and this value determines the tolerance of the algorithm to "non-physical fluctuations". The larger the noise tolerance value, the less sensitive it is to abnormal data, and it may wrongly allow distorted data that should have been excluded to enter subsequent analysis. The smaller the noise tolerance value, it may be overly sensitive to abnormal data, and normal high-frequency measurement noise may be misjudged.

[0029] In a specific embodiment, the noise tolerance is 0.005 - 0.02 MPa, and this value is set based on the typical noise level of deep-sea high-precision pressure gauges.

[0030] In the above embodiment, the morphological monotonicity index can characterize the physical stability of the pressure response, so that distorted sequences that do not conform to the laws of seepage mechanics can be excluded based on it.

[0031] In a specific embodiment, the threshold of the morphological monotonicity index is set to 0.80. If ≥0.80, it means that the instantaneous change directions of more than 80% of the data points in this data segment are correct, the curve shape is good and the monotonicity is strong, and it is judged as qualified. If <0.80, it means that there are a large amount of noises, oscillations or reverse trends in this data segment, the curve shape is not good, and it is judged as unqualified and excluded. It should be noted that the threshold of the morphological monotonicity index is a manually set value. The higher the value, the stricter the screening criteria for data, and it can be flexibly selected according to the actual sample quality and volume.

[0032] S4: Perform multi-dimensional quality evaluation on the verification sequence based on response integrity, and perform optimization according to the multi-dimensional quality evaluation results to obtain an optimized pressure transient response sequence.

[0033] In a specific embodiment, the quality indicators of the multi-dimensional quality evaluation include shut-in / injection shut-off duration, pressure change amplitude, flow duration, and flow pressure change amplitude.

[0034] In the above embodiment, the shut-in / injection shut-off duration can characterize the test sufficiency, the pressure change amplitude can characterize the degree of formation energy recovery / dissipation, and the flow duration and the flow pressure change amplitude can characterize the stability of the previous flow history. The present invention uses these four quality indicators for multi-dimensional quality evaluation, which can comprehensively ensure that the selected data has sufficient detection depth (determined by the shut-in duration) and high signal-to-noise ratio (determined by the pressure difference amplitude), and at the same time ensure that the previous flow history meets the theoretical assumptions of the constant production rate and superposition principle of the well test model, thereby significantly reducing the multi-solution nature of the interpretation results and improving the accuracy of reservoir parameter inversion.

[0035] In a specific embodiment, a multidimensional quality assessment based on response completeness is performed using the following formula: (2) In the formula: For the overall quality score; The total number of quality indicators; The weight of the k-th quality indicator; It is a normalization function; This represents the original value of the k-th quality indicator for the j-th data point. When making the best selection, the sequences are sorted according to the comprehensive quality score, and the sequences with scores exceeding the comprehensive quality score threshold are selected as the preferred pressure transient response sequences.

[0036] In the above embodiments, the weights of each quality indicator are manually set values, and the sum of all weights is 1.

[0037] In one specific embodiment, the pressure transient data filtering method based on seepage physical constraints of the present invention further includes the following steps: S5: Perform fidelity thinning on the preferred pressure transient response sequence to obtain a sparse sample set.

[0038] In the above embodiments, fidelity-preserving thinning is a technique that preserves the key features of the original data during the data thinning process, which can minimize information loss while reducing the amount of data.

[0039] In one specific embodiment, a hybrid sampling strategy combining logarithmic time distribution uniformity with pressure derivative characteristic sensitivity is employed for fidelity thinning. In this embodiment, this strategy generates a sparsified sample set that retains key seepage characteristics.

[0040] In a specific embodiment, fidelity thinning specifically includes the following sub-steps: S51: Perform a logarithmic transformation in the time domain to construct a uniform resampling grid based on a logarithmic time scale; S52: Introduces a pressure gradient sensitive mechanism to detect the pressure derivative change characteristics between adjacent sampling points. When the pressure change gradient exceeds a preset threshold, adaptive encrypted sampling is performed within the corresponding original signal interval.

