Fusion mechanism constraints and data-driven network fracture construction curve diagnosis method

Abstract

The application discloses a kind of fusion mechanism constraints and data-driven jointed network fracturing construction curve diagnosis method, belong to hydraulic fracturing technical field.The application is based on well bottom net pressure response, mechanism constraint is carried out in combination with the behavior characteristics of fracture propagation process, data driving is carried out based on microseismic monitoring data, and a multi-source feature fusion fracture propagation pattern recognition model is constructed.Compared with traditional fracture monitoring means, the method has the advantages of strong real-time, low cost and wide adaptability, and can effectively improve the scientificity and practicality of fracture propagation pattern identification, providing reliable technical support for unconventional reservoir fracturing development.

Description

Technical Field

[0001] This invention relates to the field of hydraulic fracturing technology, and in particular to a diagnostic method for fracture network fracturing construction curves that integrates mechanistic constraints and data-driven approaches. Background Technology

[0002] Shale gas, as an important unconventional natural gas resource, relies on reservoir stimulation technologies such as volumetric fracturing for efficient development to maximize reservoir stimulation volume and production capacity. During fracturing operations, the fracture propagation pattern reflects the fracture stress state and its coupling behavior with the geological structure, and is a crucial factor influencing the fracture network configuration and complexity, directly affecting the stimulation effect and final production capacity. Therefore, accurately identifying fracture propagation behavior in the formation is of great significance for optimizing fracturing parameters and efficient shale gas development.

[0003] Commonly used fracture monitoring methods include microseismic monitoring and fiber optic monitoring. While these can acquire parameters such as fracture orientation, height, complexity, and fluid ingress, they suffer from high equipment deployment costs, dependence on adjacent well conditions, and complex and subjective data interpretation, limiting their widespread application in large-scale fracturing. In contrast, pressure response data, as core information acquired in real time during fracturing, offers advantages such as convenient acquisition, low cost, and ease of deployment. Existing research indicates that the fluctuation characteristics of net pressure during fracturing can reflect the state and extension mechanism of fracture propagation, possessing potential value for fracture behavior identification.

[0004] However, most existing pressure response-based fracture identification methods lack quantitative constraints on fracture geometry and rely heavily on empirical judgment, resulting in insufficient scientific rigor and engineering adaptability. In the prior art, the invention patent with publication number CN110056336A, entitled "An Automatic Diagnosis Method for Pressure Curves in Shale Gas Fracture Network Fracturing Construction," discloses an automatic diagnosis method based on fluid dynamics and fracture propagation theory. While it can identify fractures using pressure curves, it fails to combine the actual geometric shape of the fracture with on-site monitoring characteristics for verification, thus its accuracy and applicability remain significantly limited.

[0005] Based on this, a smart diagnostic method for fracture network fracturing construction curves that integrates mechanistic constraints and data-driven approaches is proposed. Summary of the Invention

[0006] The purpose of this invention is to provide a diagnostic method for fracture network fracturing construction curves that integrates mechanistic constraints and data-driven approaches, in order to solve the problems in the background art.

[0007] To achieve the above objectives, this invention provides a diagnostic method for fracture network fracturing construction curves that integrates mechanistic constraints and data-driven approaches, comprising the following steps:

[0008] S1. Data Acquisition and Conversion: Collect wellhead pressure data and microseismic monitoring data during the fracturing operation of the target block, and calculate the bottom hole net pressure based on the wellhead pressure data;

[0009] S2. Constructing a mechanism model: Based on the trend of net pressure change at the bottom of the well, the fracture propagation behavior of the target block is divided into multiple fracture propagation modes. The propagation index is introduced as the discrimination criterion for different fracture propagation modes, and a mechanism constraint interval is constructed.

[0010] S3. Associated field data: For each crack propagation mode, extract the corresponding microseismic characteristic parameters from the microseismic monitoring data that can characterize the crack complexity and crack height.

[0011] S4. Reconstruction of discrimination criteria: By integrating the mechanism constraint interval and microseismic characteristic parameters, the XGBoost interval regression model is used to adaptively correct and optimize the mechanism constraint interval, resulting in an adaptive crack propagation mode discrimination model.

[0012] Preferably, in step S1, the formula for calculating the net pressure at the bottom of the well is:

[0013]

[0014] In the formula, P h Indicates wellhead pressure, in Pa; ΔP represents the static pressure of the sand-carrying fluid column within a single section of the wellbore. wh The integral over the vertical depth Z from the wellhead to the wellbore, Pa; The flow friction ΔP in a single wellbore section represents the frictional resistance of the flow. wf The integral over the length L from the wellhead to the wellbore, Pa; ΔP pf The frictional resistance of the orifice is expressed in Pa; σ hmin This represents the minimum horizontal principal stress, expressed in Pa.

