A repeated fracturing production data reconstruction method and system fusing physical priori

Abstract

The present application belongs to the field of oil and gas field development engineering and data processing, and discloses a repeated fracturing production data reconstruction method and system fusing physical priori, which segments data according to fracturing time points, uses a reservoir engineering model to fit a physical trend baseline which runs through the whole process and retains a step breakpoint, and uses the baseline as a constraint to build a robust decomposition objective function; the production data is decomposed into a trend component and a seasonal component through iterative optimization, and the reconstructed production data is obtained based on linear superposition of the trend component and the seasonal component, forming a closed-loop feedback. The method effectively solves the traditional defects of over-smoothing of yield mutation signals, trend fitting deviating from physical laws and lack of adaptive optimization of parameters, realizes accurate retention of fracturing step characteristics under the premise of strictly following the reservoir percolation mechanism, guarantees the physical interpretability of the trend component and the global robustness of the reconstruction results, and significantly improves the accuracy and reliability of repeated fracturing well production data cleaning and reconstruction.

Description

Technical Field

[0001] This invention belongs to the field of oil and gas field development engineering and data processing technology, and in particular relates to a method and system for reconstructing repeated fracturing production data by integrating physical priors. Background Technology

[0002] As unconventional oil and gas reservoir development enters its mid-to-late stages, problems such as the natural decline in single-well productivity and uneven reservoir utilization are becoming increasingly prominent. Repeated fracturing, as a core production enhancement measure to restore single-well productivity and expand reservoir stimulation volume, has been applied on a large scale in tight oil and gas and shale oil and gas development scenarios. Accurate and reliable single-well production time-series data is the core data foundation for quantitatively evaluating the production enhancement effect of repeated fracturing, predicting the remaining recoverable reserves of a single well, and formulating subsequent development adjustment plans. However, during field production, the raw production data collected in the field is generally characterized by strong non-stationarity, high noise content, and numerous abnormal jumps due to the combined effects of multiple factors such as measurement noise from acquisition instruments, phased well shut-in operations, and fluctuations in environmental conditions. This makes it unsuitable for direct use in production analysis and engineering calculations, necessitating data cleaning and trend reconstruction.

[0003] Currently, denoising and reconstruction techniques for oil and gas well production data can be mainly divided into two categories: one is processing methods based on moving averages or frequency domain filtering, such as moving average methods and Gaussian filtering methods. These methods are essentially low-pass filtering techniques that suppress high-frequency fluctuation components in the data through smoothing operations. The other category is purely data-driven trend decomposition methods, such as seasonal trend decomposition algorithms based on local weighted regression and standard robust seasonal trend decomposition algorithms, which extract trend terms from the data through local weighted regression fitting. However, both of the above-mentioned technologies have significant technical defects when applied to the production data processing scenario of repeatedly fractured wells: First, there is the problem of over-smoothing of production increase signals. Repeated fracturing operations will cause a step increase in the production of a single well that conforms to the physical laws of reservoir seepage. Traditional filtering methods cannot distinguish between physically meaningful production increase changes and abnormal noise. They often misjudge the peak signal of fracturing production increase as high-frequency noise and smooth or remove it, ultimately causing the fracturing effect evaluation results to be distorted. Second, the lack of physical mechanism constraints can easily lead to trend drift problems. Pure data-driven methods rely entirely on the shape of the data itself to fit the trend. In the data missing interval or the interval of large production fluctuations, the production trend obtained by decomposition often shows reverse growth or unreasonable fluctuations that violate the basic laws of reservoir seepage mechanics such as Darcy's law. The extracted trend results lack physical interpretability. Third, there is a lack of closed-loop feedback optimization mechanism. The existing data processing flow is mostly a unidirectional linear decomposition process. It is impossible to iteratively correct the model parameters based on the statistical characteristics of the residuals after decomposition, making it difficult to ensure the global optimality of the trend extraction results.

[0004] It is evident that existing production data denoising and reconstruction techniques cannot simultaneously preserve the signal of sudden changes in fracturing production, constrain reservoir physical mechanisms, and adaptively optimize model parameters, thus failing to meet the application requirements of high-precision and highly interpretable cleaning and reconstruction of production data from repeatedly fractured wells. Summary of the Invention

[0005] This invention provides a method and system for reconstructing repeated fracturing production data by integrating physical priors. This method can effectively solve the problem that existing production data denoising and reconstruction techniques are unable to simultaneously take into account the preservation of fracturing production increase abrupt signals, reservoir physical mechanism constraints, and adaptive optimization of model parameters. It can meet the application requirements of high-precision and highly interpretable cleaning and reconstruction of production data from repeated fracturing wells.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A method for reconstructing repeated fracturing production data by incorporating physical priors includes: Obtain the original production time series dataset of the target oil well; using the identified effective time point of repeated fracturing operations as the dividing point, divide the original production time series dataset into two independent physical data segments: the initial fracturing attenuation segment and the repeated fracturing production enhancement segment; Based on the pre-constructed reservoir engineering decline model, the parameters of the initial fracturing attenuation stage and the repeated fracturing production enhancement stage are fitted respectively, and a physical trend baseline is generated that runs through the entire life cycle and retains the step discontinuity at the effective time of repeated fracturing operation. Using the physical trend baseline as a constraint, the pre-constructed robust time series decomposition overall objective function is iteratively optimized and solved to obtain the optimized trend component and seasonal component. The optimized trend component is linearly superimposed with the seasonal component to obtain and output the reconstructed production data.

