Reservoir structure stress coefficient inversion method and system

Abstract

The application provides a reservoir structure stress coefficient inversion method and system, and belongs to the technical field of data processing. The method comprises the following steps: obtaining measured well diameter and lost circulation pressure data of a reservoir; constructing a neural network model taking a structure stress coefficient as a learnable parameter; calculating a horizontal ground stress according to the parameter, determining a first residual based on the horizontal ground stress and the measured well diameter, and determining a second residual based on a theoretical and measured lost circulation pressure; iteratively training the model according to the first and second residuals to construct a loss function, and taking the parameter value when the model converges as the structure stress coefficient. The application realizes the joint reverse constraint of the double physical responses of well wall collapse and formation leakage by deeply fusing continuous well diameter and sparse lost circulation pressure data, overcomes the inversion distortion caused by the small number of single data samples, and significantly improves the accuracy and robustness of the inversion result of the structure stress coefficient.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for inverting reservoir structural stress coefficients. Background Technology

[0002] In conventional drilling and oil and gas development, accurate assessment of reservoir in-situ stress is crucial for predicting wellbore instability, preventing drilling fluid loss, and optimizing fracturing design. Horizontal in-situ stress is primarily influenced by the weight of the overlying strata and geological tectonic activity; therefore, accurately determining the tectonic stress coefficient, which characterizes geological tectonic activity, is a key prerequisite for achieving precise in-situ stress assessment.

[0003] Currently, conventional methods for inverting tectonic stress coefficients mainly rely on ground leakage tests or small-scale hydraulic fracturing tests. This method involves conducting tests at specific depths in the well to obtain measured leakage pressure data of the formation. These measured leakage pressure data are then directly substituted into a single rock mechanics analytical formula, and the tectonic stress coefficient at that depth is calculated through mathematical derivation. This coefficient is then used to estimate the in-situ stress at other depths.

[0004] Due to operational costs and downhole safety risks, the number of leak tests conducted in actual engineering projects is extremely limited, resulting in a scarcity of measured leakage pressure data samples. This solution method, which relies solely on a single sparse data set and isolated mechanical formulas, fails to utilize the large amount of continuously distributed measured wellbore data along the well section. It also cannot combine wellbore collapse engineering phenomena to jointly constrain physical parameters in reverse, making it highly susceptible to parameter error propagation. Consequently, the robustness and accuracy of the inversion of the target reservoir structural stress coefficient are low. Summary of the Invention

[0005] This invention provides a method and system for inverting reservoir structural stress coefficients to address the shortcomings of existing technologies and improve the accuracy and robustness of structural stress coefficient inversion results.

[0006] This invention provides a method for inverting reservoir structural stress coefficients, comprising the following steps: Obtain logging data for multiple samples of the target reservoir, each sample corresponding to a depth point of the target reservoir. The logging data includes measured well diameter data for each sample and measured leakage pressure data for at least one sample. A neural network model is constructed, wherein the learnable parameter in the neural network model is the tectonic stress coefficient, which characterizes the degree of contribution of geological tectonic processes to horizontal geostress. Calculate the horizontal stress of each sample based on the learnable parameters; Based on the horizontal geostress and the measured well diameter data for each sample, a first deviation is determined for each sample, and a first residual is determined based on all the first deviations. For any sample with the measured leakage pressure data, the theoretical leakage pressure is calculated based on the horizontal ground stress of the corresponding sample. The second residual is determined based on the second deviation between the theoretical leakage pressure and the corresponding measured leakage pressure data. Based on the first residual and the second residual, construct a loss function; The neural network model is iteratively trained based on the loss function, and the learnable parameters are updated during the iterative training until the convergence condition is met. The value of the learnable parameters when the convergence condition is met is determined as the structural stress coefficient of the target reservoir.

[0007] According to the reservoir structural stress coefficient inversion method provided by the present invention, determining the first deviation of each sample based on the horizontal geostress and the measured well diameter data includes: Obtain the drilling fluid density and original wellbore size for each sample; Based on the horizontal geostress, drilling fluid density, original wellbore size, and measured well diameter data of each sample, calculate the radial stress and tangential stress around the well at the measured well diameter location of each sample. The first deviation for each sample is calculated based on the wellbore tangential stress and the wellbore radial stress for each sample.

[0008] According to a reservoir structural stress coefficient inversion method provided by the present invention, the step of calculating the peri-well radial stress and peri-well tangential stress at the measured well diameter location of each sample based on the horizontal geostress, the drilling fluid density, the original wellbore size, and the measured well diameter data of each sample includes: The number is calculated using the following formula. i The wellbore radial stress and the wellbore tangential stress of each sample: ; ; In the formula, For the first i The well periphery radial stress of each sample, For the first i The wellbore tangential stress of each sample, For the first i The maximum horizontal stress of each sample For the first i The minimum horizontal stress of a sample R The wellbore radius corresponding to the original wellbore size. For the first i Half of the measured well diameter corresponding to the measured well diameter data of each sample. θ For the well perimeter angle, For the first i The wellbore pressure corresponding to the drilling fluid density of each sample.

[0009] According to a reservoir structural stress coefficient inversion method provided by the present invention, the calculation of the first deviation of each sample based on the wellbore tangential stress and the wellbore radial stress includes: The number is calculated using the following formula. i The first bias of the samples: ; In the formula, For the first i The first deviation of each sample, For the first i The wellbore tangential stress of each sample, For the first i The well periphery radial stress of each sample, α The effective stress coefficient, For the first i The pore pressure corresponding to each sample For the first i The internal friction angle of the rock corresponding to each sample For the first i Rock cohesion corresponding to each sample; Determining the first residual based on all the first deviations includes: The first residual is calculated using the following formula: ; In the formula, For the first residual, n This represents the total number of samples.

[0010] According to the reservoir structural stress coefficient inversion method provided by the present invention, the step of calculating the theoretical leakage pressure based on the horizontal geostress of the corresponding sample for any sample having the measured leakage pressure data includes: The first, which has the measured leakage pressure data, is calculated using the following formula. i Theoretical leakage pressure per sample: ; ; In the formula, For the first i The theoretical leakage pressure for each sample, For the first i The maximum horizontal stress of each sample For the first i The minimum horizontal stress of a sample For the firsti The rock tensile strength corresponding to each sample For the first i The pore pressure corresponding to each sample The elastic coupling coefficient of the hole is... α The effective stress coefficient, For the first i Poisson's ratio of the rocks corresponding to each sample; The step of determining the second residual based on the second deviation between each of the theoretical leakage pressures and the corresponding measured leakage pressure data includes: The second residual is calculated using the following formula: ; In the formula, For the second residual, For the first i The measured leakage pressure data for each sample, m The number of samples containing the measured leakage pressure data.