[0015] The preferred formula for calculating the static pressure of the sand-carrying fluid column within a single section of the wellbore is as follows:

[0016] ΔP wh =[(1-V p )ρ l +V p ρ p gZ;

[0017] In the formula, ΔP wh V represents the static pressure of the sand-carrying fluid inside the wellbore, in Pa; p Indicates the percentage of proppant; ρ l This indicates the density of the fracturing fluid, in kg / m³. 3 ;ρ p This indicates the density of the proppant, in kg / m³. 3 g represents gravitational acceleration, m / s² 2Z represents the vertical depth of the wellbore, in meters (m).

[0018] Preferably, the formula for calculating the flow friction of fracturing fluid through the perforation orifice is:

[0019]

[0020] In the formula, q represents the fracturing fluid flow rate, i.e., the displacement, and m 3 / min;n pf Indicates the number of holes in a single section; d pf Indicates the diameter of the perforation hole, in meters (m); α pf ρ represents the orifice flow coefficient, ranging from 0.8 to 0.85, and is dimensionless; l This indicates the density of the fracturing fluid, in kg / m³. 3 .

[0021] The preferred formula for calculating the flow friction of a single section of wellbore is:

[0022]

[0023] In the formula, ΔP wf δ represents the frictional resistance of the fluid flow in the wellbore, Pa; f represents the hydraulic friction coefficient, dimensionless; L represents the length of the wellbore, m; D represents the diameter of the wellbore, m; v represents the fracturing fluid velocity, m / s; δ is the drag reduction rate, dimensionless.

[0024] The formula for calculating the hydraulic friction coefficient is as follows:

[0025]

[0026] In the formula, ε is the absolute roughness of the inner wall of the pipe, in meters (m); Re represents the Reynolds number, which is dimensionless.

[0027] The formula for calculating the drag reduction ratio is:

[0028]

[0029] In the formula, C g Thickener concentration, kg / m³ 3 C s This refers to the proppant concentration, in kg / m³. 3 e is the base of the natural logarithm, approximately equal to 2.71828.

[0030] Preferably, the specific steps of S2 are as follows:

[0031] S21. Based on rock mechanics and fracture propagation theory, combined with the geological characteristics and engineering practice of the target block, analyze the quantitative relationship between bottom hole net pressure and fracture propagation behavior, and classify multiple fracture propagation modes according to the quantitative relationship.

[0032] S22. By analyzing the changing trend of the net pressure curve through numerical simulation, the physical mechanisms and key characteristic behaviors corresponding to different fracture propagation modes are identified. Based on the changing trend of the bottom hole net pressure curve, the propagation index is calculated, and the mechanism constraint interval of the propagation index corresponding to different fracture propagation modes is constructed as a constraint condition at the mechanism level.

[0033] Preferably, in S2, the crack propagation modes include propagation along natural cracks, crack height limitation, smooth crack height propagation, filtration along bedding and natural cracks, and breaking through natural cracks; the propagation mode along natural cracks and the filtration mode along bedding and natural cracks are uniformly defined as the crack network propagation stage;

[0034] For the natural crack pattern, since its expansion behavior is complex and difficult to accurately characterize with a single indicator, an exclusion method is used for identification. After the other patterns are identified, the remaining unclassified crack expansion behavior samples are classified as natural crack patterns.

[0035] Preferably, in step S2, the formula for calculating the expansion index is:

[0036]

[0037] In the formula, n is the crack propagation mode exponent, which is dimensionless; t i P represents the current time point. net i For t i The corresponding net pressure at the bottom of the well at time t, in MPa; i-1 P is the time point corresponding to the previous moment. net i-1 For t i-1 The net pressure at the bottom of the well at that moment, in MPa.

[0038] Preferably, in S3, for the fracture network expansion stage consisting of natural fracture propagation mode and fracture network expansion mode along bedding and natural fracture filtration mode, the ratio of microseismic event number to reservoir stimulation volume is used as a quantitative characteristic parameter of fracture complexity; for fracture height-restricted mode, the ratio of total fracture length to fracture height is used as a characteristic parameter to evaluate the degree of vertical propagation restriction; for fracture height-smooth propagation mode, fracture height is used as a characteristic parameter to measure the sufficiency of vertical propagation; wherein, the characteristic parameters of fracture propagation mode are all extracted from microseismic monitoring data, and have strong field applicability and quantifiable characteristics.

