Stratum fracture width inversion method based on auto-encoder
By constructing a fracture leakage model by combining formation parameters and leakage data with an autoencoder, the problem of fracture width prediction in fractured formation drilling was solved, and quantitative inversion of fracture width and effective reference for plugging operations were realized, thereby improving drilling safety and efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA PETROLEUM & CHEMICAL CORP
- Filing Date
- 2024-11-18
- Publication Date
- 2026-05-19
AI Technical Summary
Existing technologies struggle to predict fracture width accurately and in a timely manner during drilling in fractured formations, which makes it difficult to provide effective references for plugging operations in the event of lost circulation, thus affecting drilling safety and efficiency.
An autoencoder was used to combine formation parameters and leakage data to construct a fracture leakage model. The autoencoder was used to extract formation feature representations and optimize model parameters to achieve inversion analysis of fracture width.
It enables quantitative inversion analysis of fracture width, provides timely reference for plugging operations, reduces losses caused by leakage, and improves drilling safety and efficiency.
Smart Images
Figure CN122065623A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to drilling technology, and more particularly to a method for inverting formation fracture width based on an autoencoder. Background Technology
[0002] In drilling and completion of fractured formations, well leakage is one of the most serious problems causing reservoir damage. Leakage leads to a drop in the annular fluid level, causing wellbore instability, increasing costs, and in severe cases, even causing blowouts and wellbore collapse. When drilling encounters a fracture and leakage occurs, the drilling fluid flows into the fracture under the pressure differential at the bottom of the well, causing an increase in pressure within the fracture and changes in the effective stress on the fracture wall, thus altering the fracture width. Fracture leakage is characterized by large volume and rapid rate of loss. Safe and efficient drilling requires avoiding leakage and taking timely measures after it occurs. Adding plugging materials of different sizes and functions to the drilling fluid is the main method for controlling leakage. Therefore, the ability to accurately predict the fracture width and prepare the appropriate plugging drilling fluid largely determines the effectiveness of the plugging.
[0003] Existing predictive models for fracture-related leakage mainly utilize fracture parameters, drilling fluid parameters, drill string parameters, and formation parameters for simulation. However, these models are all leakage prediction models, which lack strong integration with the field situation and cannot effectively provide relevant information to the drilling site. When leakage occurs due to encountering fractures, it is difficult to analyze the fracture width in a timely manner based on information such as the amount of leakage, the rate of leakage, and drilling fluid parameters provided on-site. Therefore, there is an urgent need for an inversion model that combines fracture leakage models, which can provide better reference for plugging operations based on drilling site information and effectively reduce losses caused by leakage. Summary of the Invention
[0004] The technical problem to be solved by the present invention is to provide a method for inverting formation fracture width based on an autoencoder, addressing the deficiencies in the prior art.
[0005] The technical solution adopted by this invention to solve its technical problem is: a method for inverting formation fracture width based on an autoencoder, comprising the following steps:
[0006] 1) Obtain measured data on formation fracture leakage and collect formation parameters and drilling fluid parameters;
[0007] The leakage data includes: leakage amount and leakage rate;
[0008] The formation parameters include: well depth, well depth structure, formation pressure, and normal fracture stiffness;
[0009] The drilling fluid parameters include: drill string assembly, displacement, drilling fluid density, and drilling fluid rheological parameters;
[0010] 2) Establish an autoencoder to extract stratigraphic feature representations;
[0011] 3) Construct a crack leakage model to describe the relationship between crack development and leakage;
[0012] The governing equations for drilling fluid within two-dimensional fractures in the fracture leakage model are as follows:
[0013]
[0014] Where w is the crack width; τ is the total fluid shear stress, τ y ρ is the dynamic shear force; P is the drilling fluid density; α is the fracture inclination angle; x is the model's lateral coordinate; y is the model's longitudinal coordinate; t is time; k is the consistency coefficient.
[0015] 4) Combine the formation features learned by the autoencoder with the fracture leakage model to establish a fracture leakage coupling model;
[0016] 5) Combine the leakage rate (observation data) with the crack width (labeled data), optimize the model parameters, obtain the inversion model of crack parameters, input the observation data, and obtain the predicted value of the crack width parameter.
[0017] According to the above scheme, in step 2), the formation feature representation is extracted by establishing an autoencoder. The encoder part of the autoencoder maps the formation parameters and the missing data set to a low-dimensional code, and the decoder part reconstructs the low-dimensional code to obtain the formation parameter data.
[0018] According to the above scheme, in step 2), the autoencoder includes an input layer, a hidden layer, and an output layer; wherein, the input layer to the hidden layer is used for encoding, the hidden layer to the output layer is used for decoding, and the layers are fully interconnected.
[0019] The encoding process can be expressed as follows:
[0020] z = f(w) (1) x+b (1) )
[0021] The decoding process can be expressed by the following formula:
[0022]
[0023] In the formula: x is the input data; z is the hidden variable; To reconstruct the data; w (1) w (2) b is the weight matrix; (1) b (2) f is the bias vector; f(·) is the activation function.