[0011] According to the reservoir structural stress coefficient inversion method provided by the present invention, the step of constructing a loss function based on the first residual and the second residual includes: The first residual and the second residual are weighted and summed using the first weight and the second weight to obtain the total residual, wherein the first weight and the second weight are determined based on the difference in magnitude between the first residual and the second residual; The square of the total residual is determined as the loss function.

[0012] According to the present invention, a reservoir tectonic stress coefficient inversion method is provided, wherein the learnable parameters include a first tectonic stress coefficient and a second tectonic stress coefficient; the step of calculating the horizontal geostress of each sample based on the learnable parameters includes: The number is calculated using the following formula. i Maximum and minimum horizontal geostress for each sample: ; ; In the formula, For the first i The maximum horizontal stress of each sample, For the first i The minimum horizontal stress of each sample, The first structural stress coefficient, The second structural stress coefficient, For the first i The Poisson's ratio of the rocks corresponding to each sample For the first iThe overlying strata pressure corresponding to each sample α The effective stress coefficient, For the first i The pore pressure corresponding to each sample.

[0013] According to a reservoir structural stress coefficient inversion method provided by the present invention, the method further includes: During iterative training, for samples among the plurality of samples that do not have the measured leakage pressure data, the calculation of the theoretical leakage pressure and the second deviation for the corresponding sample is skipped.

[0014] According to a reservoir structural stress coefficient inversion method provided by the present invention, the method further includes: Rock mechanics experiments were conducted on the core samples of the target reservoir to obtain rock mechanics parameters. Based on the rock mechanics parameters and the well logging data of the target reservoir, a rock mechanics parameter profile of the target reservoir was established to determine the rock mechanics parameters corresponding to each sample. The rock mechanics parameters include at least one of cohesion, internal friction angle, Poisson's ratio, elastic modulus and tensile strength. By integrating the rock density of the target reservoir, an overlying strata pressure profile of the target reservoir is established to determine the overlying strata pressure corresponding to each sample. The pore pressure is calculated based on the sonic logging data of the target reservoir, and a pore pressure profile of the target reservoir is established to determine the pore pressure corresponding to each sample.

[0015] The present invention also provides a reservoir structural stress coefficient inversion system, comprising the following modules: The data acquisition module is used to acquire logging data of multiple samples of the target reservoir. Each sample corresponds to a depth point of the target reservoir. The logging data includes the measured well diameter data of each sample and the measured leakage pressure data of at least one sample. The model building module is used to build a neural network model, in which the learnable parameter is the tectonic stress coefficient, which characterizes the contribution of geological tectonic processes to horizontal geostress. The stress calculation module is used to calculate the horizontal stress of each sample based on the learnable parameters. The first residual determination module is used to determine the first deviation of each sample based on the horizontal geostress and the measured well diameter data of each sample, and to determine the first residual based on all the first deviations. The leakage pressure calculation module is used to calculate the theoretical leakage pressure based on the horizontal ground stress of any sample with the measured leakage pressure data. The second residual determination module is used to determine the second residual based on the second deviation between each of the theoretical leakage pressures and the corresponding measured leakage pressure data. The loss function construction module is used to construct a loss function based on the first residual and the second residual; The inversion training module is used to iteratively train the neural network model based on the loss function, update the learnable parameters during iterative training until the convergence condition is met, and determine the value of the learnable parameters when the convergence condition is met as the structural stress coefficient of the target reservoir.

[0016] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the reservoir structural stress coefficient inversion method as described above.

[0017] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the reservoir structural stress coefficient inversion method as described above.

[0018] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the reservoir structural stress coefficient inversion method as described above.

[0019] In summary, one or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: By acquiring reservoir data from multiple samples of the target reservoir, a neural network model is constructed. The tectonic stress coefficient, characterizing the contribution of geological structures to horizontal in-situ stress, is used as a learnable parameter in the neural network model. The horizontal in-situ stress of each sample is calculated based on the learnable parameter. A first residual is determined based on the horizontal in-situ stress and measured wellbore data for each sample. For any sample with measured leakage pressure data, the theoretical leakage pressure is calculated based on the corresponding sample's horizontal in-situ stress. A second residual is determined based on the second deviation between the theoretical and measured leakage pressure data. A loss function is constructed based on the first and second residuals. The neural network model is iteratively trained based on the loss function, and the learnable parameters are updated during the iterative training. The process continues until the convergence condition is met. The values ​​of the learnable parameters that meet the convergence condition are then determined as the structural stress coefficients of the target reservoir. This allows for the deep fusion of a large amount of continuously distributed measured wellbore data with sparse and discrete measured leakage pressure data. This achieves joint inverse constraints on physical parameters by wellbore collapse engineering phenomena and formation leakage engineering phenomena. It overcomes the shortcomings of traditional methods, such as inversion distortion caused by the limited number of leakage test tests and scarce data samples. At the same time, by using the backpropagation mechanism of neural networks, the complex nonlinear mechanical optimization process is transformed into gradient descent calculation, avoiding the problem that the solution of a single mechanical formula is easily affected by parameter error propagation. This significantly improves the robustness and calculation accuracy of the target reservoir structural stress coefficient inversion. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in this 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 This is one of the flowcharts illustrating the reservoir structural stress coefficient inversion method provided by the present invention.

[0022] Figure 2 This is a graph showing the loss function during the training process of the neural network model provided by this invention.

[0023] Figure 3 This invention provides a curve showing the predicted and actual values ​​of the structural stress coefficient during the training process.

[0024] Figure 4 This is a schematic diagram of the reservoir structural stress coefficient inversion system provided by the present invention.