[0041] In the above embodiments, step S51 can ensure the uniformity of data point distribution on the double logarithmic diagnostic curve, and step S52 can retain the early high-frequency characteristics reflecting wellbore reservoir and skin effect.

[0042] In a specific embodiment, taking the long-term monitoring data of a permanent pressure gauge from a complex fault-block offshore oil reservoir in Nigeria as an example, the pressure transient data screening method based on seepage physical constraints described in this invention is used to process the data. In this embodiment, the reservoir has complex geological conditions, and the well types include production wells and injection wells. Its monitoring data has the following typical characteristics: the monitoring period is over five years, an extremely long span, and the data acquisition frequency is high (one data point every 30 seconds), resulting in a huge volume of raw data text files for a single well, typically containing millions of lines of data. During the long-term production and injection process, the wells underwent hundreds of start-up and shutdown operations, generating a large amount of pressure recovery or pressure drop test data, but most of this data is of poor quality due to insufficient time and drastic data fluctuations. Traditional manual screening methods are not only extremely time-consuming but also difficult to guarantee the objectivity and consistency of the screening results.

[0043] The pressure transient data filtering method based on seepage physical constraints described in this invention specifically includes the following steps: (1) Standardize and reconstruct the multi-source monitoring data of the target well to obtain a standardized time series dataset. Even when different working wells originate from the same reservoir, their raw data still suffer from inconsistent formats, disorganized header labels, and time asynchrony. In this embodiment, the raw data text file from well A1 in the target block, spanning nearly five years and containing over five million lines of data, is processed first. The raw monitoring file is exported from the oilfield database. Based on a pre-defined keyword library (containing "Time", "Pressure", "BHP", "Rate", "Q", etc.), the raw monitoring file undergoes semantic mapping and cleaning, removing outliers. The dynamic production data of well A1 before cleaning is shown in Table 1, and the dynamic production data of well A1 after cleaning is shown in Table 2.

[0044] Table 1 Dynamic production data of well A1 before cleaning

[0045] Table 2 Dynamic production data of Well A1 after cleaning

[0046] (2) Identification of transient events based on rheological state and extraction of potential pressure transient response sequences In this embodiment, a traffic status discrimination threshold is set. =1.0 m 3 / d. Based on the step change characteristics of the flow signal (i.e., from Mutation It identifies the physical moment when the fluid switches from "steady flow" to "shutdown / injection shutdown".

[0047] After comprehensive identification, 329 potential pressure transient response sequences were extracted from production well A1, and 286 potential pressure transient response sequences were extracted from injection well B1. Meanwhile, based on preset physical constraints (minimum shut-in time > 2 hours, minimum differential pressure > 0.5 MPa), invalid short-term fluctuation sequences were initially eliminated.

[0048] (3) Calculate the morphological monotonicity index of each sequence according to formula (1), and remove "pseudo-data" according to the morphological monotonicity index to obtain the verification sequence that passes the monotonicity check. In this embodiment, the seepage physical direction multiplier is adaptively matched according to the operational attributes of the target well. : For production well A1: Set the operation type to "Production" and match the physical direction multiplier. =+1 (indicating that the pressure should resume its upward trend after the well is shut in); For injection well B1: Set the job type to "Injection" and match the physical direction multiplier. =-1 (indicating that the pressure should show a downward trend after injection is stopped); Setting noise tolerance =0.005 MPa, and the physical confidence threshold (i.e., the morphological monotonicity index threshold) is set to 0.80. To visually demonstrate the screening effect and physical discrimination mechanism of this step, five representative typical time segments were selected from the 329 candidate sequences of production well A1 for detailed display, and the results are shown in Table 3: Table 3. Monotonicity verification results of some sequences

[0049] Through batch calculation and screening of the full dataset, 49 sequences out of the original 329 candidate sequences for production well A1 were deemed "qualified" because their indices exceeded the preset threshold of 0.80. These sequences were recognized as valid physical events conforming to the Darcy flow law and were allowed to proceed to the next stage of quality scoring. The remaining 280 sequences were automatically removed by the system due to morphological distortion or inconsistent physical characteristics. This fully demonstrates that the present invention can accurately identify and remove various implicit physical noises when processing massive amounts of unstructured data, ensuring the physical authenticity of the data in subsequent analysis.

