A borehole wall instability analysis method based on big data

Abstract

The present disclosure provides a wellbore instability analysis method based on big data, which relates to the field of wellbore instability. The wellbore instability prediction based on big data analysis and deep learning, combined with existing wellbore instability theory and experimental research results, finds the main control factors and related factors of wellbore instability in the studied block. Based on data classification, a big data model is established to improve the accuracy of analysis and prediction through big data learning. Through the application of the big data model, the deep learning technology is used for regional wellbore instability evaluation and research, and finally the regional wellbore instability evaluation and prevention technology based on big data is formed. The wellbore instability analysis tool based on big data analysis mainly provides the safety density window for preventing wellbore instability, the control analysis of hole diameter change, the optimization of drilling fluid performance and additive dosage, adjacent well data analysis, ECD analysis and drilling time effect analysis through data analysis, so as to reduce the hydration swelling effect of the target formation rock and prevent wellbore instability. The present disclosure is suitable for wellbore instability prediction.

Description

Technical Field

[0001] This disclosure relates to the field of wellbore instability, and in particular to a wellbore instability analysis method based on big data. Background Technology

[0002] Wellbore instability has long been a major challenge in drilling and completion operations. Especially in recent years, the geological conditions faced in drilling have become increasingly complex, and the number of wells with complex structures such as extended reach wells and ultra-long horizontal wells has increased significantly, making wellbore instability during drilling and completion even more prominent. According to statistics from Academician Luo Pingya's report, wellbore instability accounts for 24% of the losses caused by complex problems encountered in drilling, making it a common, difficult, and impactful technical challenge in complex downhole drilling accidents.

[0003] Changqing Drilling drilled 2,285 wells in 2017, 2,269 wells in 2018, and 1,323 wells in 2019. Every year, a large number of accident handling cases are accumulated. There is an urgent need to correctly analyze and process historical big data and establish an artificial intelligence program based on machine learning technology to predict wellbore instability. This will help optimize the technical solutions in the later stage and reduce the occurrence of accidents caused by wellbore instability. Summary of the Invention

[0004] This paper proposes a wellbore instability analysis method based on big data. The research on wellbore instability analysis tools based on big data analysis focuses on optimizing the drilling fluid system and performance parameters of the block formation, optimizing the dosage and configuration of treatment agents, analyzing how to reduce well diameter changes, identifying reasonable safe density windows and density warning windows, providing data support for increasing the collapse cycle to reduce the drilling time effect, and analyzing data from adjacent wells to find safe drilling parameters. Finally, a wellbore instability prediction model based on big data analysis and deep learning is established.

[0005] This disclosure provides a big data-based wellbore instability analysis method, which uses big data analysis and deep learning for wellbore instability prediction. It primarily combines existing wellbore instability theories and experimental research results to identify the main and related factors controlling wellbore instability in the studied area. Based on data classification, a big data model is established. The accuracy of analysis and prediction is improved through big data learning. By applying the big data model and utilizing deep learning technology to evaluate wellbore instability in the Changqing area, a big data-based wellbore instability evaluation and prevention technology for the Changqing region is ultimately formed.

[0006] According to a first aspect of the present disclosure, a wellbore instability analysis method based on big data is provided, the method comprising the following steps:

[0007] 101. Establish a wellbore instability analysis database and build a data interface based on the data collected by the data acquisition platform; preprocess the wellbore instability data analysis scan by traversing all drilling data from previous years, build a wellbore elastic model, obtain the stress state on the wellbore, and train the wellbore instability data from previous years.

[0008] 102. Training with historical wellbore instability data involves analyzing well diameter changes in the block formation and all construction schemes within the block. Drilling fluid density is obtained based on stress components under different conditions and substituted into the criteria for calculation to obtain shear stress on the wellbore shear surface and determine the wellbore stability.

[0009] 103. Based on the data from adjacent wells, analyze all construction plans within the block to obtain the safe density window and density warning window; based on the normalized historical logging data within the block, train and model using a BP neural network, and establish a formation pressure model through the training; this formation pressure model is a trained wellbore instability prediction neural network model for predicting wellbore instability.

[0010] Preferably, the data acquisition platform in step 101 collects data, which means collecting well site data and on-site engineering data, and preparing data collection before building machine learning for training; and builds data interfaces for extracting, cleaning and converting well caliper data, accident complexity information, gamma value, well foundation information, logging data, drilling parameters, drilling fluid performance and electrical logging information.

[0011] Preferably, the preprocessing of wellbore instability data analysis and scanning by traversing all historical drilling data in step 101 refers to acquiring parameters affecting wellbore instability, blocks, and formations prone to wellbore instability from all historical drilling data to construct a wellbore elastic model. The formula for this wellbore elastic model is as follows:

[0012] Radial stress component σ in cylindrical coordinates r for:

[0013]

[0014] Tangential stress component σ in cylindrical coordinates θ for:

[0015]

[0016] Axial stress component σ in cylindrical coordinate system z for:

[0017]

[0018] Shear stress component τ in cylindrical coordinates rθ for:

[0019]

[0020] Shear stress component τ in cylindrical coordinates θz for:

[0021]

[0022] Shear stress component τ in cylindrical coordinates rz for:

[0023]

[0024] In the formula: P P δ is the pore pressure; μ is Poisson's ratio; δ is the wellbore permeability coefficient.

[0025] θ is the wellbore angle (relative to the x-axis); φ1 is the porosity; α1 is the Biot coefficient; P W α is the fluid pressure inside the wellbore; α is the well inclination angle. R is the azimuth angle; R is the wellbore radius; r is the radial coordinate.

[0026] When the well wall is impermeable, the well wall permeability coefficient δ = 0;

[0027] When the well wall is permeable, the well wall permeability coefficient δ = 1;

[0028] Let r = R; substitute into equations (1), (2), and (3) to obtain the stress state on the well wall; the formula for the stress state on the well wall is:

[0029] σ r =Ρ W -δφ1(P W -P P (7)

[0030]

[0031]

[0032] τ rθ =0 (10)

[0033]

[0034] τ rz =0 (12)

[0035] τ rθ τ θz τ rz These are the shear stress components in cylindrical coordinates.

[0036] Preferably, step 102, training with historical wellbore instability data, involves analyzing wellbore diameter variations in the block formation and all construction schemes within the block. Drilling fluid density is obtained based on stress components under different conditions, and these are substituted into the criteria for calculation to determine the shear stress on the corresponding wellbore shear surface, thus assessing wellbore stability. This refers to calculating the radial stress component σ under different conditions. r tangential stress component σ θ axial stress component σ z The drilling fluid density was obtained and substituted into the Coulomb-Mohr criterion for calculation to obtain the corresponding shear stress on the wellbore shear surface;

[0037] The Coulomb-Moore criterion states that when the shear stress on the shear surface is greater than or equal to the rock's inherent shear strength S0 plus the frictional resistance fσ acting on the shear surface, the wellbore will experience shear failure; otherwise, the wellbore will be a stable structure.

[0038] |τ|≥fσ+S0 (13)

[0039] Where τ is the shear stress on the shear plane; S0 is the inherent shear strength of the rock; and f is the internal friction coefficient of the rock.