[0025] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0027] It should be noted that in the description of this invention, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. The terms "upper," "lower," etc., indicating orientation or positional relationships according to the accompanying drawings, are only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the system or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0028] The terms "first," "second," etc., used in this invention are used to distinguish similar objects, not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class, without limiting the number of objects; for example, a first object can be one or more. Furthermore, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0029] The following is combined Figures 1 to 5 This invention describes the reservoir structural stress coefficient inversion method, system, electronic device, storage medium, and computer program product provided by the present invention.

[0030] In practical engineering applications, the executing entity can be a formation data processing workstation deployed at the drilling site, a geostress analysis and calculation platform integrated into the oilfield data center, or an integrated geological engineering processing system equipped with a rock mechanics calculation module and a deep learning framework. This embodiment selects the geostress inversion processing system (hereinafter referred to as the system) as the executing entity to explain each step of the reservoir structural stress coefficient inversion method.

[0031] This invention provides a method for inverting reservoir structural stress coefficients, referring to... Figure 1 , Figure 1 This is a schematic flowchart of the reservoir structural stress coefficient inversion method provided by the present invention, as shown below. Figure 1 As shown, the method includes steps 110 to 180: Step 110: Obtain logging data for multiple samples of the target reservoir. Each sample corresponds to a depth point of the target reservoir. The logging data includes the measured well diameter data of each sample and the measured leakage pressure data of at least one sample.

[0032] In oil drilling and reservoir development engineering, the target reservoir contains a large number of discrete measurement depth points along the drilling depth direction. The system performs depth discretization processing on the entire well section of the target reservoir, defining each independent depth point as a sample, and collecting corresponding logging data for each sample.

[0033] In well logging data, measured wellbore diameter data is continuously collected by wellbore logging instruments during drilling. It reflects the change in the actual wellbore size relative to the original drill bit size at each depth point and is a key engineering response parameter characterizing the degree of wellbore collapse and enlargement. Measured leakage pressure data is obtained through drain tests or small-scale hydraulic fracturing tests and directly reflects the formation's pressure-bearing capacity. Due to the limited number of drain tests, only some depth points in the target reservoir have measured leakage pressure data. The system separately marks the samples with measured leakage pressure data so that they can be retrieved as needed in subsequent calculations.

[0034] Step 120: Construct a neural network model. The learnable parameter in the neural network model is the tectonic stress coefficient, which characterizes the contribution of geological tectonic processes to horizontal geostress.

[0035] The tectonic stress coefficient is a core parameter in the formula for calculating horizontal geostress, reflecting the additional contribution of geological tectonic movement to horizontal stress. The accurate value of the tectonic stress coefficient directly determines the calculation accuracy of the maximum and minimum horizontal geostress.

[0036] The system employs a fully connected neural network architecture to construct the neural network model. The model comprises an input layer, several hidden layers, and an output layer, with non-linear mappings between layers via weight matrices and activation functions. Unlike traditional neural networks that only use layer weights as learnable parameters, this system directly embeds the constructed stress coefficients into the computational graph of the neural network model, allowing them to participate as independent learnable parameters in forward propagation calculations and backpropagation gradient updates. During initialization, the system assigns reasonable initial values ​​to the constructed stress coefficients, and subsequently iteratively updates them using a gradient descent algorithm.

[0037] In one implementation, the processor constructs a fully connected neural network as the neural network model. The input layer of the neural network model includes two feature parameters: reservoir density and acoustic transit time. The neural network model has three hidden layers with 64, 128, and 32 units respectively. The output layer of the neural network model includes one feature parameter, which is the total residual. A tectonic stress coefficient is incorporated into the neural network model as a learnable parameter, which is continuously updated during backpropagation in iterative training. Each time the reservoir density and acoustic transit time parameters are input to the neural network model, the rock cohesion, rock internal friction angle, rock Poisson's ratio, drilling fluid density, overlying strata pressure, pore pressure, wellbore diameter, and measured leakage pressure at the corresponding depth point are also input to support the physical calculation process within the neural network model.

[0038] Step 130: Calculate the horizontal ground stress for each sample based on the learnable parameters.

[0039] The system utilizes the values ​​of the tectonic stress coefficients in the current iteration, combined with the corresponding rock mechanics parameters, overlying strata pressure, and pore pressure for each sample, and calculates the maximum and minimum horizontal geostress at each sample based on the theoretical framework of elasticity and porosity elasticity. The calculation of horizontal geostress simultaneously considers the lateral constraint stress component generated by the self-weight of the overlying strata and the additional tectonic stress component generated by geological tectonic processes; the superposition of these two components constitutes the final horizontal geostress result.

[0040] Step 140: Determine the first deviation for each sample based on the horizontal geostress and measured well diameter data, and determine the first residual based on all first deviations.

[0041] The first deviation reflects the degree of discrepancy between the wellbore stress state at the measured wellbore location calculated using horizontal geostress and the formation rock strength criterion, under the current tectonic stress coefficient values.

[0042] When the wellbore collapses and the diameter expands, the measured well diameter is larger than the original drill bit size. At this time, the stress state around the well meets the rock failure criterion, and the first deviation approaches zero. When the calculated structural stress coefficient has a large deviation, the equation relationship between the calculated wellbore stress and the rock failure criterion deviates, and the first deviation increases accordingly.

[0043] The system statistically summarizes the first deviation of all samples to obtain the first residual. The first residual quantifies the degree of fit of the current structural stress coefficient value to the wellbore collapse engineering phenomenon throughout the entire well section.

[0044] Step 150: For any sample with measured leakage pressure data, calculate the theoretical leakage pressure based on the horizontal ground stress of the corresponding sample.

[0045] Theoretical leakage pressure is the critical drilling fluid pressure at which tensile failure of the formation occurs, derived from the hydraulic fracturing mechanism. The calculation of theoretical leakage pressure depends on the maximum horizontal stress, minimum horizontal stress, rock tensile strength, and pore pressure at the sample location.

[0046] The system calculates the corresponding theoretical leakage pressure for each sample marked with measured leakage pressure data, so that it can be compared with the measured leakage pressure data later.

[0047] Step 160: Determine the second residual based on the second deviation between each theoretical leakage pressure and the corresponding measured leakage pressure data.

[0048] The second deviation is the difference between the theoretical leakage pressure and the measured leakage pressure data, which directly reflects the accuracy of the current tectonic stress coefficient in predicting formation leakage engineering phenomena.