[0007] Furthermore, the method for identifying the effective time point of the repeated fracturing operation is as follows: By combining on-site construction log data or using cumulative sum control chart algorithms to detect abrupt changes in the production slope, the effective time point of repeated fracturing operations can be automatically identified.

[0008] Furthermore, the reservoir engineering decline model adopts the Arps decline model, specifically expressed as follows:

[0009] In the formula, q(t) for t Theoretical output at any given time; qi This is the initial output; Di This is the initial decrease rate; b A decreasing exponent; t For time steps; The parameters of the Arps hyperbolic decline model were fitted using the least squares method for the initial fracturing attenuation stage and the repeated fracturing production enhancement stage, respectively. The initial production obtained by fitting the repeated fracturing production enhancement stage was higher than the final production of the initial fracturing attenuation stage, forming a physical step at the effective time of the repeated fracturing operation. Based on the parameters of the Arps decline model obtained by fitting two physical data segments, a physical trend baseline is generated that spans the entire life cycle. This physical trend baseline is used to characterize the sudden changes in production under the physical laws of reservoir seepage.

[0010] Furthermore, the overall objective function of the robust time series decomposition adopts an improved Robust STL optimization, which includes a reconstruction error L1 norm term, a trend term second-order difference constraint term, and a physical deviation penalty term. Among them, the reconstruction error L1 norm term is used to ensure the robustness of the decomposition process, the trend term second-order difference constraint term is used to ensure the smoothness of the trend components, and the physical deviation penalty term uses the physical trend baseline as a reference to constrain the trend components to be close to the physical trend baseline. By iteratively optimizing the overall objective function of robust time series decomposition, the original production time series dataset is decomposed into three orthogonal components: trend component, seasonal component, and residual component. The trend component is constrained by a physical deviation penalty term, fluctuates around the physical trend baseline, and retains the production increase step at the effective time point of repeated fracturing operations. The seasonal component is used to extract periodic fluctuations in the data. The residual component includes measurement noise and non-physical random disturbances.

[0011] Furthermore, the overall objective function of the robust temporal decomposition The expression is as follows:

[0012] In the formula, Let L1 norm be the reconstruction error; For the second-order difference constraint of the trend term; This is a penalty item for physical deviation; For trend components; Seasonal component; For physical constraint weights; This is the original time-series production dataset; As a baseline for physical trends; Let be the trend value at time t; This represents the total timing length. For time steps; These are the weighting coefficients.

[0013] Furthermore, after iteratively optimizing and solving the pre-constructed robust time series decomposition overall objective function, the method further includes: performing statistical tests on the residual components obtained from the decomposition; if the residual components do not satisfy the white noise statistical characteristics, then after reversing the parameters of the reservoir engineering decline model and the physical constraint weights, returning to the parameter fitting process and iterative optimization and solution process until the residual components satisfy the white noise statistical characteristics; if the residual components satisfy the white noise statistical characteristics, then outputting the optimized trend components and seasonal components.

[0014] Furthermore, the statistical test of the residual components obtained from the decomposition is performed in the following steps: For the extracted residual components, white noise test and stationarity test are performed simultaneously; If the residual component passes both tests and is determined to be pure random white noise, then the effective information has been completely extracted. If the residual component fails any test or contains unextracted trend or autocorrelation information, the parameters and / or physical constraint weights of the reservoir engineering decline model are corrected in reverse based on the autocorrelation coefficient and hysteresis characteristics of the residual. After the correction is completed, the parameter fitting process and iterative optimization solution process are returned to be executed until the residual component meets the statistical characteristics of white noise. Specifically, the residual components generated in the final iteration that satisfy the statistical characteristics of white noise are judged as invalid noise and discarded.

[0015] A system for reconstructing repeated fracturing production data by incorporating physical priors includes: The data acquisition module is used to acquire the original production time series dataset of the target oil well; using the identified effective time point of repeated fracturing operations as the dividing point, the original production time series dataset is divided into two independent physical data segments: the initial fracturing attenuation segment and the repeated fracturing production enhancement segment. The parameter fitting module is used to fit parameters to the initial fracturing attenuation stage and the repeated fracturing production enhancement stage based on a pre-built reservoir engineering decline model, and to generate a physical trend baseline that runs through the entire life cycle and retains the step breakpoint at the effective time of repeated fracturing operations. The iterative solution module is used to iteratively optimize the pre-constructed robust time series decomposition overall objective function with the physical trend baseline as the constraint benchmark, and decompose it to obtain the optimized trend component and seasonal component. The overlay module is used to linearly overlay the optimized trend component with the seasonal component to obtain and output reconstructed production data.