[0040] Preferably, the radial stress component σ r tangential stress component σ θ axial stress component σ z The expression was substituted into the Coulomb-Mohr criterion under different conditions to obtain the corresponding critical drilling fluid density P in the wellbore direction. w ;

[0041] a. When σ θ ≥σ z ≥σ r hour,

[0042] b. When σ z ≥σ θ ≥σ r hour,

[0043] c. When σ z ≥σ r ≥σ θ hour,

[0044] d. When σ r ≥σ θ ≥σ z hour,

[0045] e. When σ r ≥σ z ≥σ θ At that time, Pw ≥σ v tan 2 β-α1P f (tan 2 β-1)+C0

[0046] f, when σ θ ≥σ r ≥σ z At that time, P w ≤2σ h -σ v tan 2 β+α1P f (tan 2 β-1)-C0

[0047] Where, σ r The radial stress component in cylindrical coordinates; σ θ σ represents the tangential stress component in cylindrical coordinates. z This represents the axial stress components in cylindrical coordinates.

[0048] Preferably, step 102, training with historical wellbore instability data, involves analyzing wellbore diameter variations in the block formation and all construction schemes within the block. Drilling fluid density is obtained based on stress components under different conditions and substituted into the criteria for calculation. This refers to calculating the radial stress component σ... r tangential stress component σ θ axial stress component σ z The expression is substituted into the modified Lade criterion (Lade yield) under different conditions to solve for the corresponding critical drilling fluid density P in the wellbore direction. w ;

[0049] The Lade criterion has the following expression:

[0050]

[0051] Among them, I1=σ1+σ2+σ3; I3=σ1σ2σ3; P a Atmospheric pressure; m, η1 are material constants;

[0052] The modified Lade criterion has the following expression:

[0053] (I″1) 3 / I″3=27+η (15)

[0054] in,

[0055] I″1=(σ r +S1-P P )+(σθ +S1-P P )+(σ Z +S1-P P )

[0056]

[0057] Shear stress components in cylindrical coordinates

[0058] S1 and η are material constants; S1 = S0 / tanφ; η = 4tanφ 2 φ(9-7sinφ) / (1-sinφ)

[0059] P P φ is the pore pressure; φ is the internal friction angle.

[0060] S0 represents the inherent shear strength or internal friction of the rock.

[0061] q represents shear strength; q = (1 / 2)[(σ1 - σ2)] 2 +(σ2-σ3) 2 +(σ1-σ3) 2 ] 1 / 2

[0062] When the wellbore is impermeable, the wellbore permeability coefficient δ = 0. Substituting equation (15) into the wellbore elastic model, the critical drilling fluid density P in the wellbore direction is obtained. W The expression is:

[0063]

[0064] In the formula:

[0065] A = σ Z +S1-P P ;

[0066]

[0067]

[0068] D=(σ θn +σ Z +3S1-3P P ) 3 / (27+η);

[0069]

[0070]

[0071]

[0072] Preferably, the critical drilling fluid density value in the wellbore direction is predicted through big data learning analysis. This prediction analysis generates a three-pressure profile for the block, which includes a fracture pressure prediction assessment, a collapse pressure prediction assessment, and a pore pressure prediction assessment. At the same time, data from adjacent wells are analyzed to obtain safe drilling parameters and generate a suggested construction plan.

[0073] Preferably, analyzing data from adjacent wells to obtain safe drilling parameters refers to analyzing data from adjacent wells in each formation to obtain safe density windows and density warning windows. The data analysis of these adjacent wells includes the drilling fluid system and performance parameters of the formation, optimization of treatment agent dosage and configuration, and analysis of how to reduce well diameter changes.

[0074] The data for adjacent wells includes performance range, treatment agent dosage, mechanical drilling rate, well foundation information, drilling parameters, accident complexity, well enlargement rate, gamma value, drilling cycle, cost, and electrical logging information.

[0075] The generated suggested construction plan refers to the construction plan selected for this block according to different dimensions. The selected dimensions include wellbore enlargement rate, drilling cycle, mechanical drilling rate, and drilling cost. The suggested construction plan indicators include priority density range, standard density range, electrical logging resistance, and gamma value.

[0076] Preferably, before analyzing the data of adjacent wells in each stratum and obtaining the safe density window and density warning window, the data of the adjacent wells are selected, and data cleaning is used to process the data defects of different adjacent wells. Outliers in the data of the adjacent wells are identified and processed using a clustering algorithm. The data is then processed by pattern classification and normalization and used as input for training the BP neural network. The BP neural network is modeled using a training function. The number of nodes in the input layer of the BP neural network is set to the number of fault features. The number of nodes in the hidden layer and the learning rate parameter are set using empirical values ​​and a trial-and-error method.

[0077] Preferably, the construction plan, safety density window, and density warning window for this block are modeled based on a BP neural network after normalization of historical logging data within the block. The formation pressure model is established through this training, and the formation fracture pressure profile model and formation collapse pressure profile model are established through training based on the BP neural network.

[0078] Among them, the formation fracture pressure profile model refers to the formation fracture pressure based on the initiation of fractures or the reopening of existing fractures in the exposed formation under external forces. This formation fracture is caused by excessively high drilling fluid density, resulting in the effective tangential stress on the rock exceeding the tensile strength of the rock. Its calculation models mainly include:

[0079] Practical Rupture Pressure Prediction Model

[0080]

[0081] Prediction models for two horizontal principal geostresses:

[0082]

[0083]

[0084] In the formula:

[0085] P f Formation fracture pressure, MPa; E s The static Young's modulus of the rock is given in MPa.

[0086] μ s P represents the static Poisson's ratio of the rock, which is dimensionless. p Formation pore pressure, MPa;

[0087] α is the effective stress coefficient, which is dimensionless; σ H The maximum horizontal principal stress is expressed in MPa.

[0088] σ h The minimum principal stress is σ, MPa; v The pressure of the overlying strata is MPa;

[0089] S t The tensile strength of the rock is expressed in MPa.

[0090] ξ1 and ξ2 are the tectonic stress coefficients in two horizontal directions;

[0091] The formation collapse pressure profile calculation model mainly includes Formation Collapse Pressure Model I and Formation Collapse Pressure Practical Calculation Model II.

[0092] The formation collapse pressure model I prediction mode:

[0093]

[0094] In the formula:

[0095] ρ bt The equivalent mud density for formation shear collapse pressure is expressed in g / cm³.

[0096] η is the nonlinear correction coefficient for the wellbore rock; for general mudstone, η≤1.

[0097] σ H σ is the maximum horizontal principal stress; h The minimum principal stress is horizontal; P p Formation pore pressure;

[0098] H is the depth of the reflected layer; K is the correction factor, K = 0.5 to 0.80;

[0099] Prediction model of Practical Calculation Model II for Formation Collapse Pressure:

[0100]

[0101] Where: K t =3γ-β, is called the regional tectonic stress influence coefficient, which is a constant within the same block.

[0102] This disclosure provides a big data-based wellbore instability analysis method. Based on big data analysis and deep learning for wellbore instability prediction, and combining existing wellbore instability theories and experimental research results, it identifies the main and related factors of wellbore instability in the studied area. Based on data classification, a big data model is established, and the accuracy of analysis and prediction is improved through big data learning. Through the application of the big data model, deep learning technology is used to evaluate wellbore instability in the Changqing area, ultimately forming a big data-based wellbore instability evaluation and prevention technology for the Changqing area. The big data-based wellbore instability analysis tool mainly provides a safe density window for preventing wellbore instability, control analysis of wellbore diameter changes, optimization of drilling fluid performance and treatment agent dosage, adjacent well data analysis, ECD analysis, and drilling time effect analysis, thereby reducing the impact of hydration expansion of the target formation rock and preventing wellbore instability.