[0049] The system statistically summarizes the second deviation of all samples with measured leakage pressure data to obtain the second residual. The second residual fully utilizes the direct constraint ability of the drain test data on the structural stress coefficient.

[0050] Step 170: Construct the loss function based on the first residual and the second residual.

[0051] The loss function is composed of the first residual and the second residual. By combining the two residuals, the wellbore collapse constraint and formation leakage constraint are incorporated into the same optimization objective function, thereby realizing the joint reverse constraint of the structural stress coefficient on multiple types of engineering response data.

[0052] Step 180: Iteratively train the neural network model based on the loss function, update the learnable parameters during the iterative training until the convergence condition is met, and determine the value of the learnable parameters when the convergence condition is met as the structural stress coefficient of the target reservoir.

[0053] The system employs a backpropagation algorithm to calculate the gradient of the loss function with respect to the construction stress coefficient, and updates the construction stress coefficient according to the gradient descent direction. In each iteration, the system repeats steps 130 to 170, and updates the construction stress coefficient at the end of each iteration. The convergence criterion is that the loss function value drops below a preset threshold, or the change in the loss function value between two adjacent iterations is less than a preset accuracy requirement.

[0054] When the convergence condition is met, the system determines the current value of the tectonic stress coefficient as the final tectonic stress coefficient of the target reservoir, and applies the tectonic stress coefficient to the accurate solution of the horizontal geostress of the entire target reservoir.

[0055] Reference Figure 2 , Figure 2 This is a graph showing the loss function of the neural network model during training, as provided by this invention. The graph illustrates the trends of the training and validation loss values ​​during iterative training of the neural network model. As the number of training iterations increases, both the training and validation loss values ​​continuously decrease and stabilize after 400 iterations. Furthermore, throughout the entire iterative training process, the training and validation loss curves maintain a high degree of overlap, indicating that the iterative training process of the neural network model converges stably and does not exhibit overfitting. This ensures that the construction stress coefficient output by the neural network model possesses extremely high generalization ability and accuracy.

[0056] Reference Figure 3 , Figure 3 This invention provides a curve showing the predicted and actual values ​​of the structural stress coefficient during the training process. Figure 3 The solid blue line represents the predicted value of the first tectonic stress coefficient, the dashed blue line represents the actual value of the first tectonic stress coefficient, the solid yellow line represents the predicted value of the second tectonic stress coefficient, and the dashed yellow line represents the actual value of the second tectonic stress coefficient. The graph showing the relationship between the predicted and actual values ​​of the tectonic stress coefficients during training illustrates the trend of the predicted and actual values ​​approaching their corresponding actual values ​​with each iteration. As the number of iterations increases, the predicted and actual values ​​of the first and second tectonic stress coefficients continuously converge towards their actual values, eventually stabilizing after 600 iterations. In this stable state, the predicted and actual values ​​of both the first and second tectonic stress coefficients highly overlap, indicating that the geostress inversion system can accurately obtain the tectonic stress coefficients of the target reservoir through iterative training.

[0057] The reservoir structural stress coefficient inversion method provided in this invention uses the structural stress coefficient as a learnable parameter in a neural network model, fully utilizing the optimization characteristics of backpropagation in neural network models to transform the problem of solving the structural stress coefficient into an iterative optimization problem of the neural network model. Simultaneously, it fully utilizes continuously distributed measured wellbore diameter data and measured leakage pressure data throughout the target reservoir section, incorporating both wellbore collapse and formation leakage as dual constraints in the construction of the loss function. This overcomes the problems of limited number of leakage tests and small data samples, making it difficult to accurately assess the structural stress coefficient of the entire well or block. It also overcomes the problems of single-method solutions based on mechanical formulas being susceptible to parameter error propagation and unable to simultaneously match multiple engineering responses such as wellbore collapse and formation leakage. Therefore, it accurately inverts the structural stress coefficient, achieving precise solutions for geostress.

[0058] Based on the above embodiments, this embodiment provides a detailed description of the specific calculation method for horizontal ground stress in step 130.

[0059] In step 130, the learnable parameters include a first tectonic stress coefficient and a second tectonic stress coefficient. The first tectonic stress coefficient characterizes the contribution of geological tectonic activity to the direction of maximum horizontal stress, and the second tectonic stress coefficient characterizes the contribution of geological tectonic activity to the direction of minimum horizontal stress. The system calculates the first tectonic stress coefficient using the following formula. i Maximum and minimum horizontal geostress for each sample: ; ; In the formula, For the first i The maximum horizontal stress of each sample For the first i The minimum horizontal stress of a sample The first structural stress coefficient, The second structural stress coefficient, For the first i The Poisson's ratio of the rocks corresponding to each sample For the first i The overlying strata pressure corresponding to each sample α The effective stress coefficient, For the first i The pore pressure corresponding to each sample.

[0060] The above formula for calculating horizontal geostress is derived based on elasticity mechanics and porosity elasticity theory. In the formula... This term reflects the contribution of the overlying strata's self-weight to the lateral constraint stress in the horizontal direction, and it increases with the increase of the rock's Poisson's ratio. and The terms represent the additional contribution coefficients of tectonic stress superimposed on the lateral constraint stress, respectively. The values ​​of both are determined by the regional geological tectonic stress environment of the target reservoir. Effective stress coefficient. α Reflecting the influence of pore fluid pressure on the effective stress of the rock skeleton, To effectively control the pressure of the overlying strata.

[0061] During the forward propagation of the neural network model, density logging data and sonic transit time logging data undergo nonlinear transformations through the input layer and multiple hidden layers. Simultaneously, the system will... and Substituting the values ​​into the above formula, and combining the Poisson's ratio, overlying strata pressure, and pore pressure corresponding to each sample, the system calculates the maximum and minimum horizontal in-situ stress for each sample. During the initialization phase, the system calculates the first tectonic stress coefficient. Assign an initial value of 0.55 to the second structural stress coefficient. An initial value of 0.45 was assigned, which was set with reference to the existing geological and structural stress background knowledge of the region to ensure reasonable convergence of the iterative inversion process.

[0062] Based on the above embodiments, this embodiment describes in detail the process of determining the first deviation in step 140.

[0063] Step 141: Obtain the drilling fluid density and original wellbore size for each sample.