[0016] A device for reconstructing repeatable fracturing production data by incorporating physical priors includes: Memory, used to store computer programs; A processor is used to implement the above-described method for reconstructing repetitive fracturing production data by incorporating physical priors when executing the computer program.

[0017] A computer-readable storage medium storing a computer program, which, when executed by a processor, is used to implement the above-described method for reconstructing repeated fracturing production data by incorporating physical priors.

[0018] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a method for reconstructing production data from repeated fracturing operations by incorporating physical priors. First, the data is segmented based on fracturing time points. A physical trend baseline, spanning the entire process and preserving step breakpoints, is fitted using a reservoir engineering model. This baseline serves as a constraint to construct a robust decomposition objective function. Through iterative optimization, the production data is decomposed into trend and seasonal components. The reconstructed production data is obtained by linearly superimposing these components, forming a closed-loop feedback loop. This method effectively addresses the shortcomings of traditional methods, such as over-smoothing of abrupt production increases, deviation of trend fitting from physical laws, and lack of adaptive parameter optimization. It accurately preserves fracturing step characteristics while strictly adhering to reservoir seepage mechanisms, ensuring the physical interpretability of the trend components and the global robustness of the reconstruction results. This significantly improves the accuracy and reliability of cleaning and reconstructing production data from repeatedly fracturing wells. Attached Figure Description

[0019] Figure 1 A flowchart illustrating the implementation of a method for reconstructing repeated fracturing production data that integrates physical priors, provided in an embodiment of the present invention; Figure 2 This is a comparison chart of the trend extraction effects of different denoising methods provided in the embodiments of the present invention; Figure 3 The convergence curve of the adaptive correction process of model parameters provided in the embodiments of the present invention; Figure 4 A comparison chart of the final reconstructed output and the original data provided for an embodiment of the present invention; Figure 5 This is a core flowchart of a method for reconstructing repeated fracturing production data that integrates physical priors, provided in an embodiment of the present invention. Figure 6 This is a schematic diagram of a reconstructed system for repeated fracturing production data that integrates physical priors, provided as an embodiment of the present invention. Detailed Implementation

[0020] To further understand the content of this invention, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments are merely illustrative and not limiting of the invention.

[0021] The technical terms involved in this invention are explained as follows: Arps decline model: a classic empirical model proposed by petroleum engineer J.J. Arps in 1945, used to describe the production change law of oil and gas fields or single wells in the boundary control flow stage (i.e. the natural decline stage of production).

[0022] STL stands for Seasonal and Trend decomposition using Loess, which is a method of decomposing seasonal and trend trends using locally weighted regression.

[0023] Robust STL: Robust Seasonal and Trend decomposition using Loess, which is a method for decomposing seasonal and trend trends using locally weighted regression.

[0024] Ljung-Box white noise test: This is a statistical test used to determine whether the autocorrelation coefficient of a time series data significantly deviates from zero overall within a certain lag order.

[0025] The Phillips-Perron (PP) stationarity test is a statistical test used to determine whether time series data has a "unit root".

[0026] As described in the background section, traditional methods suffer from the following significant drawbacks when processing data from repeatedly fractured wells: First, the problem of "over-smoothing" of abrupt production increases: Repeated fracturing can lead to a "step-like" increase in production that conforms to physical laws. Traditional filters cannot distinguish between physical abrupt changes and abnormal noise, often misjudging the peak production increase from fracturing as high-frequency noise and removing or smoothing it, resulting in distorted evaluation of fracturing effectiveness. Second, the lack of physical mechanism constraints leads to "trend drift": Purely data-driven methods rely entirely on data patterns. In areas with missing data or large fluctuations, the decomposed trends often exhibit reverse growth or unreasonable fluctuations that violate reservoir flow mechanics (such as Darcy's law), lacking interpretability. Third, the lack of a closed-loop feedback mechanism: Existing processes are typically unidirectional linear processes where "decomposition ends," unable to correct model parameters based on the statistical characteristics of the residuals, making it difficult to guarantee the optimality of the extracted results.

[0027] To address the aforementioned issues, this embodiment provides a method for reconstructing repeated fracturing production data by integrating physical priors. This method aims to solve problems such as overly smoothed production surge signals, lack of physical constraints, and insufficient closed-loop feedback. By introducing an Arps segmented decreasing analysis model as a hard constraint and constructing an adaptive closed loop based on residual statistical characteristics, this method achieves high-fidelity production signal reconstruction that conforms to physical laws. This method realizes a deep integration of physical mechanisms and data mining, accurately restoring the instantaneous production surge characteristics caused by repeated fracturing while eliminating high-frequency noise, ensuring the dual fidelity of the reconstructed data in terms of statistics and physics.