[0039] Preferably, in step S4, an XGBoost interval regression model is introduced, using the mechanism constraint interval as a soft label and the microseismic characteristic parameters as input, with minimizing the loss function as the training objective. The specific steps are as follows:

[0040] S41. Construct a training dataset by fusing the mechanism constraint interval and microseismic characteristic parameters. The training dataset is represented as follows:

[0041]

[0042] In the formula, D train The training dataset contains multiple sample points, each consisting of microseismic characteristic parameters and corresponding mechanistic constraint intervals; x i This represents the microseismic characteristic parameter vector of the i-th crack propagation mode. represent the lower and upper limits of the corresponding mechanism constraint intervals, respectively, and N represents the total number of training set samples;

[0043] S42. Establish an XGBoost interval regression model and define a loss function for fitting and learning. After training, statistically analyze the model output f(x) for all sample points under each crack propagation mode type. i The distribution analysis is performed, and the corrected pattern discrimination interval is output.

[0044] Preferably, in the loss function of S42, there is no penalty if the output of the XGBoost interval regression model falls within the mechanistic constraint interval, and there is a squared penalty if it exceeds the mechanistic constraint interval; the training objective is to minimize the overall loss function.

[0045] Preferably, in step S42, the loss function is expressed as:

[0046]

[0047] In the formula, f(x) i ) represents the output of the XGBoost interval regression model, and represents the predicted expansion index;

[0048] The training objective is represented as:

[0049]

[0050] In the formula, This refers to the structure and parameter set of the model; This is the interval loss function value for the i-th sample point in the training dataset.

[0051] Preferably, the XGBoost interval regression model in S4 is expressed as:

[0052]

[0053] In the formula, K is the number of regression trees, f k (x iLet be the k-th subtree, and F represent the function space formed by all regression trees. Each tree maps the input sample points to a certain leaf node and outputs the corresponding predicted value.

[0054] Therefore, this invention presents a diagnostic method for fracture network fracturing construction curves that integrates mechanistic constraints and data-driven approaches. By combining the dynamic response of bottom hole net pressure and microseismic monitoring characteristics during fracturing operations, a theoretically interpretable and data-driven fracture propagation pattern index system is constructed. Furthermore, an XGBoost-based interval regression model combined with a loss function is introduced for correction, forming an adaptive intelligent discrimination model. This effectively overcomes the limitations of traditional mechanistic models in terms of accuracy and the strong subjectivity of empirical judgments, significantly improving the accuracy and adaptability of fracture propagation behavior discrimination. Simultaneously, the model possesses strong interpretability and field applicability, providing real-time guidance for fracturing parameter optimization and fracture propagation control. This provides a scientific theoretical basis and reliable data support for efficient fracturing development of shale gas reservoirs, contributing to improved reservoir stimulation effects and the achievement of stable and increased production targets.

[0055] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0056] Figure 1 This is a schematic diagram of the overall process of an embodiment of the present invention;

[0057] Figure 2 This is a microseismic monitoring image of well FY according to an embodiment of the present invention;

[0058] Figure 3 The numerical simulation results of different fracture propagation modes and the corresponding bottom hole net pressure curves of the present invention are shown in the embodiments of the present invention; wherein, (a), (b), (c), (d), and (e) represent the numerical simulation results of fracture height restriction, filtration along bedding and natural fractures, smooth fracture height propagation, propagation along natural fractures, and breaking through natural fracture modes, respectively; and (f), (g), (h), (i), and (j) are the corresponding bottom hole net pressure curves, respectively.

[0059] Figure 4 This describes the optimization process for the number of trees during the hyperparameter optimization of the model in this embodiment of the invention.

[0060] Figure 5 This describes the optimization process for the maximum tree depth during the hyperparameter optimization of the model in this embodiment of the invention.

[0061] Figure 6 The diagram shows the diagnostic results of the third segment of the FY well according to an embodiment of the present invention; where (a) is the diagnostic results of the fracture propagation mode index; and (b) is the corresponding diagnostic results of the fracture propagation mode.

[0062] Figure 7This is a correlation analysis diagram between the diagnostic results of the fracture network complexity of the entire well section model of well FY and the microseismic monitoring results in an embodiment of the present invention.