[0024] According to the above solution, in step 3), the model boundary conditions are set as follows:
[0025] The fracture boundary condition can be approximated as a constant pressure boundary;
[0026] Assume that the initial pressure in the fracture remains unchanged as the formation pressure Pi, and the boundary condition is set as a constant pressure boundary:
[0027]
[0028] According to the above solution, in step 3), the pressure loss during drilling is determined by the rheological mode, viscosity, density, flow regime, and wellbore size of the drilling fluid.
[0029] According to the above solution, in step 3), the method for determining the pressure loss during drilling is as follows:
[0030] First, calculate the Reynolds number, determine whether the flow regime is laminar or turbulent based on the Reynolds number, and then calculate the annulus pressure drop according to the pressure drop formula;
[0031] Reynolds number in the annulus:
[0032]
[0033] Critical Reynolds number in the annulus:
[0034]
[0035] When NRe < NRec, the flow in the annulus is laminar, and the formula for calculating the pressure drop per unit length is:
[0036]
[0037] When NRe > NRec, the flow in the annulus is turbulent, and the formula for calculating the pressure drop per unit length is:
[0038]
[0039] In the formula, D in is the wellbore diameter; D out is the outer diameter of the drill pipe; q is the leakage rate.
[0040] The present invention also provides an electronic device, including: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors are caused to execute the method according to any one of the above solutions.
[0041] The present invention also provides a computer-readable storage medium, on which a computer program is stored, and characterized in that when the computer program is executed by a processor, the method according to any one of the above solutions is implemented.
[0042] The beneficial effects of this invention are:
[0043] This invention combines an autoencoder and a crack leakage model to propose a crack width inversion method based on measured leakage data. It couples the feature representation of the autoencoder with the mathematical model, thereby realizing quantitative inversion analysis of crack width. Attached Figure Description
[0044] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings:
[0045] Figure 1 This is a flowchart of a method according to an embodiment of the present invention. Detailed Implementation
[0046] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0047] like Figure 1 As shown, a method for inverting formation fracture width based on an autoencoder includes the following steps:
[0048] 1) Obtain measured data on formation fracture leakage and collect formation parameters and drilling fluid parameters;
[0049] The leakage data includes: leakage amount and leakage rate;
[0050] The formation parameters include: well depth, well depth structure, formation pressure, and normal fracture stiffness;
[0051] The drilling fluid parameters include: drill string assembly, displacement, drilling fluid density, and drilling fluid rheological parameters;
[0052] 2) Establish an autoencoder and extract stratigraphic feature representations using the established autoencoder;
[0053] Using the collected formation parameters and missing data, formation feature representations are extracted:
[0054] The encoder part of the autoencoder maps the formation parameters and the missing data set to a low-dimensional code, and the decoder part reconstructs the formation parameter data from the low-dimensional code; and makes the reconstructed data as similar as possible to the original input data.
[0055] An autoencoder consists of an input layer, a hidden layer, and an output layer; the input layer to the hidden layer is used for encoding, the hidden layer to the output layer is used for decoding, and the layers are fully connected to each other.
[0056] The encoding process can be expressed as follows:
[0057] z = f(w) (1) x+b (1) )
[0058] The decoding process can be expressed by the following formula:
[0059]
[0060] In the formula: x is the input data; z is the hidden variable; To reconstruct the data; w (1) w (2) b is the weight matrix; (1) b (2) f is the bias vector; f(·) is the activation function;
[0061] 4) Construct a crack leakage model to describe the relationship between crack development and leakage;
[0062] Control equations of drilling fluid in two-dimensional fractures
[0063]
[0064] Where w is the crack width; τ is the total fluid shear stress, τ y ρ is the dynamic shear force; P is the drilling fluid density; α is the fracture inclination angle; x is the model's lateral coordinate; y is the model's longitudinal coordinate; t is time; k is the consistency coefficient.
[0065] When the edge of the fracture is connected to a large storage space such as a large cave or underground river, the fracture boundary condition can be approximated as a constant pressure boundary.
[0066] The initial pressure within the fracture is assumed to be constant, with the formation pressure Pi remaining constant. The boundary condition is set to a constant pressure boundary.
[0067]
[0068] Pressure loss during drilling is determined by the rheological mode, viscosity, density, flow regime and wellbore size of the drilling fluid. In this embodiment, the Herba model is used for calculation.
[0069] First, determine the Reynolds number to identify whether the flow is laminar or turbulent. Then, calculate the annular pressure drop using the formula.
[0070] Reynolds number in the annulus:
[0071]
[0072] Critical Reynolds number for annular space:
[0073]
[0074] When NRe < NRec, the flow in the annulus is laminar, and the formula for calculating the pressure drop per unit length is:
[0075]
[0076] When NRe > NRec, the flow in the annulus is turbulent, and the formula for calculating the pressure drop per unit length is:
[0077]
[0078] In the formula, D in is the wellbore diameter; D out is the outside diameter of the drill pipe; q is the leakage rate;
[0079] 4) Combine the formation characteristics learned by the autoencoder with the fracture leakage model to establish a fracture leakage coupling model;
[0080] 5) Combine the leakage rate as the observed data with the fracture width as the marked data to optimize the model parameters, obtain the inversion model of the fracture parameters, input the observed data, and obtain the predicted value of the fracture width parameter.