[0028] For example, such as Figure 5 As shown, this embodiment provides a method for reconstructing repeated fracturing production data by incorporating physical priors, including: Obtain the original production time series dataset of the target oil well; using the identified effective time point of repeated fracturing operations as the dividing point, divide the original production time series dataset into two independent physical data segments: the initial fracturing attenuation segment and the repeated fracturing production enhancement segment; Based on the pre-constructed reservoir engineering decline model, the parameters of the initial fracturing attenuation stage and the repeated fracturing production enhancement stage are fitted respectively, and a physical trend baseline is generated that runs through the entire life cycle and retains the step discontinuity at the effective time of repeated fracturing operation. Using the physical trend baseline as a constraint, the pre-constructed robust time series decomposition overall objective function is iteratively optimized and solved to obtain the optimized trend component and seasonal component. The optimized trend component is linearly superimposed with the seasonal component to obtain and output the reconstructed production data.

[0029] The reconstruction method provided in this embodiment will be further explained below: For example, such as Figure 1 As shown in the figure, this embodiment provides a method for reconstructing repeated fracturing production data by incorporating physical priors, including: Step 1: Identification of operating events and time-series segmentation; Step 2: Constructing segmented physical prior benchmarks; Step 3: Robust decomposition under physical constraints; Step 4: Residual diagnosis and adaptive parameter correction; Step 5: High-fidelity signal reconstruction.

[0030] Furthermore, step 1 specifically includes the following sub-steps: Step 101: Raw Data Acquisition. Obtain the raw production time-series data set of the target oil well. , is represented as: (1) In the formula: Indicates at time Actual production data; This represents the total length of the data.

[0031] Step 102: Determining the Fracturing Time Point. By combining on-site construction log data or using cumulative sum control chart algorithms to detect abrupt changes in the production slope, the effective time point of repeated fracturing operations is automatically identified. .

[0032] Step 103: Physical Stage Division. (Based on...) Using this as the dividing point, the original production time series dataset... The data is divided into two independent physical stage segments: the initial fracturing attenuation segment. and repeated fracturing production enhancement section : (2) Furthermore, step 2 specifically includes the following sub-steps: Step 201: Physical Model Definition. The Arps decline model (Arps hyperbolic decline model) from the field of reservoir engineering is introduced as the fundamental physical prior. Its mathematical expression is: (3) In the formula: for Theoretical output at any given time; This is the initial output; This is the initial decrease rate; It is a decreasing index (reflecting reservoir seepage characteristics).

[0033] Step 202: Piecewise parameter fitting. This applies to the segmentation in step 103. and Two data sets were used to fit the parameters of the aforementioned Arps model using the least squares method. In particular, due to the increased production effect of repeated fracturing, Initial output obtained from segment fitting Will be significantly higher than The output at the end of the period, thus in A physical step is formed at that point.

[0034] Step 203: Generate the full-cycle baseline. Based on the fitted parameters, generate a physical trend baseline that spans the entire lifecycle. The baseline is at The step breakpoint is preserved, which represents the sudden change in output under the physical law.

[0035] Furthermore, step 3 specifically includes the following sub-steps: Step 301: Objective Function Construction. Construct an improved RobustSTL optimization problem.

[0036] In the objective function of traditional trend decomposition, a physical deviation penalty term is introduced, and its overall objective function becomes... Defined as: (4) In the formula: The L1 norm of the reconstruction error (to ensure robustness); The second-order difference constraint for the trend term (to ensure smoothness); This is a penalty item for physical deviation; Let be the trend component to be solved; Seasonal component; These are the physical constraint weights.

[0037] Step 302: Optimization and Component Extraction. By iteratively solving the above optimization problem, the original data... Decomposed into three orthogonal components: Trend Components Constrained by physical penalty terms, closely aligned with the physical trend baseline. Fluctuations, preserved The production increase step at the location; Seasonal components Extract periodic fluctuations from the data (such as well shut-in cycles and the effects of seasonal temperature differences). residual components : It mainly includes measurement noise and non-physical random disturbances; This is the original production time series data.

[0038] Furthermore, step 4 specifically includes the following sub-steps: Step 401: Residual Statistical Test. This involves testing the residual components extracted in step 302. Perform the Ljung-Box white noise test and the PP (Phillips-Perron) stationarity test.

[0039] Step 402: Closed-loop feedback judgment. Set the significance level (e.g., significance level). If the test passes (residual) If the result is determined to be pure random white noise, it means that all the valid information has been extracted, and proceed to step 5; if the test fails (the residual still contains unextracted trends or autocorrelation information), proceed to step 403.

[0040] Step 403: Parameter Backward Correction. Based on the autocorrelation coefficient and lag characteristics of the residuals, the Arps model parameters in Step 2 are backward corrected (e.g., fine-tuning the decreasing exponent). (or the physical constraint weights in step 3) After correction, return to steps 2 through 4 until the residuals satisfy the statistical properties of white noise.

[0041] Furthermore, step 5 specifically includes the following sub-steps: Step 501: Noise Removal. Discard the residual components generated in the final iteration. This is considered invalid noise.

[0042] Step 502: Synthesize the output. Combine the optimized physical trend components. Seasonal component By performing linear superposition, the reconstructed output is obtained. : (5) Final output reconstruction yield The curves, while smoothing out high-frequency noise, fully preserve the peak production increase from repeated fracturing, and the overall trend conforms to the reservoir seepage mechanics characteristics.