[0050] (4) Calculate the comprehensive quality score of each sequence according to formula (2), and sort and optimize according to the comprehensive quality score. The comprehensive quality score is calculated for the verification sequences (49 production wells A1 and 42 injection wells B1) obtained in step (3) that have passed the monotonicity check. In this embodiment, the quality indicators of multidimensional quality evaluation include shut-in / injection stop duration (M1), pressure change amplitude (M2), flow duration (M3), and flow pressure change amplitude (M4). For the above quality indicators with different physical units, the minimum-maximum scaling method is used to eliminate the difference in dimensions. Based on the high sensitivity of well test analysis to the quality of shut-in section data, a weight vector W=[0.4,0.4,0.1,0.1] focusing on shut-in data is set.

[0051] After fully scoring and ranking the 49 qualified sequences of production well A1, in order to intuitively demonstrate the discriminative power of the evaluation model, typical sequences with high and low rankings were selected for comparison and display. The specific data is shown in Table 4: Table 4. Overall quality scores of 49 qualified sequences for production well A1.

[0052] Based on the overall quality score The values ​​are sorted in descending order. In this embodiment, the top 5 test events of the highest quality were selected for well A1 (i.e., the top 5 sequences were selected). The results are highly consistent with the optimal results obtained by experienced engineers through manual screening at great expense, and even include high-quality events that may have been missed during manual screening, which fully demonstrates the objectivity and efficiency of the present invention.

[0053] (5) Perform fidelity thinning on the preferred pressure transient response sequence obtained in step (4). Taking the top-ranked sequence (original monitoring data sampled at a frequency of 30 seconds / sample, total duration 55.2 hours, containing 6,624 original data points) as an example, adaptive downsampling based on key feature preservation is performed. Given the extremely high computational redundancy when using the full dataset directly in the analytical model, and the fact that late-stage high-frequency small fluctuations severely interfere with the smoothness of the pressure derivative curve, this step employs a hybrid strategy of "logarithmic time domain + pressure gradient sensitivity" to perform fidelity-preserving data thinning.

[0054] Specifically, firstly, taking the shut-in moment as the zero point, the flow time Δt for each sampling point in the original dataset is calculated. Then, the logarithmic sampling density is set to 20 points per log cycle, and the logarithmic step size is calculated, thus generating a series of theoretically logarithmically evenly spaced nodes on the time axis. Subsequently, the actual sampling points in the original dataset that are closest to these theoretical node times are found and retained. This step ensures that the data points are evenly distributed along the horizontal axis on the double logarithmic well test chart.

[0055] Next, recursive encryption based on pressure gradients is performed. The generated sequence of real sampling points is traversed, and the absolute value of the pressure difference |ΔP| between two adjacent real sampling points is calculated. A pressure change sensitivity threshold δ = 0.02 MPa is set. If |ΔP| > δ between two adjacent points, it indicates that the pressure change is drastic during that time period, and the real sampling point has lost key features. At this point, the original dataset is backtracked, and the original data point at the middle position in the time interval corresponding to the two skeleton points is extracted and inserted into the sequence. The process of "calculating the difference - comparing the threshold - inserting data" is repeated for the newly formed interval until the pressure difference between all adjacent points is less than the threshold or the minimum sampling interval is reached.

[0056] After processing, the number of data points was precisely reduced from 6,624 to 385. In the first 0.1 hours of well shut-in, due to the drastic pressure changes, the sampling points were automatically retained at a very high density, perfectly reproducing the "hump" shape of the pressure derivative curve. In the later 50 hours of well shut-in, as the pressure tended to stabilize, the sampling points automatically became sparse, thus physically eliminating high-frequency sawtooth noise.