[0103] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0104] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0105] Figure 1 This is a flowchart of a wellbore instability analysis method based on big data provided in an embodiment of this disclosure;

[0106] Figure 2 This is a flowchart of a wellbore instability prediction modeling method based on big data, provided in an embodiment of this disclosure.

[0107] Figure 3 This is a schematic diagram of the safety density window and density early warning window structure of a wellbore instability analysis method based on big data provided in this disclosure embodiment;

[0108] Figure 4This is a schematic diagram of the three-pressure profile calculation model structure of a wellbore instability analysis method based on big data provided in this embodiment of the disclosure;

[0109] Figure 5 This is a schematic diagram of the geostress parameter model structure of a wellbore instability analysis method based on big data provided in this embodiment of the disclosure. Detailed Implementation

[0110] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0111] Example 1

[0112] This disclosure provides a wellbore instability analysis method based on big data, such as... Figure 1 As shown, the fault diagnosis method includes the following steps:

[0113] 101. Establish a wellbore instability analysis database and build a data interface based on the data collected by the data acquisition platform; preprocess the wellbore instability data analysis scan by traversing all drilling data from previous years, build a wellbore elastic model, obtain the stress state on the wellbore, and train the wellbore instability data from previous years.

[0114] 102. Training with historical wellbore instability data involves analyzing well diameter changes in the block formation and all construction schemes within the block. Drilling fluid density is obtained based on stress components under different conditions and substituted into the criteria for calculation to obtain shear stress on the wellbore shear surface and determine the wellbore stability.

[0115] 103. Based on the data from adjacent wells, analyze all construction plans within the block to obtain the safe density window and density warning window; based on the normalized historical logging data within the block, train and model using a BP neural network, and establish a formation pressure model through the training; this formation pressure model is a trained wellbore instability prediction neural network model for predicting wellbore instability.

[0116] In one embodiment, the data acquisition platform in step 101 collects data, which means collecting well site data and on-site engineering data, and preparing data collection before building machine learning for training; and building a data interface for extracting, cleaning and converting well caliper data, accident complexity information, gamma value, well foundation information, logging data, drilling parameters, drilling fluid performance and electrical logging information.

[0117] In one embodiment, the preprocessing of wellbore instability data analysis and scanning by traversing all historical drilling data in step 101 refers to acquiring parameters affecting wellbore instability, blocks, and formations prone to wellbore instability from all historical drilling data to construct a wellbore elasticity model. The formula for this wellbore elasticity model is as follows:

[0118] Radial stress component σ in cylindrical coordinates r for:

[0119]

[0120] Tangential stress component σ in cylindrical coordinates θ for:

[0121]

[0122] Axial stress component σ in cylindrical coordinate system z for:

[0123]

[0124] Shear stress component τ in cylindrical coordinates rθ for:

[0125]

[0126] Shear stress component τ in cylindrical coordinates θz for:

[0127]

[0128] Shear stress component τ in cylindrical coordinates rz for:

[0129]

[0130] In the formula: P P δ is the pore pressure; μ is Poisson's ratio; δ is the wellbore permeability coefficient.

[0131] θ is the wellbore angle (relative to the x-axis); φ1 is the porosity; α1 is the Biot coefficient; P W α is the fluid pressure inside the wellbore; α is the well inclination angle. R is the azimuth angle; R is the wellbore radius; r is the radial coordinate.

[0132] When the well wall is impermeable, the well wall permeability coefficient δ = 0;

[0133] When the well wall is permeable, the well wall permeability coefficient δ = 1;

[0134] Let r = R; substitute into equations (1), (2), and (3) to obtain the stress state on the well wall; the formula for the stress state on the well wall is:

[0135] σ r =Ρ W -δφ1(P W -P P (7)

[0136]

[0137]

[0138] τ rθ =0 (10)

[0139]

[0140] τ rz =0 (12)

[0141] τ rθ τ θz τ rz These are the shear stress components in cylindrical coordinates.

[0142] In one embodiment, step 102, training with historical wellbore instability data, involves analyzing wellbore diameter variations in the block formation and all construction schemes within the block. Drilling fluid density is obtained based on stress components under different conditions, and these are substituted into criteria for calculation to determine the corresponding shear stress on the wellbore shear surface, thus assessing wellbore stability. This refers to calculating the radial stress component σ under different conditions. r tangential stress component σ θ axial stress component σ z The drilling fluid density was obtained and substituted into the Coulomb-Mohr criterion for calculation to obtain the corresponding shear stress on the wellbore shear surface;

[0143] The Coulomb-Moore criterion states that when the shear stress on the shear surface is greater than or equal to the rock's inherent shear strength S0 plus the frictional resistance fσ acting on the shear surface, the wellbore will experience shear failure; otherwise, the wellbore will be a stable structure.

[0144] |τ|≥fσ+S0 (13)

[0145] Where τ is the shear stress on the shear plane; S0 is the inherent shear strength of the rock; and f is the internal friction coefficient of the rock.

[0146] In one embodiment, the radial stress component σ r tangential stress component σ θ axial stress component σ zThe expression was substituted into the Coulomb-Mohr criterion under different conditions to obtain the corresponding critical drilling fluid density P in the wellbore direction. w ;

[0147] a. When σ θ ≥σ z ≥σ r hour,

[0148] b. When σ z ≥σ θ ≥σ r hour,

[0149] c. When σ z ≥σ r ≥σ θ hour,

[0150] d. When σ r ≥σ θ ≥σ z hour,

[0151] e. When σ r ≥σ z ≥σ θ At that time, P w ≥σ v tan 2 β-α1P f (tan 2 β-1)+C0

[0152] f, when σ θ ≥σ r ≥σ z At that time, P w ≤2σ h -σ v tan 2 β+α1P f (tan 2 β-1)-C0

[0153] Where, σ r The radial stress component in cylindrical coordinates; σ θ σ represents the tangential stress component in cylindrical coordinates. z This represents the axial stress components in cylindrical coordinates.

[0154] In one embodiment, step 102, training with historical wellbore instability data, involves analyzing wellbore diameter variations in the block formation and all construction schemes within the block. Drilling fluid density is obtained based on stress components under different conditions and substituted into the criteria for calculation. This refers to calculating the radial stress component σ... rtangential stress component σ θ axial stress component σ z The expression is substituted into the modified Lade criterion (Lade yield) under different conditions to solve for the corresponding critical drilling fluid density P in the wellbore direction. w ;

[0155] The Lade criterion has the following expression:

[0156]

[0157] Among them, I1=σ1+σ2+σ3; I3=σ1σ2σ3; P a Atmospheric pressure; m, η1 are material constants;

[0158] The modified Lade criterion has the following expression:

[0159] (I″1) 3 / I″3=27+η (15)

[0160] in,

[0161] I″1=(σ r +S1-P P )+(σ θ +S1-P P )+(σ Z +S1-P P )

[0162]

[0163] Shear stress components in cylindrical coordinates

[0164] S1 and η are material constants; S1 = S0 / tanφ; η = 4tanφ 2 φ(9-7sinφ) / (1-sinφ)

[0165] P P φ is the pore pressure; φ is the internal friction angle.

[0166] S0 represents the inherent shear strength or internal friction of the rock.