[0043] For example, in order to verify the effectiveness of the reconstruction method provided in the above embodiments, this embodiment constructs a dedicated data processing environment.

[0044] Hardware Platform: This embodiment uses a high-performance workstation equipped with an Intel Xeon Gold 6248R CPU (3.0GHz, 24 cores) and 64GB DDR4 memory for computation. However, this method is not limited to this; any computing unit with floating-point capabilities, including but not limited to ordinary personal computers, cloud computing instances (such as AWS EC2), or embedded industrial control computers, can be used as the execution subject of this method.

[0045] Software environment: The algorithm is written in Python 3.8, optimized using SciPy 1.7.0, and time-series processing is performed using Pandas. However, in practical applications, the logic of this method can be ported to any programming language such as C++, MATLAB, or Java, and does not depend on any specific third-party library version.

[0046] Based on the hardware platform and software environment established above, this reconstruction method is implemented in practice. The specific implementation process is as follows: This embodiment selects the daily gas production data of a certain shale gas horizontal well "Well-X" throughout its entire life cycle as the input object, and the data length is... The well was on the [day]. Repeated fracturing operations were carried out that day. The specific implementation steps are as follows: Phase 1: Operating Condition Identification and Physical Segmentation Perform step 101 to export the raw production sequence from the field SCADA system. .

[0047] Step 102: This embodiment uses the CUSUM algorithm to detect abrupt changes in slope. A detection threshold is set. ( (This is the standard deviation of the preceding window). However, it should be noted that this threshold can be adjusted based on the data noise level. The algorithm can be flexibly adjusted without affecting the positioning accuracy of the abrupt change point. In this embodiment, the algorithm automatically locks the effective time point of repeated fracturing. sky.

[0048] Execution step 103: The system uses The data is divided into the initial fracturing section as a boundary. (Days 1-649) and repeated fracturing sections (Days 650-1200).

[0049] Phase Two: Physical Benchmark Construction Execution steps 201 and 202: Introduction hyperbolic decreasing model Fit the data to the two data segments separately. The initial decline rate was obtained by fitting the segment. Decreasing exponent ;exist The initial yield was obtained by fitting the segment. It was significantly higher than the output at the end of the previous period.

[0050] Step 203: Join the two curve segments to generate a physical trend baseline. The baseline is at The area exhibits a distinct "step-like" physical characteristic.

[0051] Phase 3: Decomposition of Physical Constraints Step 301: Construct the objective function with a physics penalty term:

[0052] In this embodiment, physical constraint weights The preferred setting is 0.5. However, in practical applications, this parameter can be adjusted between 0.1 and 10.0 depending on the data signal-to-noise ratio: when the noise is extremely high, it can be appropriately increased. To enhance the guiding role of the physical model; when the data quality is good, the adjustment can be reduced. To retain more data details.

[0053] Step 302: Solve the above convex optimization problem using ADMM (Alternating Direction Multiplier Method), setting the number of iterations to 100, and extract the preliminary trend components. Seasonal components and residual components .

[0054] Phase 4: Closed-loop feedback correction Execute step 401: Process the residuals Perform Ljung-Box white noise test and set the lag order. .

[0055] Step 402: The calculated p-value is 0.01, which is less than the significance level. This indicates that the residuals still contain unextracted autorelated information (test failed).

[0056] Step 403 (Adaptive Correction): The system automatically triggers a feedback mechanism to fine-tune the decreasing exponent in the Arps model based on the residual autocorrelation characteristics. Then, change it from 1.2 to 1.15 and re-execute steps 2 to 3.

[0057] After three rounds of iteration, the p-value was retested and increased to 0.23. The residual is determined to be white noise, and the iteration stops.

[0058] Phase 5: Signal Reconstruction Perform steps 501 and 502: Discard the final white noise residual and apply the optimized physical trend. Seasonal component Superimpose and output the final reconstructed curve. .

[0059] Furthermore, the implementation effect analysis of this embodiment is as follows: To visually demonstrate the technical effectiveness of this method, the following three sets of comparative experiments were conducted in this embodiment: like Figure 2 As shown, with time (days) as the horizontal axis and daily gas production as the vertical axis, the trend extraction effects of different noise reduction methods are intuitively compared. Figure 2 In the diagram, gray scatter dots represent the original noisy data, blue dashed lines represent trend terms obtained by decomposition using the traditional STL algorithm, and red solid lines represent trend terms extracted using the reconstruction method (incorporating physical priors) provided in this embodiment. Through the analysis of A magnified view of the localized moment during repeated fracturing reveals that the traditional STL method (blue dashed line) exhibits a smooth, sloping transition, showing a clear "oversmoothing" phenomenon and losing the abrupt change in yield characteristics. In contrast, the present invention (red solid line) accurately presents a "vertical step" consistent with physical reality at this moment, and in the subsequent decline phase, it conforms to the Arps decline law more closely than the blue dashed line, unaffected by individual outliers. This fully demonstrates that the present invention has significantly better physical fidelity than traditional pure data-driven algorithms when handling strong non-stationary abrupt change signals.