[0064] Drilling fluid density is a parameter that is monitored and recorded in real time on the ground using a drilling fluid density meter during drilling operations. The system uses an interpolation method to map the drilling fluid density data of each measurement point to each sample at the corresponding depth, ensuring that each sample has a complete drilling fluid density value.

[0065] The original wellbore size is determined by the drill bit diameter. The original wellbore size is usually a fixed value within the target reservoir. The system directly reads and stores the original wellbore size as a known geometric parameter.

[0066] Step 142: Based on the horizontal geostress, drilling fluid density, original wellbore size, and measured well diameter data of each sample, calculate the radial stress and tangential stress around the well at the measured well diameter location of each sample.

[0067] Perimeter stress is the stress state of the rock surrounding a circular wellbore under the combined action of horizontal geostress and wellbore fluid column pressure. Specifically, it includes radial stress along the radial direction and tangential stress along the tangential direction.

[0068] The system uses the Kirch formula for stress distribution around a circular hole in elasticity to calculate the wellbore stress. The calculation point is selected at the actual measured well diameter location, that is, half of the actual measured well diameter is used as the distance from the wellbore center to the calculation point, so that the actual wellbore enlargement degree is directly introduced into the wellbore stress calculation.

[0069] Step 143: Calculate the first deviation for each sample based on the wellbore tangential stress and wellbore radial stress.

[0070] At the depth point where wellbore collapse occurs, tangential stress is the maximum principal stress, and radial stress is the minimum principal stress. Together, they determine the stress state of the surrounding rock.

[0071] The system substitutes the wellbore tangential stress and wellbore radial stress into the rock failure criterion, calculates the difference between the stress state and the failure criterion equation, and the difference is the first deviation.

[0072] When the structural stress coefficient is accurately determined, the wellbore stress state obtained by theoretical calculation is in high agreement with the rock failure criterion, and the first deviation approaches zero.

[0073] This invention uses drilling fluid density, original wellbore size, and measured well diameter data to perform refined calculations of wellbore stress. It directly couples the actual wellbore collapse and diameter expansion phenomenon to the wellbore stress calculation process, so that the first deviation accurately characterizes the difference between the stress calculation result under the current learnable parameter values ​​and the measured engineering response of wellbore collapse, providing accurate physical constraints for the iterative training of the neural network model.

[0074] Based on the above embodiments, this embodiment provides a detailed description of the specific calculation formulas for the radial stress and tangential stress around the well in step 142.

[0075] The system calculates the first using the following formula. i The radial and tangential stresses around the wellbore for each sample are as follows: ; ; In the formula, For the first i The well perimeter radial stress of each sample, For the first i Well perimeter tangential stress of a sample For the first i The maximum horizontal stress of each sample For the first i The minimum horizontal stress of a sample R The wellbore radius corresponding to the original wellbore size. For the first iHalf of the measured well diameter corresponding to the measured well diameter of each sample. θ For the well perimeter angle, For the first i The wellbore pressure corresponding to the drilling fluid density of each sample.

[0076] The above formula is derived from the analytical solution of stress in a circular wellbore under plane strain conditions in elasticity, and is applicable to describing the stress distribution around a circular wellbore in an anisotropic horizontal geostress field. In the formula, The term reflects the geometric attenuation effect with reference to the original wellbore radius. This reflects the stress concentration effect caused by the horizontal stress difference. The results indicate that the wellbore stress varies periodically with azimuth angle, reaching extreme values ​​in the directions of maximum and minimum horizontal stress. Wellbore angle The value is determined by the wellbore measurement azimuth corresponding to each sample. The system determines the value of each sample based on the azimuth angle recorded by the logging instrument. value. It is calculated by multiplying the drilling fluid density by the product of gravitational acceleration and vertical depth, and represents the supporting pressure exerted by the drilling fluid column on the well wall in the wellbore.

[0077] This invention, through the aforementioned wellbore stress calculation formula, unifies the maximum horizontal stress, minimum horizontal stress, wellbore pressure, and measured wellbore diameter data into a wellbore stress field calculation framework for a circular wellbore. The stress calculation location is determined by the wellbore radius corresponding to the original wellbore size and half of the measured wellbore diameter. This ensures that the calculation results of the wellbore radial stress and wellbore tangential stress accurately reflect the true stress state at the measured wellbore diameter location after wellbore collapse and expansion, thereby improving the accuracy of the first deviation calculation.

[0078] Based on the above embodiments, this section provides a detailed description of the specific calculation formula for the first deviation and the calculation method for the first residual in step 143.

[0079] The system calculates the first using the following formula. i First bias of the sample: ; In the formula, For the first i The first bias of the sample For the first i Well perimeter tangential stress of a sample For the first i The well perimeter radial stress of each sample, α The effective stress coefficient, For the first i The pore pressure corresponding to each sample For the first iThe internal friction angle of the rock corresponding to each sample For the first i The rock cohesion corresponding to each sample.

[0080] The above formula is derived from the Mohr-Coulomb rock failure criterion. The Mohr-Coulomb criterion describes the critical relationship between the maximum and minimum principal stresses when rock undergoes shear failure: ; In the formula, For the maximum principal stress, For the minimum principal stress, It is the internal friction angle. C It represents cohesive force.

[0081] In the wellbore collapse region, the effective maximum principal stress is obtained by subtracting the pore pressure contribution from the tangential stress around the wellbore, and the effective minimum principal stress is obtained by subtracting the pore pressure contribution from the radial stress around the wellbore. Substituting these effective principal stress relationships into the Mohr-Coulomb criterion and subtracting both sides of the equation, we obtain the... i The first bias of each sample When the wellbore is in a critical failure state, The value equals zero; when the calculated stress state exceeds the failure criterion, Take a positive value; when the stress state does not reach the failure criterion, Take the negative value.

[0082] After calculating all n After the first deviation of each sample, the system calculates the first residual using the following formula: ; In the formula, For the first residual, n The total number of samples is denoted as . The first residual is the arithmetic mean of the absolute values ​​of the first deviations of all samples, treating positive and negative deviations equally to ensure that the optimization direction is always toward reducing the overall deviation.