[0167] q represents shear strength; q = (1 / 2)[(σ1 - σ2)] 2 +(σ2-σ3) 2 +(σ1-σ3) 2 ] 1 / 2

[0168] When the wellbore is impermeable, the wellbore permeability coefficient δ = 0. Substituting equation (15) into the wellbore elastic model, the critical drilling fluid density P in the wellbore direction is obtained. W The expression is:

[0169]

[0170] In the formula:

[0171] A = σ Z +S1-P P ;

[0172]

[0173]

[0174] D=(σ θn +σ Z +3S1-3P P ) 3 / (27+η);

[0175]

[0176]

[0177]

[0178] In one embodiment, the critical drilling fluid density value in the wellbore direction is predicted through big data learning analysis. This prediction analysis generates a three-pressure profile for the block, which includes a fracture pressure prediction assessment, a collapse pressure prediction assessment, and a pore pressure prediction assessment. At the same time, data from adjacent wells are analyzed to obtain safe drilling parameters and generate a recommended construction plan.

[0179] In one embodiment, analyzing data from adjacent wells to obtain safe drilling parameters refers to analyzing data from adjacent wells in each formation to obtain safe density windows and density warning windows. The data analysis of these adjacent wells includes the drilling fluid system and performance parameters of the formation, optimization of treatment agent dosage and configuration, and analysis of how to reduce well diameter changes.

[0180] The data for adjacent wells includes performance range, treatment agent dosage, mechanical drilling rate, well foundation information, drilling parameters, accident complexity, well enlargement rate, gamma value, drilling cycle, cost, and electrical logging information.

[0181] The generated suggested construction plan refers to the construction plan selected for this block according to different dimensions. The selected dimensions include wellbore enlargement rate, drilling cycle, mechanical drilling rate, and drilling cost. The suggested construction plan indicators include priority density range, standard density range, electrical logging resistance, and gamma value.

[0182] In one embodiment, before analyzing the data of adjacent wells in each formation and obtaining the safe density window and density warning window, the data of the adjacent wells is selected, and data cleaning is used to process the defects of different adjacent well data. Outliers in the data of the adjacent wells are identified and processed using a clustering algorithm. The data is then processed by pattern classification and normalization and used as input for training a BP neural network. The BP neural network is modeled using a training function. The number of nodes in the input layer of the BP neural network is set to the number of fault features. The number of hidden layer nodes and the learning rate parameter are set using empirical values ​​and a trial-and-error method.

[0183] In one embodiment, the construction plan, safety density window, and density warning window for this block are modeled based on a BP neural network after normalization of historical logging data within the block. The formation pressure model is established through this training, and the formation fracture pressure profile model and formation collapse pressure profile model are established through training based on the BP neural network.

[0184] Among them, the formation fracture pressure profile model refers to the formation fracture pressure based on the initiation of fractures or the reopening of existing fractures in the exposed formation under external forces. This formation fracture is caused by excessively high drilling fluid density, resulting in the effective tangential stress on the rock exceeding the tensile strength of the rock. Its calculation models mainly include:

[0185] Practical Rupture Pressure Prediction Model

[0186]

[0187] Prediction models for two horizontal principal geostresses:

[0188]

[0189]

[0190] In the formula:

[0191] P f Formation fracture pressure, MPa; E s The static Young's modulus of the rock is given in MPa.

[0192] μ s P represents the static Poisson's ratio of the rock, which is dimensionless. p Formation pore pressure, MPa;

[0193] α is the effective stress coefficient, which is dimensionless; σ H The maximum horizontal principal stress is expressed in MPa.

[0194] σ h The minimum principal stress is σ, MPa;v The pressure of the overlying strata is MPa;

[0195] S t The tensile strength of the rock is expressed in MPa.

[0196] ξ1 and ξ2 are the tectonic stress coefficients in two horizontal directions;

[0197] The formation collapse pressure profile calculation model mainly includes Formation Collapse Pressure Model I and Formation Collapse Pressure Practical Calculation Model II.

[0198] Formation collapse pressure model I prediction mode:

[0199]

[0200] In the formula:

[0201] ρ bt The equivalent mud density for formation shear collapse pressure is expressed in g / cm³.

[0202] η is the nonlinear correction coefficient for the wellbore rock; for general mudstone, η≤1.

[0203] σ H σ is the maximum horizontal principal stress; h The minimum principal stress is horizontal; P p Formation pore pressure;

[0204] H is the depth of the reflected layer; K is the correction factor, K = 0.5 to 0.80;

[0205] Prediction model of Practical Calculation Model II for Formation Collapse Pressure:

[0206]

[0207] Where: K t =3γ-β, is called the regional tectonic stress influence coefficient, which is a constant within the same block.

[0208] This disclosure provides a big data-based wellbore instability analysis method. The wellbore instability prediction method, based on big data analysis and deep learning, mainly combines existing wellbore instability theories and experimental research results to identify the main and related factors of wellbore instability in the studied area. Based on data classification, a big data model is established, and the accuracy of analysis and prediction is improved through big data learning. By applying the big data model, deep learning technology is used to evaluate wellbore instability in the Changqing area, ultimately forming a big data-based wellbore instability evaluation and prevention technology for the Changqing area. The big data-based wellbore instability analysis tool mainly provides a safe density window for preventing wellbore instability, control analysis of wellbore diameter changes, optimization of drilling fluid performance and treatment agent dosage, adjacent well data analysis, ECD analysis, and drilling time effect analysis, etc., to reduce the impact of hydration expansion of the target formation rock and prevent wellbore instability.

[0209] Based on the above Figure 1 The corresponding embodiments describe a wellbore instability analysis method based on big data, and the following are specific embodiments disclosed in this invention.

[0210] Example 2

[0211] This invention provides a wellbore instability analysis method based on big data, such as... Figure 2 As shown, the steps of this method are as follows:

[0212] Step S1 establishes a wellbore instability analysis database, which includes all data related to wellbore instability from the integrated database and the drilling fluid information technology integrated platform database.

[0213] Specifically, the integrated storage system enables wireless networking among multiple disciplines such as well testing, acid fracturing, and testing, making the transmission and storage of real-time data faster, more accurate, and more centralized. In addition, the Sichuan-Chongqing integrated storage platform can perform comprehensive data acquisition and management, professional data analysis and application, and production dynamic tracking across the entire wellbore professional chain, and has the capability to process massive amounts of data.

[0214] The integrated drilling fluid information technology platform database covers related specialties such as drilling, logging, well logging, and downhole operations. Through the platform, it has completed various applications including unique data acquisition, data management, data publishing, and data analysis, laying the foundation for information technology construction. In the overall system architecture design, the data acquisition platform uses an integrated well site data collector and data acquisition system to collect on-site engineering data; the data transmission platform uses well site networking and the Sichuan-Chongqing integrated dedicated network for data transmission; the data storage platform uses standard data formats to store data in the integrated database; and the comprehensive application platform conducts remote data monitoring, engineering early warning, and drilling and merging decision-making applications at the rear base.

[0215] The wellbore instability analysis database in this invention is established by integrating all data related to wellbore instability into an integrated database and a drilling fluid information technology integrated platform database. This enables the wellbore instability analysis database to combine existing wellbore instability theories and experimental research results to identify the main and related factors of wellbore instability in the studied area. Based on data classification, a big data model is established, and the accuracy of analysis and prediction is improved through big data learning.

[0216] In one embodiment: Step S2 involves data collection and preparation, which involves data interface processing based on the data in the established database. The data interface includes an extraction, cleaning, and conversion interface for well caliper data, accident complexity information, gamma value, well foundation information, logging data, drilling parameters, drilling fluid performance, and electrical logging information.

[0217] There are many reasons for wellbore instability, mainly divided into two aspects: natural and man-made.