[0060] like Figure 3 As shown, a dual-axis chart illustrates the model convergence and index changes during the "closed-loop feedback" process. Figure 3 In the middle, the left-axis bar chart shows the Ljung-Box test statistic for each iteration. The p-value is displayed on the right-axis line chart, with the horizontal axis representing the iteration number. The horizontal dashed line in the chart represents the significance threshold. As observed, in the first iteration, the Q-value was high and the p-value was below the threshold, indicating that the model was not perfectly fitted. With the feedback mechanism correcting the physical parameters, the Q-value decreased round by round, and the p-value finally broke through the threshold in the third iteration. This process confirms that the adaptive correction mechanism based on residual statistical characteristics proposed in this invention can automatically guide the model to converge to the statistically optimal solution, effectively avoiding the subjectivity of manual parameter tuning. Here, the p-value is the probability of observing the current sample data assuming the null hypothesis is true.

[0061] like Figure 4 As shown, the comparison results between the reconstructed output curve and the original data are presented. Figure 4 In the middle, the background layer is the original noisy data with a light gray background. It exhibits significant high-frequency oscillations; the foreground layer is reconstructed data in dark green. , Figure 4 The Chinese label is marked The initial decreasing trend of the stage and The phase exhibits a "decreasing trend after increased production." The results show that the reconstructed curve... The data is extremely smooth, effectively filtering out measurement noise such as glitch signals, while perfectly matching the central trend of the original data and preserving long-term fluctuation details such as seasonal effects, without losing effective information due to noise reduction. This proves that the present invention can reproduce high-fidelity production data that conforms to reservoir seepage mechanics under extremely low signal-to-noise ratio conditions, providing a reliable data foundation for subsequent production capacity evaluation.

[0062] For example, such as Figure 6As shown, this embodiment also provides a system for reconstructing repeated fracturing production data by integrating physical priors, including: a data acquisition module for acquiring the original production time series dataset of the target oil well; using the identified effective time point of repeated fracturing operation as the dividing point, dividing the original production time series dataset into two independent physical data segments: the initial fracturing attenuation segment and the repeated fracturing production enhancement segment; The parameter fitting module is used to fit parameters to the initial fracturing attenuation stage and the repeated fracturing production enhancement stage based on a pre-built reservoir engineering decline model, and to generate a physical trend baseline that runs through the entire life cycle and retains the step breakpoint at the effective time of repeated fracturing operations. The iterative solution module is used to iteratively optimize the pre-constructed robust time series decomposition overall objective function with the physical trend baseline as the constraint benchmark, and decompose it to obtain the optimized trend component and seasonal component. The overlay module is used to linearly overlay the optimized trend component with the seasonal component to obtain and output reconstructed production data.

[0063] The present invention also provides a device for reconstructing repeated fracturing production data by incorporating physical priors, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the method for reconstructing repeated fracturing production data by incorporating physical priors.

[0064] The present invention also provides a computer program product, including a computer program / instruction that, when executed by a processor, implements the steps of the repetitive fracturing production data reconstruction method that incorporates physical priors.

[0065] When the processor executes the computer program, it implements the above-mentioned steps of reconstructing repeated fracturing production data by fracturing with physical prior knowledge, for example: obtaining the original production time series dataset of the target oil well; using the identified effective time point of repeated fracturing operation as the dividing point, dividing the original production time series dataset into two independent physical data segments: the initial fracturing attenuation segment and the repeated fracturing production enhancement segment. Based on the pre-constructed reservoir engineering decline model, the parameters of the initial fracturing attenuation stage and the repeated fracturing production enhancement stage are fitted respectively, and a physical trend baseline is generated that runs through the entire life cycle and retains the step discontinuity at the effective time of repeated fracturing operation. Using the physical trend baseline as a constraint, the pre-constructed robust time series decomposition overall objective function is iteratively optimized and solved to obtain the optimized trend component and seasonal component. The optimized trend component is linearly superimposed with the seasonal component to obtain and output the reconstructed production data.

[0066] Exemplarily, the computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing preset functions, the instruction segments describing the execution process of the computer program in the fusion physical prior repetitive fracturing production data reconstruction device. For example, the computer program can be divided into a data acquisition module, a parameter fitting module, an iterative solution module, and an overlay module; the specific functions are as follows: The data acquisition module is used to acquire the original production time series dataset of the target oil well; using the identified effective time point of repeated fracturing operations as the dividing point, the original production time series dataset is divided into two independent physical data segments: the initial fracturing attenuation segment and the repeated fracturing production enhancement segment; The parameter fitting module is used to perform parameter fitting on the initial fracturing attenuation segment and the repeated fracturing production enhancement segment based on a pre-constructed reservoir engineering decline model, respectively, to generate a physical trend baseline that runs through the entire life cycle and retains a step breakpoint at the effective time point of repeated fracturing operations; The iterative solution module is used to iteratively optimize and solve the pre-constructed robust time series decomposition overall objective function with the physical trend baseline as the constraint benchmark, and decompose it to obtain the optimized trend component and seasonal component; The overlay module is used to linearly overlay the optimized trend component and the seasonal component to obtain and output the reconstructed production data.