[0083] In this embodiment of the invention, the Mohr-Coulomb criterion is modified by rearranging terms to construct a deviation function carrying measured wellbore data. The first deviation physically characterizes the distance between the wellbore stress state and the critical state of rock shear failure at the measured wellbore location. The average of the absolute values ​​of the first deviations of all samples is used as the first residual, so that the wellbore logging data continuously distributed throughout the target reservoir section participate in the inverse constraint of the learnable parameters, significantly expanding the number of effective constraint samples and improving the stability and accuracy of the structural stress coefficient inversion results.

[0084] Based on the above embodiments, this embodiment provides a detailed description of the specific calculation formula for the theoretical leakage pressure in step 150 and the calculation method for the second residual in step 160.

[0085] The system calculates the first [unit / item] with measured leakage pressure data using the following formula. i Theoretical leakage pressure per sample: ; ; In the formula, For the first i The theoretical leakage pressure per sample For the first i The maximum horizontal stress of each sample For the first i The minimum horizontal stress of a sample For the first i The rock tensile strength corresponding to each sample For the first i The pore pressure corresponding to each sample α The effective stress coefficient, For the first i The Poisson's ratio of the rocks corresponding to each sample For the first i The pore elastic coupling coefficient corresponding to each sample.

[0086] The above theoretical leakage pressure calculation formula is derived based on the tensile failure mechanism of hydraulic fracturing. When the drilling fluid pressure rises to the critical value for tensile fracturing of the formation, the minimum effective circumferential stress at the wellbore drops to zero, and formation leakage occurs. In the formula, This term reflects the influence of horizontal stress anisotropy on the degree of stress concentration in the wellbore. The term represents the ability of strata rocks to resist tensile failure. This item reflects the corrective effect of pore pressure on effective stress.

[0087] The system calculates the second residual using the following formula: ; In the formula, For the second residual, For the first i Measured leakage pressure data for each sample. m The number of samples with measured leakage pressure data is given. The second residual is the arithmetic mean of the absolute values ​​of the differences between the theoretical leakage pressure and the measured leakage pressure data of all samples with measured leakage pressure data, which quantifies the prediction accuracy of the current structural stress coefficient value for the drain test data.

[0088] This invention establishes a formula for calculating theoretical leakage pressure based on the maximum horizontal ground stress, the minimum horizontal ground stress, the tensile strength of rock, and pore pressure. The average value of the absolute value of the deviation between the theoretical leakage pressure and the measured leakage pressure data is used as the second residual. This transforms the measured engineering response of formation leakage into a direct inverse constraint on learnable parameters, making full use of the limited ground leakage test data in the process of structural stress coefficient inversion. This complements the first residual and significantly improves the constraint strength and accuracy of the inversion results.

[0089] Based on the above embodiments, this section describes in detail the specific construction method of the loss function in step 170.

[0090] The system uses the first weight With the second weight The total residual is obtained by weighted summation of the first and second residuals. : ; The first and second weights are determined based on the difference in magnitude between the first and second residuals. In practical engineering, and Because of differences in physical dimensions and numerical ranges, residuals can differ in magnitude. Directly adding them would mask the contribution of the smaller residual to the loss function, causing the corresponding physical constraints to fail. Therefore, when setting weights, the system analyzes the typical numerical ranges of the two residuals in the initial iteration rounds and determines the weights according to the principle of making the contributions of the two weighted residuals comparable in magnitude. q 1 and q The value of 2. In this embodiment, the system will... The value is 1, The value is set to 10 to compensate. Compared to Smaller numerical order of magnitude.

[0091] The system determines the square of the total residuals as the final loss function. L : ; The loss function is constructed by squaring the total residual. On the one hand, this ensures that the loss function is differentiable everywhere, which satisfies the gradient calculation requirements of the backpropagation algorithm. On the other hand, the squaring operation has an amplifying penalty effect on larger total residuals, prompting the iterative optimization process to give stronger correction to states with large deviations, thus accelerating the convergence speed of the loss function.

[0092] like Figure 2 As shown, the loss function increases with the number of iterations. LThe loss function curve continues to decrease, eventually stabilizing at a minimum value when the convergence condition is met. The monotonically decreasing trend of the loss function curve indicates that the joint optimization mechanism with dual physical constraints can stably guide the structural stress coefficient to converge toward the true value.

[0093] In this embodiment of the invention, a first weight and a second weight are determined based on the difference in magnitude between the first residual and the second residual. The first residual and the second residual are then weighted and summed. This avoids the problem that the constraint effect of a certain type of measured engineering response is weakened during iterative training due to the different magnitudes of the two types of residuals. This ensures that the wellbore collapse constraint and the formation leakage constraint are balanced in the loss function. At the same time, the square of the total residual is determined as the loss function, which enhances the sensitivity of the loss function to deviation, accelerates the convergence speed of the neural network model iterative training, and improves the accuracy of the structural stress coefficient inversion.

[0094] Based on the above embodiments, this section describes in detail the sample processing method for missing or lost pressure data during iterative training.

[0095] During the iterative training in step 180, the system processes all samples one by one in each iteration. For samples that do not have measured leakage pressure data, the system skips the theoretical leakage pressure calculation and second deviation calculation for the corresponding sample and directly proceeds to the processing flow of the next sample.

[0096] In actual drilling projects, due to limitations in the implementation conditions of ground leakage tests, each well typically only has 1 to 2 depth points with measured leakage pressure data, while a large number of depth points lack such data. During the initial data reading phase, the system marks each sample to indicate whether it has measured leakage pressure data. In each iteration, the system reads the marking status of the current sample. If the marking status indicates no measured leakage pressure data, the theoretical leakage pressure calculation and the second deviation calculation are skipped. If the marking status indicates measured leakage pressure data, the theoretical leakage pressure calculation and the second deviation calculation are performed normally, and the second deviation is included in the statistical summary of the second residual. Only for those with measured leakage pressure data m The samples are averaged, and no zero-padding or interpolation is performed on samples with missing data, thereby ensuring the accuracy of the physical meaning of the second residual.

[0097] This invention skips the calculation of theoretical leakage pressure and second deviation for samples without measured leakage pressure data, enabling the neural network model to complete iterative training normally under actual engineering conditions where measured leakage pressure data is sparse. It takes into account both the sparsity of measured data and the integrity of the inversion process, reduces invalid calculations, and improves the efficiency of iterative training and the applicability of the method in actual engineering.