[0218] Natural factors include: geological structure type and in-situ stress, lithology and shape of strata, type of clay minerals, presence and dip angle of weak planes, cementation of bedding planes, stratum strength, development of fractures and joints, porosity, permeability and fluid pressure in pores, etc.

[0219] Human factors include: drilling fluid properties (water loss, viscosity, rheology, density), the composition of the drilling fluid and the strength of chemical interaction with shale (hydration, swelling), the depth and extent of the drilling fluid invasion zone around the well, the time of wellbore exposure, the annular return velocity of the drilling fluid, the erosion effect on the well wall, the circulating dynamic pressure and the fluctuation pressure during tripping in and out of the well, the shape of the wellbore trajectory, and the friction and collision between the drill string and the well wall.

[0220] Data related to influencing factors and their acquisition

[0221] (1) Formation rock mechanical parameters

[0222] This mainly includes rock elastic modulus, compressive strength, cohesion, friction angle, Poisson's ratio, etc. Data acquisition methods: Drawing on existing research findings, or obtaining data through sonic logging combined with actual rock sample experiments; in practical research, applying existing mature theories and combining them with the actual conditions of the study block, establishing methods for determining formation rock mechanical parameters using sonic logging data, obtaining formation rock mechanical parameters, and, where core samples are available, obtaining accurate rock mechanical parameter data through experimental testing.

[0223] (2) Geostress parameters

[0224] Data acquisition methods: Drawing on existing research findings, or obtaining data through sonic logging and other methods combined with actual rock sample experiments; establishing a method for determining geostress parameters using sonic logging data to obtain more accurate geostress parameters for the study block, providing a foundation for subsequent quantification of influencing factors.

[0225] (3) Pore pressure

[0226] Data acquisition methods: draw on existing research results, or obtain data through sonic logging and other methods combined with actual drilling tests; establish a method for determining pore pressure parameters using sonic logging data, obtain more accurate pore pressure parameters for the research block, and provide a basis for subsequent quantification of influencing factors.

[0227] (4) Study the rock and mineral composition of the target strata

[0228] Data acquisition methods: drawing on existing research findings or obtaining data through actual rock sample experiments.

[0229] (5) Study the hydration and swelling properties of rocks in the target formation

[0230] Data acquisition methods: draw on existing research results or obtain data through actual rock samples and experimental testing of in-use drilling fluid systems, and combine big data analysis theory to reasonably quantify influencing factors.

[0231] (6) Mechanical analysis results of the influence of wellbore inclination and azimuth on wellbore stability

[0232] Data acquisition methods: draw on existing research results or obtain wellbore stress state through classical mechanics analysis, and rationally determine the quantitative influencing factors.

[0233] (7) Drilling fluid system and performance; Data acquisition method: draw on existing research results or obtain data through drilling database parameter analysis, and reasonably quantify the influencing factors.

[0234] (8) Drilling hydraulic parameters; Data acquisition method: refer to existing research results or obtain them through drilling database parameter analysis. The study needs to comprehensively consider wellbore size, drilling fluid performance, etc., to provide a basis for determining the actual drilling density window.

[0235] (9) The time effect of wellbore stability under the drilling fluid system; Data acquisition method: refer to existing research results or obtain data through drilling database parameter analysis. The study focuses on referencing existing research results and reasonably quantifies the influencing factors.

[0236] (10) Drilling fluid density window; Data acquisition method: Determine the formation collapse pressure by referring to existing research results, or determine the formation collapse pressure through theoretical analysis, and then determine the drilling fluid density window;

[0237] In one embodiment, step S3 involves data preprocessing, analyzing and scanning drilling data on wellbore instability. This includes data on blocks and formations prone to wellbore instability, specific well types, well categories, and situations where wellbore instability frequently occurs in project departments, comparative analysis of drilling data from adjacent wells in the same block, the time effect of wellbore stability under the drilling fluid system in use, and the drilling fluid density window, to form an early warning scheme for frequent wellbore instability in the same block.

[0238] Before drilling begins, compressive stress already exists in the rock strata. Except in geologically complex areas, geostress can be categorized into vertical stress or overburden pressure σ. V and two horizontal stresses: σ H (maximum horizontal stress) and σ h (Minimum horizontal stress). Generally speaking, the two horizontal stresses are not equal. After drilling begins, the pressure of the drilling mud column replaces the support provided to the wellbore by the drilled rock formation, causing a redistribution of stress around the wellbore. This redistributed stress can be called σ. θ Circumferential stress along the wellbore), σ r Radial stress along the wellbore), σ z Stress along the axis of the parallel wellbore. In deviated wells, it also includes the shear stress component τ. θz .

[0239] Wellbore stress can be easily calculated, primarily relying on stress-strain characteristics chosen to simulate the rock formation's response to loads. The most common assumed characteristics are formation homogeneity, isotropy, and linear elasticity. Based on these, stress can be described by some simple equations; other characteristic models require numerical solutions. Linear elastic analysis is simple to apply and therefore the most common method. Other complex models struggle to obtain complete input parameters, which are difficult to accurately determine in the field. Using the linear elastic method, the stress state on the wellbore is generally at a critical point. Sometimes, in underbalanced drilling and considering the viscous effects of pore fluid migration, most critical points occur within the rock rather than on the wellbore. The three principal stresses are described as follows:

[0240] Maximum principal stress σ1:

[0241] Aziσ1 is the angle between σ1 and the due north direction;

[0242] Incσ1 is the angle between the downward tilt direction of σ1 and the vertical direction;

[0243] Intermediate principal stress σ2:

[0244] Tiltσ2 is the angle between σ2 and the vertical direction;

[0245] Minimum principal stress σ3;

[0246] ① Wellbore elastic model:

[0247]

[0248]

[0249]

[0250]

[0251]

[0252]

[0253] In the formula:

[0254] σ r The radial stress component is in cylindrical coordinates.

[0255] σ θ For the tangential stress components in cylindrical coordinates;

[0256] σ z For the axial stress components in cylindrical coordinates;

[0257] τ rθ The shear stress components are in cylindrical coordinates.

[0258] τ θz The shear stress components are in cylindrical coordinates.

[0259] τ rz The shear stress components are in cylindrical coordinates.

[0260] P P Pore ​​pressure;

[0261] μ is Poisson's ratio;

[0262] δ is the wellbore permeability coefficient; when the wellbore is impermeable, δ = 0; when the wellbore is permeable, δ = 1.

[0263] θ is the well-circumference angle (relative to the x-axis);

[0264] φ1 represents porosity;

[0265] α1 is the Biot coefficient;

[0266] P W The pressure of the liquid inside the wellbore;

[0267] α is the well inclination angle;

[0268] It is the azimuth angle;

[0269] R is the radius of the wellbore;

[0270] r is the radial coordinate;

[0271] ②The stress state on the well wall is as follows:

[0272] Let r = R:

[0273] σ r =Ρ W -δφ1(P W -P P )

[0274]

[0275]

[0276] τ rθ =0

[0277]

[0278] τ rz =0

[0279] Specifically, an integrated intelligent operation and maintenance platform for wellbore instability analysis is built through a set data interface to undertake the entire operation and maintenance support work and comprehensively monitor data that may lead to drilling instability. Through network topology diagrams, automated operation and maintenance, and customized functions, and through data analysis, it provides safe density windows for preventing wellbore instability, control analysis of well diameter changes, optimization of drilling fluid performance and treatment agent dosage, adjacent well data analysis, ECD analysis, and drilling time effect analysis, etc., and performs intelligent task scheduling to reduce the impact of hydration expansion of target formation rocks and prevent wellbore instability.