[0081] [[ID=2,5]]Use the characteristics learned by the autoencoder as the input and adopt the gradient descent method for the inversion process;
[0082] The reconstruction error used to optimize the model parameters is:
[0083]
[0084] By minimizing the reconstruction error, the network parameters θ = {w (1) , b (1) , w (2) , b (2)} can be effectively learned.
[0085] In this process, we set the initial guess and select appropriate inversion parameters. Then, the inversion parameters are continuously updated through iteration, and the inversion result is continuously optimized. In each iteration, it is judged whether the convergence condition is reached. If not, the iteration continues.
[0086] After obtaining the optimal inversion result through iteration, the estimated value of the fracture width parameter is obtained. Then, the obtained result is verified and error analyzed to ensure the accuracy and reliability of the inversion result. Finally, the obtained fracture width is interpreted and applied to the actual plugging operation.
[0087] It should be understood that those of ordinary skill in the art can make improvements or transformations according to the above description, and all such improvements and transformations should fall within the protection scope of the appended claims of this invention.
Claims
1. A method for inverting formation fracture width based on an autoencoder, characterized in that, It includes the following steps: 1) Obtain the measured data of formation fracture leakage, and collect formation parameters and drilling fluid parameters; The leakage data includes: leakage volume and leakage rate; The formation parameters include: well depth, well depth structure, formation pressure, normal fracture stiffness; The drilling fluid parameters include: drill string assembly, displacement, drilling fluid density, drilling fluid rheological parameters; 2) Establish an autoencoder to extract formation feature representations; 3) Construct a fracture leakage model to describe the relationship between fracture development and leakage; 4) Combine the formation features learned by the autoencoder with the fracture leakage model to establish a fracture leakage coupling model; 5) Combine the leakage rate with the fracture width to optimize the model parameters, obtain an inversion model of the fracture width parameter, input the leakage rate into the inversion model, and obtain the predicted value of the fracture width parameter.
2. The formation fracture width inversion method based on autoencoder according to claim 1, characterized in that, In step 2), the formation feature representation is extracted through the established autoencoder. Among them, the encoder part of the autoencoder maps the formation parameters and the leakage data set to a low-dimensional code, and the decoder part reconstructs the low-dimensional code to obtain the formation parameter data.
3. The method for inverting formation fracture width based on an autoencoder according to claim 1, characterized in that, In step 2), the autoencoder includes an input layer, a hidden layer, and an output layer; among them, the input layer to the hidden layer is used for encoding, and the hidden layer to the output layer is used for decoding, and all layers are fully connected to each other; The encoding process is expressed by the formula: z=f(w (1) x+b (1) ) The decoding process is expressed by the formula: In the formula: x is the input data; z is the hidden variable; To reconstruct the data; w (1) w (2) b is the weight matrix; (1) b (2) f is the bias vector; f(·) is the activation function.
4. The method for inverting formation fracture width based on an autoencoder according to claim 1, characterized in that, The control equation of the drilling fluid in the two-dimensional fracture in the fracture leakage model in step 3) is as follows: Where w is the crack width; τ is the total fluid shear stress, τ y ρ is the dynamic shear force; P is the drilling fluid density; α is the fracture inclination angle; x is the model's lateral coordinate; y is the model's longitudinal coordinate; t is time; k is the consistency coefficient.
5. The method for inverting formation fracture width based on an autoencoder according to claim 1, characterized in that, In step 3), the model boundary conditions are set as follows: The fracture boundary condition can be approximated as a constant pressure boundary; Assume that the initial pressure in the fracture remains unchanged as the formation pressure Pi, and the boundary condition is set as a constant pressure boundary:
6. The method for inverting formation fracture width based on an autoencoder according to claim 1, characterized in that, In step 3), in the fracture leakage model, it is set that the pressure loss during the drilling process is determined by the rheological mode, viscosity, density, flow regime and wellbore size of the drilling fluid.
7. The formation fracture width inversion method based on an autoencoder according to claim 6, characterized in that, In step 3), the method for determining the pressure loss during the drilling process is as follows: First, calculate the Reynolds number, judge the flow regime as laminar flow or turbulent flow through the Reynolds number, and then calculate the annulus pressure drop according to the pressure drop formula; Reynolds number in the annulus: Critical Reynolds number in the annulus: When NRe < NRec, the flow in the annulus is laminar flow, and the formula for calculating the pressure drop per unit length is: When NRe > NRec, the flow in the annulus is turbulent flow, and the formula for calculating the pressure drop per unit length is: In the formula, D in It is the wellbore diameter; D out q is the outer diameter of the drill pipe; q is the leakage rate.
8. An electronic device, characterized in that it includes: one or more processors; and a storage device for storing one or more programs, wherein, when the one or more programs are executed by the one or more processors, the one or more processors are caused to execute the method according to any one of claims 1 to 7.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program realizes the method according to any one of claims 1 to 7 when executed by a processor.