[0067] The device for reconstructing repeated fracturing production data using fusion physical priors can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. This device may include, but is not limited to, processors and memory. Those skilled in the art will understand that the above are examples of devices for reconstructing repeated fracturing production data using fusion physical priors and do not constitute a limitation on such devices. The device may include more components than described above, or combine certain components, or use different components. For example, the device may also include input / output devices, network access devices, buses, etc.

[0068] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor, or any conventional processor. The processor is the control center for the fusion physical prior repetitive fracturing production data reconstruction, connecting various parts of the fusion physical prior repetitive fracturing production data reconstruction equipment via various interfaces and lines.

[0069] The memory can be used to store the computer program and / or modules. The processor realizes various functions of the fusion physical prior repetitive fracturing production data reconstruction device by running or executing the computer program and / or modules stored in the memory and calling the data stored in the memory.

[0070] The memory may primarily include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a function (such as sound playback, image playback, etc.). The data storage area may store data created based on the use of the mobile phone (such as audio data, phonebook, etc.). Furthermore, the memory may include high-speed random access memory and non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart media cards (SMC), secure digital cards (SD cards), flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0071] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the described method for reconstructing repetitive fracturing production data by incorporating physical priors.

[0072] If the modules / units of the integrated physical prior repetitive fracturing production data reconstruction system are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0073] Based on this understanding, the present invention can implement all or part of the processes in the above-mentioned method for reconstructing repeated fracturing production data by incorporating physical priors, or it can be accomplished by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the above-mentioned method for reconstructing repeated fracturing production data by incorporating physical priors. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or a preset intermediate form, etc.

[0074] The computer-readable storage medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0075] It should be noted that the content contained in the computer-readable storage medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable storage medium does not include electrical carrier signals and telecommunication signals.

[0076] Compared with existing reconstruction measures, this invention provides a method and system for reconstructing repeated fracturing production data by incorporating physical priors, which has the following advantages: First, this invention solves the problem of oversmoothing "production increase mutation signals" in traditional algorithms: By introducing a "segmented physical hard constraint" mechanism, this invention clarifies that production mutations are an endogenous feature of the physical model. Through segmented modeling with the effective time point of repeated fracturing operations as the boundary, the algorithm can accurately distinguish between "legitimate mutations brought about by construction" and "illegal noise brought about by measurement," thereby achieving strong noise reduction while retaining 100% of the peak characteristics of fracturing production increase, providing a reliable basis for evaluating fracturing effects.

[0077] Secondly, this invention improves the interpretability and physical consistency of the denoising results. By constructing a denoising objective function that integrates a "mechanistic model and a statistical model," the Arps production decline formula from reservoir engineering is embedded into the RobustSTL loss function. This forces the denoised trend term to approximate the physical true value, correcting the "trend drift" problem commonly found in pure data-driven methods in areas with missing data, and ensuring that the processed data strictly conforms to the principles of reservoir seepage mechanics (such as Darcy's law).

[0078] Third, adaptive optimization of model parameters is achieved: This invention establishes a "model parameter adaptive correction closed loop" based on residual statistical characteristics. By using the Ljung-Box test statistic as a feedback signal, it overcomes the limitations of the traditional "one-way execution" technique. This mechanism realizes the transformation from "passive denoising" to "active optimization and reconstruction," ensuring that the final separated residual sequence satisfies the strict white noise statistical assumption and extracts the effective information from the data to the greatest extent.

[0079] The above embodiments are merely one of the implementation methods for achieving the technical solution of the present invention. The scope of protection claimed by the present invention is not limited to this embodiment, but also includes any variations, substitutions and other implementation methods that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention.

[0080] 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 protection scope of the present invention.

Claims

1. A method for reconstructing repeated fracturing production data by incorporating physical priors, characterized in that, include: Obtain the original production time series dataset of the target oil well; using the identified effective time point of repeated fracturing operations as the dividing point, divide the original production time series dataset into two independent physical data segments: the initial fracturing attenuation segment and the repeated fracturing production enhancement segment; Based on the pre-constructed reservoir engineering decline model, the parameters of the initial fracturing attenuation stage and the repeated fracturing production enhancement stage are fitted respectively, and a physical trend baseline is generated that runs through the entire life cycle and retains the step discontinuity at the effective time of repeated fracturing operation. Using the physical trend baseline as a constraint, the pre-constructed robust time series decomposition overall objective function is iteratively optimized and solved to obtain the optimized trend component and seasonal component. The optimized trend component is linearly superimposed with the seasonal component to obtain and output the reconstructed production data.

2. The method for reconstructing repeated fracturing production data by incorporating physical priors as described in claim 1, characterized in that, The method for identifying the effective time point of the repeated fracturing operation is as follows: By combining on-site construction log data or using cumulative sum control chart algorithms to detect abrupt changes in the production slope, the effective time point of repeated fracturing operations can be automatically identified.