[0098] Based on the above embodiments, this embodiment describes in detail the methods for establishing the target reservoir rock mechanical parameter profile, the overlying strata pressure profile, and the pore pressure profile. The reservoir structural stress coefficient inversion method also includes steps 101 to 103.

[0099] Step 101: Conduct rock mechanics experiments on the core of the target reservoir to obtain rock mechanics parameters. Based on the rock mechanics parameters and the well logging data of the target reservoir, establish a rock mechanics parameter profile of the target reservoir to determine the rock mechanics parameters corresponding to each sample. The rock mechanics parameters include at least one of cohesion, internal friction angle, Poisson's ratio, elastic modulus, and tensile strength.

[0100] In the rock mechanics experimental stage, core samples retrieved from the target reservoir were mechanically processed to prepare standard cylindrical cores with a diameter of 1 inch and a height of 2 inches. Based on the experimental results, uniaxial compression tests were conducted to determine the rock's elastic modulus and uniaxial compressive strength; triaxial compression tests were performed under different confining pressures to determine the rock's cohesion and internal friction angle using the strength envelopes under different confining pressures; and Brazilian splitting tests were conducted to determine the rock's tensile strength by loading and failing the disc-shaped core in the diameter direction. Rock mechanics parameters include at least one of cohesion, internal friction angle, Poisson's ratio, elastic modulus, and tensile strength.

[0101] In the stage of establishing rock mechanics parameter profiles, the system establishes a correlation between the rock mechanics experimental results at finite depth points and the continuously distributed density logging data and sonic transit time logging data. Through the continuous profile calculation method driven by logging curves, the system interpolates and calculates the cohesion, internal friction angle, Poisson's ratio, elastic modulus and tensile strength of each sample in the entire target reservoir section, forming a complete rock mechanics parameter profile.

[0102] Step 102: Integrate the rock density of the target reservoir to establish the overlying strata pressure profile of the target reservoir, so as to determine the overlying strata pressure corresponding to each sample.

[0103] Overlying stratum pressure is the vertical pressure generated by the self-weight of all rocks above a certain depth point at that depth. The system integrates the density logging curve layer by layer from the surface to the target reservoir along the vertical depth direction. The integration formula is the cumulative summation of the product of the overlying stratum density, gravitational acceleration, and vertical depth micro-element, thereby forming a continuous overlying stratum pressure profile throughout the entire well section of the target reservoir.

[0104] Step 103: Calculate the pore pressure based on the sonic logging data of the target reservoir, establish the pore pressure profile of the target reservoir, and determine the pore pressure corresponding to each sample.

[0105] The system employs a pore pressure prediction method based on sonic transit time logging data. It compares the measured sonic transit time with the normal compaction trend line, quantifies the formation overpressure using the sonic transit time deviation, and calculates the pore pressure at each depth point using the effective stress principle, forming a continuous pore pressure profile. The normal compaction trend line is determined by fitting sonic transit time data from the normally compacted strata above the target reservoir.

[0106] This invention establishes a rock mechanics parameter profile of the entire target reservoir section by combining rock mechanics experiments with well logging data. It also establishes overlying strata pressure profiles and pore pressure profiles by integrating rock density and using sonic logging data. This ensures that the sample corresponding to each depth point of the target reservoir has complete and accurate rock mechanics parameters, overlying strata pressure, and pore pressure, providing reliable basic data support for horizontal stress calculation, first deviation calculation, and theoretical leakage pressure calculation. Overall, this ensures the accuracy of the tectonic stress coefficient inversion results.

[0107] Reference Figure 4 , Figure 4 This is a schematic diagram of the reservoir structural stress coefficient inversion system provided by the present invention. The system includes: The data acquisition module is used to acquire logging data of multiple samples from the target reservoir. Each sample corresponds to a depth point in the target reservoir. The logging data includes the measured well diameter data of each sample and the measured leakage pressure data of at least one sample. The model building module is used to build a neural network model. The learnable parameter in the neural network model is the tectonic stress coefficient, which characterizes the contribution of geological tectonic processes to horizontal geostress. The stress calculation module is used to calculate the horizontal stress of each sample based on learnable parameters. The first residual determination module is used to determine the first deviation of each sample based on the horizontal geostress and measured well diameter data of each sample, and to determine the first residual based on all the first deviations. The leakage pressure calculation module is used to calculate the theoretical leakage pressure based on the horizontal stress of any sample with measured leakage pressure data. The second residual determination module is used to determine the second residual based on the second deviation between each theoretical leakage pressure and the corresponding measured leakage pressure data. The loss function construction module is used to construct the loss function based on the first residual and the second residual; The inversion training module is used to iteratively train the neural network model based on the loss function. During the iterative training, the learnable parameters are updated until the convergence condition is met. The values ​​of the learnable parameters when the convergence condition is met are determined as the structural stress coefficients of the target reservoir.

[0108] It should be noted that the reservoir structural stress coefficient inversion system provided by the present invention can execute the reservoir structural stress coefficient inversion method of any of the above embodiments during specific operation, which will not be elaborated in this embodiment.

[0109] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 5 As shown, the electronic device may include a processor 510, a communication interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other through the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute the reservoir structural stress coefficient inversion method provided in the above embodiments.

[0110] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0111] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, and when the program instructions are executed by a computer, the computer is able to execute the reservoir structural stress coefficient inversion method provided in the above embodiments.

[0112] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the reservoir structural stress coefficient inversion method provided in the above embodiments.

[0113] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0114] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0115] 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 the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for inverting reservoir structural stress coefficients, characterized in that, include: Obtain logging data for multiple samples of the target reservoir, each sample corresponding to a depth point of the target reservoir. The logging data includes measured well diameter data for each sample and measured leakage pressure data for at least one sample. A neural network model is constructed, wherein the learnable parameter in the neural network model is the tectonic stress coefficient, which characterizes the degree of contribution of geological tectonic processes to horizontal geostress. Calculate the horizontal stress of each sample based on the learnable parameters; Based on the horizontal geostress and the measured well diameter data for each sample, a first deviation is determined for each sample, and a first residual is determined based on all the first deviations. For any sample with the measured leakage pressure data, the theoretical leakage pressure is calculated based on the horizontal ground stress of the corresponding sample. The second residual is determined based on the second deviation between the theoretical leakage pressure and the corresponding measured leakage pressure data. Based on the first residual and the second residual, construct a loss function; The neural network model is iteratively trained based on the loss function, and the learnable parameters are updated during the iterative training until the convergence condition is met. The value of the learnable parameters when the convergence condition is met is determined as the structural stress coefficient of the target reservoir.