[0063] Figure 8 This is a correlation analysis diagram between the diagnostic results of fracture height expansion in the entire well section model of well FY and the microseismic monitoring results in an embodiment of the present invention. Detailed Implementation

[0064] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0065] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0066] Example

[0067] like Figure 1 As shown, this invention provides a diagnostic method for fracture network fracturing construction curves that integrates mechanistic constraints and data-driven approaches. Taking the F block in Southwest China as the implementation area, this method fits and dynamically corrects the mechanistic constraint interval, and then uses the corrected fracture propagation mode index (propagation index) range to determine the fracture propagation mode of a continental shale horizontal well, FY. The specific details are as follows:

[0068] Well FY was designed to fracture 26 sections. Microseismic methods were used for fracture monitoring during construction. The monitoring results for each fractured section are summarized below. Figure 2 See Table 1.

[0069] Table 1. On-site fracture monitoring results for each section of Well FY

[0070]

[0071] S1. Data Acquisition and Conversion: Net pressure is converted based on the wellhead pressure data monitored from multiple wells in Block F, serving as the basis for subsequent analysis.

[0072] S2. Constructing a mechanistic model to provide prior information and initial parameter values ​​for subsequent data-driven models: This embodiment considers the influence of natural cracks, geostress fields, and fluid properties, and combines the mechanical mechanisms and characterization equations of different crack propagation behaviors to classify crack propagation modes into five categories (e.g., Figure 3 (As shown). Then, based on the changing trend of the net pressure curve, calculate the crack propagation mode index (propagation index) n.

[0073] The propagation index is used to characterize the overall trend of net pressure change during crack propagation. Its physical meaning lies in reflecting the differences in pressure response characteristics caused by typical behaviors such as crack obstruction, filtration, and smooth propagation.

[0074] In this embodiment, the specific judgment conditions corresponding to the 5 types of patterns are as follows:

[0075] Pattern 1: Propagation along natural cracks: n>0.5;

[0076] Mode 2, Seam Height Restricted: 0 <n<0.5;

[0077] Mode 3, suture height extension proceeded smoothly: -0.3 <n<0;

[0078] Mode 4, filtration along bedding and natural cracks: -0.5 <n<-0.3;

[0079] Mode 5, Breaking through natural cracks: n < -0.5.

[0080] The mesh expansion stage is defined as Mode 1 + Mode 4.

[0081] S3. To enhance the correlation between fracture propagation mode discrimination results and field monitoring data, a corresponding characteristic parameter matching scheme is proposed: calculate the ratio of microseismic event number to reservoir stimulation volume (SRV) as a quantitative indicator of fracture complexity (mode 1, mode 4); calculate the ratio of total fracture length to fracture height as a quantitative indicator of fracture height restriction; and extract the monitored fracture height data to assess the smoothness of fracture height propagation.

[0082] S4. Reconstructing the discrimination criteria: Constructing a training dataset and using an XGBoost-based interval regression model to perform corrected fitting of the mechanism constraint interval (exponential interval of crack propagation mode). The specific process is as follows:

[0083] The training dataset is constructed as follows:

[0084]

[0085] Wherein, the input feature vector x i = (Number of events / SRV, Total crack length / Height, Crack height), fitted using an XGBoost interval regression model (random forest regression model), with the initial hyperparameter Θ set as follows:

[0086] The regression trees are set to 100; the maximum tree depth is 6; the learning rate is 0.1; the sample subsampling ratio is 0.8; the feature ratio used per tree is 0.8; and the minimum leaf node sample weight is 1. Optimization is performed to minimize the loss function. The minimum loss function of 0.0081 is achieved when the hyperparameters Θ are 125, 14, 0.7, 0.9, and 3. Figure 4 , Figure 5 (As shown).

[0087] After training, the model output f(x) for all sample points under each crack propagation mode type is statistically analyzed. i The distribution analysis was performed, and the corrected pattern discrimination interval was constructed, as follows:

[0088] Pattern 1: Propagation along natural cracks: n>0.55;

[0089] Mode 2, suture height limited: 0.05 <n<0.55;

[0090] Mode 3, suture height expansion proceeded smoothly: -0.25 <n<0.05;

[0091] Mode 4, filtration along bedding and natural cracks: -0.45 <n<-0.25;

[0092] Mode 5, Breaking through natural cracks: n < -0.45.

[0093] To evaluate the agreement between the optimized model results and field monitoring data, the fracture network expansion and fracture height expansion results identified by the model for the entire well section were plotted on the same coordinate system with the corresponding field microseismic monitoring results, using the correlation coefficient R0. 2 To measure the degree of matching, such as Figure 7 , Figure 8 As shown.

[0094] The application results in the corresponding fractured section of well FY show that the results calculated by this method are ≥80% consistent with the field fracture monitoring results, which has strong field application value and provides a new method for identifying shale gas fracture propagation modes and evaluating post-fracture effects.