[0280] In this embodiment: Step S4 is based on BP neural network training to construct a wellbore elasticity model. Wellbore diameter variation analysis suggests construction plans for wells with relatively stable wellbores and provides early warning plans for wells with relatively unstable wellbores. The plans include the block, formation, performance range, and treatment agent dosage. This mainly combines existing wellbore instability theories and experimental research results to identify the main and related factors of wellbore instability in the studied block. Based on data classification, a big data model is established, and the accuracy of analysis and prediction is improved through big data learning. The study optimizes the drilling fluid system and performance parameters of the block's formation, treatment agent dosage and configuration, analyzes how to reduce wellbore diameter variation, identifies reasonable safe density windows and density warning windows, provides data support for increasing the collapse cycle to reduce drilling time effects, and analyzes data from adjacent wells to find safe drilling parameters, etc. Figure 5 As shown, the geostress parameter model focuses on key strata within the studied area. A neural network model is established by combining existing wellbore instability theories and experimental research results.

[0281] Based on a review of domestic and international literature and taking into account various factors affecting the wellbore, a wellbore instability prediction model based on big data analysis and deep learning was established.

[0282] The following table provides formulas for calculating drilling fluid density under different conditions:

[0283]

[0284] This situation is obtained under certain conditions. In practice, σ should be... r σ θ σ z Substituting the expression into the Coulomb-Moore criterion, we obtain the corresponding conditions.

[0285] The Lade criterion and the Coulomb-Moore criterion have similar approaches to finding the corresponding solutions.

[0286] The Lade criterion has the following expression:

[0287]

[0288] Among them, I1=σ1+σ2+σ3; I3=σ1σ2σ3; P a Atmospheric pressure; m, η1 are material constants;

[0289] The modified Lade criterion has the following expression:

[0290] (I″1) 3 / I″3=27+η (15)

[0291] in,

[0292] I″1=(σ r +S1-P P )+(σ θ +S1-P P )+(σ Z +S1-P P )

[0293]

[0294] Shear stress components in cylindrical coordinates

[0295] S1 and η are material constants; S1 = S0 / tanφ; η = 4tanφ 2 φ(9-7sinφ) / (1-sinφ)

[0296] P P φ is the pore pressure; φ is the internal friction angle.

[0297] S0 represents the inherent shear strength or internal friction of the rock.

[0298] q represents shear strength; q = (1 / 2)[(σ1 - σ2)] 2 +(σ2-σ3) 2 +(σ1-σ3) 2 ] 1 / 2

[0299] Substituting the stress calculation expression under the pore elastic model into the above formula, the corresponding P can be obtained. W .

[0300] When the well wall is impermeable, δ = 0, and at this time:

[0301]

[0302] In the formula:

[0303] A = σ Z +S1-P P

[0304]

[0305]

[0306] D=(σ θn +σ Z +3S1-3P P ) 3 / (27+η)

[0307]

[0308]

[0309]

[0310] Specifically, the early warning scheme of this invention optimizes the drilling fluid system and performance parameters of the preferred formation, the dosage and configuration of the treatment agent, analyzes how to reduce well diameter changes, and identifies a reasonable safe density window and density early warning window, such as... Figure 3 As shown, this study provides data to reduce the drilling time effect by increasing the collapse cycle, and supports wellbore instability prediction based on big data analysis and deep learning. It mainly combines existing wellbore instability theories and experimental research results to find the main and related factors of wellbore instability in the studied block. Based on data classification, a big data model is established, and the accuracy of analysis and prediction is improved through big data learning. At the same time, data from adjacent wells are analyzed to find safe drilling parameters, and finally, a suggested construction plan is generated.

[0311] In this embodiment: Step S5 suggests a construction plan, wherein the construction plan analysis optimizes the construction plan for this block according to different dimensions. The optimization dimensions include wellbore enlargement rate, drilling cycle, mechanical drilling rate, and drilling cost. The indicators include preferred density range, standard density range, electrical logging resistance, and gamma value.

[0312] Regarding wellbore stability, wellbore instability is essentially a mechanical failure problem. Because numerous factors are involved, the main factors can be summarized into five categories: rock strength, geostress, hydration expansion (including time effects), drilling fluid density window, and drilling engineering factors.

[0313] The effects of hydration swelling involve pore pressure, electrolyte chemistry, and physicochemical properties in the drilling fluid, and many of these effects are interrelated, making quantification difficult in current research. Further research needs to draw upon existing mature theories and combine them with the practical engineering aspects of drilling in the block to provide reasonable quantitative data.

[0314] In practical research, it is difficult to reasonably determine the weights using direct quantitative data. It is necessary to reasonably determine the main controlling influencing factors and quantify them according to the five major data categories, theoretical methods, existing mature research results at home and abroad, and the actual situation of Changqing drilling project.

[0315] Step S6 establishes a wellbore prediction model based on a BP neural network. This involves analyzing the performance range of adjacent wells, treatment agent dosage, mechanical drilling rate, ECD value, drilling time, accident complexity, wellbore enlargement rate, gamma ray value, drilling cycle, cost, and electrical logging information for each formation. The research focuses on optimizing the drilling fluid system and performance parameters for the block's formations, optimizing treatment agent dosage and configuration, analyzing how to reduce wellbore variation, identifying reasonable safe density windows and density warning windows, providing data support for increasing the collapse cycle to mitigate the drilling time effect, and analyzing data from adjacent wells to find safe drilling parameters.

[0316] Specifically, the selection of acquired data is crucial because not all data can be fitted using the same model. If the data used in model training differs significantly in value, distribution, etc., the model's fitting performance will inevitably deteriorate. This problem is particularly evident in drilling data. Typically, during oil drilling, the geographical location and significantly different environments between multiple wells directly lead to substantial differences in drilling data. If data from multiple wells are used to train the same model, the resulting model's fitting performance is likely to be poor. Furthermore, due to the influence of drilling depth, drilling data can vary considerably at different drilling stages within the same well. Therefore, selecting appropriate data is essential before processing the data.

[0317] 1) Data cleaning is typically performed to address data quality issues. Raw data obtained from real-world applications inevitably contains problems such as missing data or invalid features. Depending on the specific circumstances of drilling data, this step primarily addresses different data defects.

[0318] 2) Outlier handling: This involves detecting and processing outliers in the data. Outliers are typically samples that are far from other data points and significantly deviate from them. In terms of distribution, normal samples are usually densely distributed in a small area, while outliers are usually far from the normal sample distribution and sparsely distributed. Outliers may indicate data anomalies, errors in the sampling process, or that a certain theory is not applicable in a specific context. Therefore, clustering algorithms are needed to identify, identify, and process them accordingly.

[0319] 3) After outlier processing is completed and before the data is input into the model for training, the data needs to be segmented and normalized.

[0320] 4) We will try various machine learning algorithms, such as neural networks, to model and compare them to select the prediction model with the best performance.

[0321] After establishing a wellbore instability prediction model based on big data analysis and deep learning, a large amount of data is needed for practical application research to further improve the prediction model and continuously optimize it. Comprehensive and accurate data acquisition is crucial because it involves a wide range of data, a large volume of data, and requires a lot of important basic data (such as core samples, cuttings, and logging data). If the target formation is not a reservoir, problems may arise such as difficulty in obtaining some necessary basic data or difficulty in obtaining accurate data.