[0057] Finally, a standard format file containing these 385 key points was output. Using any well test interpretation software, this file was loaded for nonlinear regression fitting. After importing the data into the mainstream well test interpretation software KAPPA, the formation permeability of well A1 was successfully derived to be 1393 md, the skin factor to be 3.6, and the formation factor to be 50125 md·m. This result is in high agreement with the parameters extrapolated from previous geological static modeling and well tests involving interference from adjacent wells.

[0058] In summary, this invention can efficiently screen and obtain high-quality transient pressure data, significantly improving the objectivity and accuracy of reservoir dynamic analysis and providing a high-quality data foundation for reservoir parameter inversion. Compared with existing technologies, this invention represents a significant advancement.

[0059] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A pressure transient data screening method based on seepage physical constraints, characterized in that, The method comprises the following steps: S1: obtaining multi-source monitoring data of a target well, and performing standardized reconstruction processing on the multi-source monitoring data to obtain a standardized time series data set; S2: extracting a potential pressure transient response sequence from the standardized time series data set based on transient event identification of rheological state; S3: performing monotonicity checking on the potential pressure transient response sequence based on seepage physical constraints, eliminating sequences that fail the monotonicity checking, and obtaining a verified sequence that passes the monotonicity checking; S4: performing multi-dimensional quality evaluation based on response completeness on the verified sequence, and performing optimization according to the multi-dimensional quality evaluation result to obtain an optimized pressure transient response sequence.

2. The pressure transient data screening method based on seepage physical constraints of claim 1, wherein, In step S3, the monotonicity checking is performed by the following formula: （1） wherein: is a monotony index; is the total number of data points in the potential pressure transient response sequence; is an indicator function; is a physical direction multiplier; is the pressure value of the i+1th data point; is the pressure value of the ith data point; is a noise tolerance; If the morphological monotonicity index of the sequence If the value is less than the morphological monotonicity index threshold, the sequence fails the monotonicity check; otherwise, it passes.

3. The pressure transient data screening method based on seepage physical constraints of claim 2, wherein, the physical direction multiplier is determined according to the operation type of the target well, when the operation type of the target well is a production well, the physical direction multiplier = +1, when the operation type of the target well is an injection well, the physical direction multiplier = -1.

4. The pressure transient data screening method based on seepage physical constraints of claim 1, wherein, In step S4, the quality indicators of the multi-dimensional quality evaluation include shut-in / drainage duration, pressure change amplitude, flow duration, and flow pressure change amplitude.

5. The pressure transient data screening method based on seepage physical constraints of claim 4, wherein, The multi-dimensional quality evaluation based on response completeness is performed by the following formula: （2） wherein: is the overall quality score; is the total number of quality indicators; is the weight of the kth quality indicator; is a normalization function; is the raw value of the kth quality indicator for the jth data. When performing optimization, the sequences whose scores exceed a comprehensive quality score threshold are selected as the optimized pressure transient response sequence according to the sorting of the comprehensive quality scores.

6. The pressure transient data screening method based on physical constraints of flow percolation according to any one of claims 1-5, characterized in that, The method further comprises the following steps: S5: performing fidelity thinning on the optimized pressure transient response sequence to obtain a sparse sample set.

7. The pressure transient data screening method based on seepage physical constraints of claim 6, wherein, In step S5, a hybrid sampling strategy combining log time distribution uniformity and pressure derivative characteristic sensitivity is used to perform fidelity thinning.

8. The pressure transient data screening method based on seepage physical constraints of claim 7, wherein, The fidelity thinning specifically comprises the following sub-steps: S51: performing log transformation in the time domain to construct a uniform resampling grid based on a log time scale; S52: introducing a pressure gradient sensitive mechanism to detect the pressure derivative change characteristics between adjacent sampling points, and when the pressure change gradient exceeds a preset threshold, performing adaptive encryption sampling in the corresponding original signal interval.

Images