[0095] Therefore, this invention presents a diagnostic method for fracture network fracturing construction curves that integrates mechanistic constraints and data-driven approaches. By integrating the dynamic response of bottom hole net pressure and microseismic monitoring characteristics during fracturing operations, a theoretically interpretable and data-driven fracture propagation pattern index system is constructed. An XGBoost-based interval regression model is introduced, and a specific loss function is used for model fitting and correction, effectively mitigating the uncertainty caused by dependence on a single data source and significantly improving the accuracy of fracture propagation behavior identification. Simultaneously, the constructed pattern discrimination model has strong interpretability and field adaptability, guiding the optimization of fracturing design parameters and the control of fracture propagation direction, thereby contributing to efficient stimulation and stable production increase of shale gas reservoirs.

[0096] 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 them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method of diagnosing a fracture curve of a network fracture operation by fusing mechanism constraints and data-driven, characterized in that, The method comprises the following steps: S1, data acquisition and conversion: collecting wellhead pressure data and microseismic monitoring data in the fracturing process of the target block, and calculating the bottomhole net pressure according to the wellhead pressure data; S2, constructing a mechanism model: based on the trend of the bottomhole net pressure, the fracture propagation behavior is divided into multiple fracture propagation modes, an expansion index is introduced as a discrimination criterion for different fracture propagation modes, and a mechanism constraint interval is constructed; The calculation formula of the expansion index is: ； In the formula, is a time point corresponding to the current time, is is a wellbore bottom hole net pressure corresponding to the current time, is a time point corresponding to the previous time, is is a wellbore bottom hole net pressure corresponding to the previous time; S3, correlating field data: extracting microseismic characteristic parameters corresponding to the fracture propagation mode from the microseismic monitoring data; The microseismic characteristic parameters of the natural fracture propagation mode, the filtration mode along the bedding and natural fractures, and the fracture height limited mode are the ratio of the number of microseismic events to the reservoir reconstruction volume; the microseismic characteristic parameter of the fracture height smooth expansion mode is the ratio of the total length of the fracture to the height of the fracture; and the microseismic characteristic parameter of the fracture height smooth expansion mode is the height of the fracture; S4, reconstructing the discrimination criterion: fusing the mechanism constraint interval and the microseismic characteristic parameters, and using an XGBoost interval regression model to adaptively correct the mechanism constraint interval to obtain an adaptive fracture propagation mode discrimination model; The specific steps are: S41, fusing the mechanism constraint interval and the microseismic characteristic parameters to construct a training data set, which is represented as: ； In the formula, a microseismic characteristic parameter vector representing the crack propagation mode, , respectively represent the lower limit and the upper limit of the corresponding mechanism constraint interval, representing the total number of training set samples; S42, establishing an XGBoost interval regression model and defining a loss function for fitting learning, and outputting the corrected mode discrimination interval after training.

2. The method according to claim 1, wherein the method is characterized in that, In the S1, the calculation formula of the bottomhole net pressure is: ； where Pw represents wellhead pressure, Pf represents single-zone wellbore internal sand-carrying fluid column pressure z represents vertical depth from wellhead to wellbore, the integral of Ff represents single-zone wellbore flow friction, the integral of from wellhead to wellbore length, Fp represents perforation friction, σh represents minimum horizontal principal stress.

3. The method according to claim 1, wherein, The specific steps of the S2 are: S21, based on rock mechanics and fracture propagation theory, combining the geological characteristics and engineering practice of the target block, analyzing the quantitative relationship between the bottomhole net pressure and the fracture propagation behavior, and dividing multiple fracture propagation modes according to the quantitative relationship; S22, calculating the expansion index according to the trend of the bottomhole net pressure curve, and constructing the mechanism constraint interval of the expansion index corresponding to different fracture propagation modes.

4. The method according to claim 1, wherein the method is characterized in that: In the S2, the fracture propagation modes include propagation along natural fractures, fracture height limited, fracture height smooth expansion, filtration along bedding and natural fractures, and breakthrough of natural fractures. Among them, the breakthrough of natural fractures is identified by exclusion method.

5. The method according to claim 4, wherein the method is characterized in that: In the loss function of the S42, if the output of the XGBoost interval regression model falls within the mechanism constraint interval, there is no penalty, and if it exceeds the mechanism constraint interval, there is a square penalty; the training target is to minimize the overall loss function.

6. The method according to claim 5, wherein, The loss function is represented as: ； In the formula, represents the output result of the XGBoost interval regression model.

7. The method according to claim 6, wherein, The XGBoost interval regression model in the S4 is represented as: ； wherein is the number of regression trees, is the number of the subtree, represents a function space composed of all regression trees.

Images