3. The method for reconstructing repeated fracturing production data by incorporating physical priors as described in claim 1, characterized in that, The reservoir engineering decline model adopts the Arps decline model, which is specifically expressed as follows: In the formula, q(t) for t Theoretical output at any given time; qi This is the initial output; Di This is the initial decrease rate; b A decreasing exponent; t For time steps; The parameters of the Arps hyperbolic decline model were fitted using the least squares method for the initial fracturing attenuation stage and the repeated fracturing production enhancement stage, respectively. The initial production obtained by fitting the repeated fracturing production enhancement stage was higher than the final production of the initial fracturing attenuation stage, forming a physical step at the effective time of the repeated fracturing operation. Based on the parameters of the Arps decline model obtained by fitting two physical data segments, a physical trend baseline is generated that spans the entire life cycle. This physical trend baseline is used to characterize the sudden changes in production under the physical laws of reservoir seepage.

4. The method for reconstructing repeated fracturing production data by incorporating physical priors as described in claim 1, characterized in that, The robust time series decomposition overall objective function adopts an improved Robust STL optimization overall objective function, including a reconstruction error L1 norm term, a trend term second-order difference constraint term, and a physical deviation penalty term. Among them, the reconstruction error L1 norm term is used to ensure the robustness of the decomposition process, the trend term second-order difference constraint term is used to ensure the smoothness of the trend components, and the physical deviation penalty term uses the physical trend baseline as a reference to constrain the trend components to be close to the physical trend baseline. By iteratively optimizing the overall objective function of robust time series decomposition, the original production time series dataset is decomposed into three orthogonal components: trend component, seasonal component, and residual component. The trend component is constrained by a physical deviation penalty term, fluctuates around the physical trend baseline, and retains the production increase step at the effective time point of repeated fracturing operations. The seasonal component is used to extract periodic fluctuations in the data. The residual component includes measurement noise and non-physical random disturbances.

5. The method for reconstructing repeated fracturing production data by incorporating physical priors as described in claim 4, characterized in that, The robust temporal decomposition overall objective function The expression is as follows: In the formula, Let L1 norm be the reconstruction error; For the second-order difference constraint of the trend term; This is a penalty item for physical deviation; For trend components; Seasonal component; For physical constraint weights; This is the original time-series production dataset; As a baseline for physical trends; Let be the trend value at time t; This represents the total timing length. For time steps; These are the weighting coefficients.

6. The method for reconstructing repeated fracturing production data by incorporating physical priors as described in claim 1, characterized in that, After iteratively optimizing and solving the pre-constructed robust time series decomposition overall objective function, the method further includes: performing statistical tests on the residual components obtained from the decomposition; if the residual components do not satisfy the white noise statistical characteristics, then after reversing the parameters of the reservoir engineering decline model and the physical constraint weights, returning to the parameter fitting process and iterative optimization and solution process until the residual components satisfy the white noise statistical characteristics; if the residual components satisfy the white noise statistical characteristics, then outputting the optimized trend components and seasonal components.

7. The method for reconstructing repeated fracturing production data by incorporating physical priors as described in claim 6, characterized in that, The statistical test of the residual components obtained from the decomposition is performed in the following steps: For the extracted residual components, white noise test and stationarity test are performed simultaneously; If the residual component passes both tests and is determined to be pure random white noise, then the effective information has been completely extracted. If the residual component fails any test or contains unextracted trend or autocorrelation information, the parameters and / or physical constraint weights of the reservoir engineering decline model are corrected in reverse based on the autocorrelation coefficient and hysteresis characteristics of the residual. After the correction is completed, the parameter fitting process and iterative optimization solution process are returned to be executed until the residual component meets the statistical characteristics of white noise. Specifically, the residual components generated in the final iteration that satisfy the statistical characteristics of white noise are judged as invalid noise and discarded.

8. A system for reconstructing repeated fracturing production data by integrating physical priors, characterized in that, include: The data acquisition module is used to acquire the original production time series dataset of the target oil well; using the identified effective time point of repeated fracturing operations as the dividing point, the original production time series dataset is divided into two independent physical data segments: the initial fracturing attenuation segment and the repeated fracturing production enhancement segment. The parameter fitting module is used to fit parameters to the initial fracturing attenuation stage and the repeated fracturing production enhancement stage based on a pre-built reservoir engineering decline model, and to generate a physical trend baseline that runs through the entire life cycle and retains the step breakpoint at the effective time of repeated fracturing operations. The iterative solution module is used to iteratively optimize the pre-constructed robust time series decomposition overall objective function with the physical trend baseline as the constraint benchmark, and decompose it to obtain the optimized trend component and seasonal component. The overlay module is used to linearly overlay the optimized trend component with the seasonal component to obtain and output reconstructed production data.

9. A device for reconstructing repeated fracturing production data by integrating physical priors, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the repetitive fracturing production data reconstruction method according to any one of claims 1-7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it is used to implement the reconstructing method for repeated fracturing production data that incorporates physical priors as described in any one of claims 1-7.

Images