[0322] Step S7 establishes a formation pressure model. Through a review of domestic and international literature, and by comprehensively considering various factors affecting the wellbore, a wellbore instability prediction model based on big data analysis and deep learning training of a BP neural network wellbore prediction model is established.

[0323] Specifically, the research will focus on key stratigraphic levels within the regional study area. A neural network model will be established by combining existing wellbore instability theories and experimental findings.

[0324] 1. Formation fracture pressure profile model

[0325] Formation fracturing pressure refers to the pressure exerted on exposed formations in a wellbore by external forces, causing fracturing or reopening of existing fractures. Formation fracturing occurs when the drilling fluid density is too high, causing the effective tangential stress on the rock to exceed its tensile strength. Its calculation models mainly include:

[0326] 1.1 Practical Rupture Pressure Prediction Model I

[0327] Practical Rupture Pressure Prediction Model

[0328]

[0329] Prediction models for two horizontal principal geostresses:

[0330]

[0331]

[0332] In the formula:

[0333] P f The formation fracture pressure is expressed in MPa.

[0334] E s The static Young's modulus of the rock is given in MPa.

[0335] μ s The static Poisson's ratio for the rock is dimensionless.

[0336] P p Formation pore pressure, MPa;

[0337] α is the effective stress coefficient, which is dimensionless;

[0338] σ H The maximum horizontal principal stress is expressed in MPa.

[0339] σ h The minimum principal stress is in MPa.

[0340] σ v The pressure of the overlying strata is MPa;

[0341] S t denoted as the tensile strength of the rock, MPa.

[0342] 2. Calculation model for formation collapse pressure profile

[0343] Wellbore collapse is closely related to the stress state, strength characteristics, and drilling fluid density of the surrounding rock. Currently, there are two main models for calculating formation collapse pressure, namely, Practical Calculation Model I and Practical Calculation Model II.

[0344] 2.1 Practical Collapse Pressure Calculation Model I

[0345] Prediction pattern:

[0346]

[0347] In the formula:

[0348] ρ btThe equivalent mud density for formation shear collapse pressure is expressed in g / cm³.

[0349] η is the nonlinear correction coefficient for the wellbore rock; for general mudstone, η = 0.95.

[0350] The meanings and determination methods of other parameters are the same as before.

[0351] 2.1 Practical Collapse Pressure Calculation Model II

[0352] Prediction pattern:

[0353]

[0354] In the formula:

[0355] K t =3γ-β, which is called the regional tectonic stress influence coefficient. It is a constant within the same block and has no dimension.

[0356] The meanings and determination methods of other parameters are the same as before.

[0357] Using the model above, we can perform computer programming using VB6.0. Using the programmed code, we can calculate the formation elastic parameters and then use the following criteria to determine the wellbore stability.

[0358] Predict the generation of block three pressure profiles, such as Figure 4 As shown, this includes prediction and assessment of fracture pressure, collapse pressure, and pore pressure. Regarding wellbore stability, wellbore instability is essentially a mechanical failure problem. Because numerous influencing factors are involved, the main factors can be summarized into five categories: rock strength influence, in-situ stress influence, hydration expansion influence (including time effects), wellbore drilling fluid density window influence, and drilling engineering factors.

[0359] The impact of hydration expansion involves pore pressure, the chemistry and physicochemical properties of electrolytes in the drilling fluid, and many of these effects are interrelated, making quantification difficult in current research. This study needs to draw upon existing mature theories and combine them with the engineering practice of drilling in the Changqing block to provide reasonable quantitative data.

[0360] The S8 step predicts wellbore instability using a computer program. It mainly combines existing wellbore instability theories and experimental research results to identify the main and related factors of wellbore instability in the studied area. Based on data classification, a big data model is established to improve the accuracy of analysis and prediction through big data learning.

[0361] Specifically, a neural network model is established by combining existing wellbore instability theories and experimental research results. Wellbore instability prediction based on big data analysis and machine learning mainly involves combining existing wellbore instability theories and experimental research results to find the main and related factors of wellbore instability in the studied block, establishing a big data model based on data classification, and improving the accuracy of analysis and prediction through big data learning.

[0362] Formations are relatively stable sedimentary intervals formed over a period of time in a relatively unchanging sedimentary environment. Therefore, theoretically, well logging curves should consist of stepped segments. In reality, due to the contributions of adjacent layers, well logging curves appear as smooth curves. Well logging analysts stratify well logging curves based on their morphological characteristics and amplitude. This involves abstracting the symbolic features of the well logging curves and then using empirical knowledge and logical rules to perform logical reasoning to arrive at a conclusion. The principle of the artificial intelligence automatic stratification method is to simulate human thought processes using computers. It expresses the knowledge and rules used in manual comparison work using computer language and describes the logical thought process of human stratification, thus using computers to simulate human reasoning.

[0363] Because the big data analysis of wellbore stability covers a wide range of areas, including geostress distribution, pore pressure prediction, formation rock mechanical parameters, formation rock clay mineral types and contents, formation rock hydration and expansion properties, drilling fluid system and properties, drilling fluid circulation and downhole pressure changes, wellbore inclination and azimuth, and the time effect of hydration and expansion of target formation rocks, a large amount of data is required and the data acquisition sources involve multiple departments.

[0364] The establishment of an integrated wellbore instability analysis platform can correspond to block formation lithology analysis, wellbore physicochemical analysis, drilling fluid column pressure and ECD analysis, and enable the weight allocation of various influencing factors of wellbore instability in the prevention of complex accidents in each block.

[0365] The above technical solution utilizes a trained neural network to accurately analyze and process historical big data from a large number of accident handling cases collected annually. It establishes an artificial intelligence program based on machine learning technology to predict wellbore instability. S1 and S2 involve data collection and preparation before training the machine learning system. S3 involves preprocessing the data from previous years for training. S4 and S5 involve finding early warnings and solutions through training, ensuring the neural network training results closely match actual conditions and improving problem-solving efficiency. S6 and S7 involve finding or selecting suitable solution models for various complex situations, based on a wellbore instability prediction model trained using a BP neural network. S8 then uses this trained neural network wellbore instability prediction model to predict wellbore instability.

[0366] This disclosure provides a big data-based wellbore instability analysis method. The wellbore instability prediction method, based on big data analysis and deep learning, mainly combines existing wellbore instability theories and experimental research results to identify the main and related factors of wellbore instability in the studied area. Based on data classification, a big data model is established, and the accuracy of analysis and prediction is improved through big data learning. By applying the big data model, deep learning technology is used to evaluate wellbore instability in the Changqing area, ultimately forming a big data-based wellbore instability evaluation and prevention technology for the Changqing area. The big data-based wellbore instability analysis tool mainly provides a safe density window for preventing wellbore instability, control analysis of wellbore diameter changes, optimization of drilling fluid performance and treatment agent dosage, adjacent well data analysis, ECD analysis, and drilling time effect analysis, etc., to reduce the impact of hydration expansion of the target formation rock and prevent wellbore instability.