2. The method according to claim 1, characterized in that, The determination of the first deviation for each sample based on the horizontal geostress and the measured well diameter data includes: Obtain the drilling fluid density and original wellbore size for each sample; Based on the horizontal geostress, drilling fluid density, original wellbore size, and measured well diameter data of each sample, calculate the radial stress and tangential stress around the well at the measured well diameter location of each sample. The first deviation for each sample is calculated based on the wellbore tangential stress and the wellbore radial stress for each sample.

3. The method according to claim 2, characterized in that, The calculation of the radial and tangential stress around the wellbore at the measured wellbore location for each sample, based on the horizontal geostress, drilling fluid density, original wellbore size, and measured wellbore diameter data, includes: The number is calculated using the following formula. i The wellbore radial stress and the wellbore tangential stress of each sample: ； ； In the formula, For the first i The well periphery radial stress of each sample, For the first i The wellbore tangential stress of each sample, The maximum horizontal ground stress for the i-th sample. For the first i The minimum horizontal stress of a sample R The wellbore radius corresponding to the original wellbore size. For the first i Half of the measured well diameter corresponding to the measured well diameter data of each sample. θ For the well perimeter angle, For the first i The wellbore pressure corresponding to the drilling fluid density of each sample.

4. The method according to claim 2, characterized in that, The calculation of the first deviation for each sample based on the wellbore tangential stress and the wellbore radial stress includes: The number is calculated using the following formula. i The first bias of the samples: ； In the formula, For the first i The first deviation of each sample, For the first i The wellbore tangential stress of each sample, For the first i The well periphery radial stress of each sample, α The effective stress coefficient, For the first i The pore pressure corresponding to each sample For the first i The internal friction angle of the rock corresponding to each sample For the first i Rock cohesion corresponding to each sample; Determining the first residual based on all the first deviations includes: The first residual is calculated using the following formula: ； In the formula, For the first residual, n This represents the total number of samples.

5. The method according to claim 1, characterized in that, The calculation of theoretical leakage pressure based on the horizontal ground stress of the corresponding sample for any sample with the measured leakage pressure data includes: The first, which has the measured leakage pressure data, is calculated using the following formula. i Theoretical leakage pressure per sample: ； ； In the formula, For the first i The theoretical leakage pressure for each sample, For the first i The maximum horizontal stress of each sample For the first i The minimum horizontal stress of a sample For the first i The rock tensile strength corresponding to each sample For the first i The pore pressure corresponding to each sample The elastic coupling coefficient of the hole is... α The effective stress coefficient, For the first i Poisson's ratio of the rocks corresponding to each sample; The step of determining the second residual based on the second deviation between each of the theoretical leakage pressures and the corresponding measured leakage pressure data includes: The second residual is calculated using the following formula: ； In the formula, For the second residual, For the first i The measured leakage pressure data for each sample, m The number of samples containing the measured leakage pressure data.

6. The method according to claim 1, characterized in that, The step of constructing a loss function based on the first residual and the second residual includes: The first residual and the second residual are weighted and summed using the first weight and the second weight to obtain the total residual, wherein the first weight and the second weight are determined based on the difference in magnitude between the first residual and the second residual; The square of the total residual is determined as the loss function.

7. The method according to claim 1, characterized in that, The learnable parameters include a first tectonic stress coefficient and a second tectonic stress coefficient; the calculation of the horizontal ground stress of each sample based on the learnable parameters includes: The number is calculated using the following formula. i Maximum and minimum horizontal geostress for each sample: ； ； In the formula, For the first i The maximum horizontal stress of each sample, For the first i The minimum horizontal stress of each sample, The first structural stress coefficient, The second structural stress coefficient, For the first i The Poisson's ratio of the rocks corresponding to each sample For the first i The overlying strata pressure corresponding to each sample α The effective stress coefficient, For the first i The pore pressure corresponding to each sample.

8. The method according to claim 1, characterized in that, The method further includes: During iterative training, for samples among the plurality of samples that do not have the measured leakage pressure data, the calculation of the theoretical leakage pressure and the second deviation for the corresponding sample is skipped.

9. The method according to claim 1, characterized in that, The method further includes: Rock mechanics experiments were conducted on the core samples of the target reservoir to obtain rock mechanics parameters. Based on the rock mechanics parameters and the well logging data of the target reservoir, a rock mechanics parameter profile of the target reservoir was established to determine the rock mechanics parameters corresponding to each sample. The rock mechanics parameters include at least one of cohesion, internal friction angle, Poisson's ratio, elastic modulus and tensile strength. By integrating the rock density of the target reservoir, an overlying strata pressure profile of the target reservoir is established to determine the overlying strata pressure corresponding to each sample. The pore pressure is calculated based on the sonic logging data of the target reservoir, and a pore pressure profile of the target reservoir is established to determine the pore pressure corresponding to each sample.

10. A reservoir structural stress coefficient inversion system, characterized in that, include: The data acquisition module is used to acquire logging data of multiple samples of the target reservoir. Each sample corresponds to a depth point of the target reservoir. The logging data includes the measured well diameter data of each sample and the measured leakage pressure data of at least one sample. The model building module is used to build a neural network model, in which the learnable parameter is the tectonic stress coefficient, which characterizes the contribution of geological tectonic processes to horizontal geostress. The stress calculation module is used to calculate the horizontal stress of each sample based on the learnable parameters. The first residual determination module is used to determine the first deviation of each sample based on the horizontal geostress and the measured well diameter data of each sample, and to determine the first residual based on all the first deviations. The leakage pressure calculation module is used to calculate the theoretical leakage pressure based on the horizontal ground stress of any sample with the measured leakage pressure data. The second residual determination module is used to determine the second residual based on the second deviation between each of the theoretical leakage pressures and the corresponding measured leakage pressure data. The loss function construction module is used to construct a loss function based on the first residual and the second residual; The inversion training module is used to iteratively train the neural network model based on the loss function, update the learnable parameters during iterative training until the convergence condition is met, and determine the value of the learnable parameters when the convergence condition is met as the structural stress coefficient of the target reservoir.

Images