[0367] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for analyzing wellbore instability based on big data, characterized in that, The method comprises: Step 101, establishing a wellbore instability analysis database and constructing a data interface according to data collected by a data collection platform; pre-processing wellbore instability data analysis scanning by traversing all historical drilling data, constructing a wellbore elasticity model, and obtaining stress states on the wellbore, Step 102, substituting stress components under different conditions into the criterion for operation respectively to obtain a critical drilling fluid density; at the same time, normalizing historical logging data and adjacent well data in the block as inputs of a wellbore instability prediction model trained based on a BP neural network; Expressions for the radial stress component , tangential stress component , and axial stress component are substituted into the Coulomb-Mohr criterion under different conditions to obtain the corresponding critical drilling fluid density for the wellbore direction; a. when time, b. when time, c. when time, d. when time, e. when time, f. when time, wherein, is the radial stress component in cylindrical coordinates; is the tangential stress component in cylindrical coordinates; is the axial stress component in cylindrical coordinates, is the formation fracture pressure, MPa; The expressions of the radial stress component , the tangential stress component , and the axial stress component are substituted into the modified Lade criterion under different conditions to obtain the corresponding critical drilling fluid density of the wellbore direction ; The Lade criterion has the following expression: (14) wherein ; ; is atmospheric pressure; is a material constant; The modified Lade criterion has the following expression: (15) wherein, are the shear stress components in cylindrical coordinates , is a material constant; ; Pp is the pore pressure; φ is the internal friction angle; is the intrinsic shear strength or internal friction of the rock; shear strength; When the well wall is impermeable, the well wall permeability coefficient The critical drilling fluid density in the wellbore direction is calculated by substituting the formula (15) into the well wall elastic model The expression is: （16） In the formula: ； ； ； ； ； ； ； Step 103, obtaining a safe density window and a density early warning window according to all construction schemes in the block analyzed based on adjacent well data; training and modeling based on a BP neural network to establish a formation pressure model; the formation pressure model is a trained wellbore instability prediction neural network model for wellbore instability prediction.

2. The method of claim 1, wherein, The data collected by the data collection platform in the step 101 refers to collecting well site data and field engineering data for data collection and preparation before training of the machine learning; the data interface constructed is an extraction, cleaning and conversion interface of caliper data, accident complex information, gamma value, well foundation information, logging data, drilling parameter, drilling fluid performance and electric logging information; The pre-processing of the wellbore instability data analysis scanning by traversing all historical drilling data in the step 101 refers to obtaining wellbore instability parameters, blocks and strata prone to wellbore instability data to construct a wellbore elasticity model; the wellbore elasticity model formula is as follows: Radial stress component in cylindrical coordinate system is: = + ( ) (1 ) + ( ) (1+3 ) + (1+3 )sin + (1) Tangential stress component in cylindrical coordinates is: = + ( ) (1+ ) )(1+3 ) (1+3 ) sin + (2) Axial stress component in cylindrical coordinate system is: = [2( ) +4 ] + （3） Shear stress components in cylindrical coordinates are: = ( ) (1 +2 ) + (1 +2 ) （4） Shear stress components in cylindrical coordinates are: = ( + ) (1+ )（5） Shear stress components in cylindrical coordinates are: = ( + ) (1 )（6） wherein: Pp is the pore pressure; Poisson's ratio; is the wellbore permeability coefficient; is the azimuthal angle relative to the x-axis; is the porosity; is the permeability; is the fluid pressure in the wellbore; is the deviation angle; is the azimuthal angle; is the wellbore radius; is the radial coordinate; When the wellbore wall is impermeable, the wellbore wall permeability coefficient = 0. When the wellbore wall is permeable, the wellbore wall permeability coefficient = 1; Let ; the stress state on the well wall is obtained by substituting formula (1), (2) and (3); the stress state on the well wall is: - （7） + （8） ] + （9） （10） （11） （12） 、 、 are the shear stress components in cylindrical coordinates, respectively.

3. The method of claim 1, wherein, The radial stress component according to different conditions in step 102 The tangential stress component The axial stress component The critical drilling fluid density is obtained by substituting the Coulomb-Mohr criterion into the operation respectively, and the shear stress on the corresponding wellbore shear surface is obtained. The Coulomb-Mohr criterion is: when the shear stress on the shear surface is greater than or equal to the inherent shear strength of the rock plus the frictional resistance acting on the shear surface , the well wall is in shear failure, otherwise, the well wall is a stable structure; （13） wherein, is the shear stress on the shear plane; is the inherent shear strength of the rock; is the internal friction coefficient of the rock.

4. The method of claim 1, wherein, The critical drilling fluid density value of the borehole direction is predicted by big data learning and analysis; the prediction analysis generates a block three-pressure profile, which includes fracture pressure prediction evaluation, collapse pressure prediction evaluation and pore pressure prediction evaluation; at the same time, the data of adjacent wells are analyzed to obtain safe drilling parameters and generate a recommended construction scheme.

5. The method of claim 4, wherein, The analysis of the data of adjacent wells to obtain safe drilling parameters refers to analyzing the data of adjacent wells of each stratum to obtain a safe density window and a density early warning window; the analysis of the data of adjacent wells includes drilling fluid system, performance parameter, treatment agent dosage and configuration optimization of the block stratum, and analysis of how to reduce the change of caliper; The data of adjacent wells include performance range, treatment agent dosage, rate of penetration, well foundation information, drilling parameter, accident complexity, caliper expansion rate, gamma value, drilling cycle, cost and electric logging information; The generation of the recommended construction scheme refers to selecting a construction scheme of the block according to different dimensions, and the dimensions include caliper expansion rate, drilling cycle, rate of penetration and drilling cost; the recommended construction scheme indicators include preferred density range, standard density range, electric logging resistance and gamma value.

6. The method of borehole instability analysis based on big data according to claim 5, wherein, Before the data analysis of each stratum adjacent well, the safe density window and the density early warning window are obtained, the data of the adjacent well is selected, the data defects of different adjacent wells are processed by data cleaning, the outliers existing in the processed adjacent well data are processed, the outlier data is identified and processed by clustering algorithm, and then the data is subjected to pattern classification and normalization processing to serve as the input of BP neural network training.

7. The method of claim 6, wherein, According to the construction scheme of the block, the safe density window and the density early warning window, the historical logging data in the block is normalized and trained based on the BP neural network to establish a formation pressure model, and the formation fracture pressure profile model and the formation collapse pressure profile model are established based on the BP neural network training.

8. The method of borehole instability analysis based on big data according to claim 7, characterized in that, The formation fracture pressure profile model refers to the formation fracture pressure under the action of external force on the exposed formation of the wellbore to make it crack or reopen the original crack, and the formation fracture is caused by the over-high drilling fluid density in the well, which makes the effective tangential stress on the rock exceed the tensile strength of the rock; the calculation model mainly includes: Practical fracture pressure prediction model （17） Two horizontal main ground stress prediction models: （18） （19） In the formula: is the formation fracture pressure, MPa; is the rock static Young's modulus, MPa; Poisson's ratio of the rock, dimensionless; Pp is the pore pressure of the formation, MPa; effective stress coefficient, dimensionless; horizontal maximum principal stress, MPa; is the horizontal minimum principal stress, MPa; is the overburden pressure, MPa; Tensile strength of rock, MPa; is the coefficient of the construction stress in two horizontal directions; The formation collapse pressure profile calculation model mainly includes the formation collapse pressure model I and the practical calculation model II of the formation collapse pressure. The formation collapse pressure model I prediction model: （20） In the formula: Equivalent Mud Density for Shear Collapse of Formation, g / cm3; For the nonlinear correction factor of the well wall rock, the general mudstone rock is taken ; is the horizontal maximum principal stress; is the horizontal minimum principal stress; is the formation pore pressure; H is the depth of the reflection layer; K is the correction coefficient, K=0.5~0.80; The practical calculation model II of the formation collapse pressure prediction model: （21） In the formula: , called the regional tectonic stress influence coefficient, is constant within a block.

Images